You are on page 1of 21

PUBLICATIONS

Geochemistry, Geophysics, Geosystems


RESEARCH ARTICLE Calculation of water-bearing primary basalt and estimation of
10.1002/2014GC005329
source mantle conditions beneath arcs: PRIMACALC2 model for
Key Points: WINDOWS
 Model calculation of primary arc
basalt magma Jun-Ichi Kimura1 and Alexey A. Ariskin2
 Thermodynamic fractional
crystallization model 1
Institute for Research on Earth Evolution, Japan Agency for Marine-Earth Science and Technology, Yokosuka, Japan,
 Mantle equilibrium determined using 2
Vernadsky Institute, Russian Academy of Science, Moscow, Russia
petrogenetic grids

Supporting Information: Abstract We present a new method for estimating the composition of water-bearing primary arc basalt
 Readme
and its source mantle conditions. The PRIMACALC2 model uses a thermodynamic fractional crystallization
 Supporting Information S1-S3
model COMAGMAT3.72 and runs with an Excel macro to examine the mantle equilibrium and trace element
calculations of a primary basalt. COMAGMAT3.72 calculates magma fractionation in 0–10 kb at various com-
Correspondence to:
J.-I. Kimura, positions, pressure, oxygen fugacity, and water content, but is only applicable for forward calculations. PRI-
jkimura@jamstec.go.jp MACALC2 first calculates the provisional composition of a primary basalt from an observed magma. The
basalt composition is then calculated by COMAGMAT3.72 for crystallization. Differences in elemental con-
Citation: centrations between observed and the closest-match calculated magmas are then adjusted in the primary
Kimura, J.-I., and A. A. Ariskin (2014),
basalt. Further iteration continues until the calculated magma composition converges with the observed
Calculation of water-bearing primary
basalt and estimation of source mantle magma, resulting in the primary basalt composition. Once the fitting is satisfied, back calculations of trace
conditions beneath arcs: PRIMACALC2 elements are made using stepwise addition of fractionated minerals. Mantle equilibrium of the primary
model for WINDOWS, Geochem.
basalt is tested using the Fo-NiO relationship of olivine in equilibrium with the primary basalt, and thus with
Geophys. Geosyst., 15, 1494–1514,
doi:10.1002/2014GC005329. the source mantle. Source mantle pressure, temperature, and degree of melting are estimated using petro-
genetic grids based on experimental data obtained in anhydrous systems. Mantle melting temperature in a
Received 7 MAR 2014 hydrous system is computed by adjusting T with a parameterization for a water-bearing system. PRIMA-
Accepted 7 APR 2014 CALC2 can be used either in dry or water-bearing arc magmas and is also applicable to mid-ocean ridge
Accepted article online 10 APR 2014 basalts and nonalkalic ocean island basalts.
Published online 29 APR 2014

1. Introduction
Because erupted magmas have been subjected to fractional crystallization and assimilation, the estimation
of chemical composition of primary magmas is important for examining source mantle processes. However,
even if a subordinate role of crustal assimilation is assumed, the reconstruction of fractionation conditions
of a primary magma is a complex procedure because of the various possible crystallization paths within
intracrustal magma chamber systems [Almeev et al., 2013a, 2013b; Grove and Kinzler, 1986].
Olivine maximum fractionation calculations have been used in the estimation of primary basalts [Danyush-
evsky et al., 2000; Herzberg and Asimow, 2008; Herzberg et al., 2007] and have been proven to be useful for
less-fractionated magmas. In many arc lavas, however, the role of fractionation is generally large because of
the existence of a thick sialic crust and because fractionation includes crystals other than olivine [Grove and
Baker, 1984; Tatsumi and Suzuki, 2009].
Once a fractionating magma reaches multiple saturation and begins to move along a cotectic, it is not pos-
sible to trace which side of the cotectic the magma originated from, nor where it joined the cotectic [e.g.,
Ariskin et al., 1993]. Moreover, arc magmas contain varying amounts of water [Plank et al., 2013] and volatiles
[Wallace, 2005], and water in particular affects the liquid line of descent (LLD) [Almeev et al., 2013a, 2013b;
Sisson and Grove, 1993]. Therefore, back calculation models are not able to identify the unique primary
magma composition that fractionated to an observed composition [Danyushevsky and Plecov, 2011], and
only a family of plausible candidates can be identified.
Although difficult to obtain, a reasonable estimate of a primary basalt composition is important, since it ena-
bles better examination of the genetic conditions of the basalt in the source mantle using (1) a petrogenetic
grid for mantle melting (e.g., PRIMELT2 model) [Herzberg and Asimow, 2008; Herzberg et al., 2007; Niu, 1997;

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1494
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

Till et al., 2012]; (2) thermodynamic phase equilibria model calculations (e.g., pMELTS) [Asimow et al., 2004;
Ghiorso et al., 2002; Kuritani et al., 2014b]; or (3) forward mass balance calculations (e.g., Arc Basalt Simulator
(ABS) 3 and 4 models) [Kimura et al., 2009, 2010, 2006].
To achieve the aforementioned goal, we developed an iterative algorithm designed to calculate the compo-
sition of primary basalt in relation to 10 major and 26 trace elements from basalt to basaltic andesite arc
magmas, mid-ocean ridge magmas, and subalkalic ocean island magmas. PRIMACALC version 2 (hereafter
referred to as PRIMACALC2) is an Excel based WINDOWS application software which allows estimates of a
primary basalt using back calculations along the fractional crystallization path of a magma and the source
mantle conditions in a dry to wet system. In this paper, we describe the calculation scheme within PRIMA-
CALC2 and the applications of the model.
The PRIMACALC2 program package, containing a PRIMACALC_2.00(COM3.72).xls Excel spreadsheet and a
COMAGMAT folder containing COMAGMAT3.72 FORTRAN code and relevant files, can be found with a brief
installation guide in supporting information S1. We have confirmed that PRIMACALC2 runs with WINDOWS
XP, 7 and 8, and with Excel 2003, 2007, and 2013 using a 32 bit mode. A Mac PC can also run PRIMACALC2
using WINDOWS OS mode. For 64 bit Excel users, PRIMACALC_2.00(COM3.72)w64.xls is available upon
request to the corresponding author of this study.

2. Calculation Scheme
Figure 1 shows a screen shot of PRIMACALC2 with the calculation flow chart. PRIMACALC2 is built on an
Excel spreadsheet consisting of CONTROL_PANEL, PRIMACALC1, PRIMELT2_MOD, PLOT, COM_Result,
TRACECALC, PRIMACOT, TRIPLOT, CMAS_CALC, PREMELT2_MOD2, Katz, and INPUT1_file Worksheets (see
PRIMACALC_2.00(COM3.72).xls in supporting information S1). In this chapter, we use a step-by-step expla-
nation of the method used by PRIMACALC2 to calculate the composition of a primary basalt.

2.1. Source Data and Treatment


PRIMACALC2 uses 11 major (SiO2, TiO2, Al2O3, FeO, MnO, MgO, CaO, Na2O, K2O, P2O5, and H2O) and 26 trace
elements (Ni, Rb, Ba, Th, U, Nb, Ta, La, Ce, Pb, Pr, Sr, Nd, Sm, Zr, Hf, Eu, Gd, Tb, Dy, Y, Ho, Er, Tm, Yb, and Lu)
contained in natural basalt to basaltic andesite (see (1) in Figure 1). We use the major element, Ni, and H2O
in the composition of natural magmas for 1. calculations of the phase equilibria and melt/mineral composi-
tions in fractional crystallization within the intracrustal magma chamber using COMAGMAT3.72 [Ariskin,
1999; Ariskin and Barmina, 2004; Ariskin et al., 1993] to identify the primary basalt; 2. calculations of the
source mantle conditions in equilibrium with the primary basalt, using petrogenetic grids modified from
PRIMELT2 [Herzberg and Asimow, 2008] and by extending these to wet conditions by parameterization of
Katz et al. [2003]; 3. then, back calculations for all the trace element compositions of the primary basalt are
made, by back-tracking the fractionation sequence with mineral/melt partitioning (see the calculation flow
in Figure 1).
H2O content in natural magmas is normally poorly constrained. PRIMACALC2 explores the effect of H2O
along with the other parameters, P and fO2, in the calculation shown in step 1 (details of the calculations
are shown below).

2.2. First Back Calculation with PRIMACALC1


In order to apply the forward model of COMAGMAT3.72, we need to initially estimate the provisional com-
position of a primary basalt. With the major element composition of a natural magma, the PRIMACALC1
Worksheet calculates the composition of a primary basalt as the initial forecast. To achieve this, we prepared
a typical LLD template for an arc basalt. We chose a basaltic andesite composition from the averaged arc
magma proposed by Tatsumi and Suzuki [2009]. The basaltic andesite has an olivine, pyroxene, and plagio-
clase multiple saturated condition in a shallow magma chamber (Figure 2). We added 10 wt % equilibrated
olivine stepwise, thereby forcing the basaltic andesite to be equilibrated with a depleted mantle. The result-
ant primary basalt composition contained MgO 5 14 wt %, Mg# (Mg/[Mg 1 Fe21]) 5 0.68 (see the red stars
in Figure 2).
We then performed forward fractional crystallization calculations using COMAGMAT3.72 with the condition
of the primary basalt in the magma chamber as P 5 0.3 GPa, fO2 5 QFM12, at various H2O 5 0–2.5 wt %,

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1495
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

PRIMACALC ver.2.00 calculation flow (bottom) with corresponding screen panel (upper)
PRIMACALC_2.00_20131224 CODE INPUT COMAGMAT 3.72 CODE EXAMPLE Powered by COMAGMAT_3.72 and PRIMELT2, programed by JIK OL to CS-MS-A
50
CLYSTALIZATION MODE FRC 1 1: FRC Default Element INPUT NORM COM372 MODEL MODEL OUTPUT A
PRESSURE_CONDITION ISO 1 1: ISO 2: DEC Isobaric/Decomp. Sample VCH-02 VCH-02 Cs/Condi. OL.max P.Bas_1 P.Bas_2 Recalc.
45 Next Cs Sample DIFF(T-C) RND_TXT
SET_SET P_MAX P [kbar] 5 5 3 10 <10kb SiO2 47.63 48.29 48.10 47.78 47.22 47.22 48.29 47.32 SiO2 0.094 47.22
40
SET_MAX P_STEP P [kbar] 11 11 11 0 .1 TiO2 1.15 1.16 1.15 1.08 1.01 1.01 1.16 1.01 TiO2 0.006 1.01 A
SET_NUL/ MIN P [kbar] 0 0 0 2.5 Al2O3 17.60 17.84 17.73 16.49 15.48 15.48 35
17.84 15.54 Al2O3 0.056 15.48 Y2
EDGE
MINERAL_CHEMISTRY 3 Unused FeO 9.09 9.21 9.60 9.37 9.57 9.57 9.21
30 9.37 FeO -0.195 9.57
DiEnPy
LOW-Ca_Px_MODEL 1 1 0: PIG 1: OPX Default MnO 0.17 0.17 0.16 0.17 0.17 0.17 0.17 0.17 MnO 0.004 0.17 COT3
25 Py
NUMBER OF COMPOSITION 1 Unused MgO 8.54 8.66 8.65 11.74 13.80 13.80 8.66 13.80 MgO 0.003 13.80 COT4

OXYGEN_SYSTEM
OXYGEN_BUFFER_CONDI
OPE
CON
1
1
3
1: OPE
1: CON
Default
Default
CaO
Na2O
110.22
2.72
10.36
2.76
10.33
2.73
29.59
2.55
69.00
2.40
89.00
2.40
10.36
20
2.76
15
9.02
2.41
CaO
Na2O
0.015
0.014
9.00
2.40
COT5
COT6
COT7
INIT_OXYGEN_BUFFER NNO 2 1: QFM 2: NNO 3: IW 4: HM K2O 1.31 1.33 1.32 1.23 1.15 1.15 1.33 1.15 K2O 0.002 1.15 TD
LSiPy
LOG_UNIT_SHIFT_OXBUFFER 0 0 P2O5 0.23 0.23 0.23 - 0.20 0.20 0.23
10 0.20 P2O5 0.001 0.20 Pxs
FINAL_OXYGEN_BUFFER (VAR) Unused PRESS for START SUM 99.55 100.00 100.00 100.00 100.00 100.00 0.00 Thermal devide
0.00 Cr2O3 8.991 0.00
5
XSTALLIZATION_INCRIM.[%] 1 Default COMAGMAT Rb 41.64 41.64 - 38.55 35.16 35.16 0.00 0.00 LOI 9.064 0.00
XSTALLIZATION_RANGE_MAX 75 Default Ba 401.69 401.69 - 371.94 339.19 339.19 0
100.00 100.00 Column 17
CALC. 50 60 70 80 90 100
TEMP_CONVERGENCE 0.5 Default Th 2.58 2.58 - 2.39 2.18 2.18 CS Di En(MS)
PHASE_COMP[mol%]
#
0.25
0
Default
Unused 3
PRESS for START
PMC2_CALC. CONDITIONS
U
Nb
0.72
2.68
0.72
2.68
-
-
0.67
2.48
0.61
2.27
0.61
2.27
100
7 AN An
FIT
H2O_IN_PRIMARY_MELT[wt.%] 2 2 <7wt.% 2 Precision (Diff / X) Ta 0.16 0.16 - 0.14 0.13 0.13 90
EDGE
OUTPUT_FILE_NAME 00000001 Default 4 Iteration run (no.#) K 10858 10858 - 10054 9170 9170 TD
80 0GPa
MAJOR_COMP 47.13 1.00 15. Excel PMC2_MANTLE COMP. La 12.09 12.09 - 11.20 10.23 10.23 1GPa
TRACE_COMP (DUMMY) 0.0000 Unused 90.0 PMC1_OL.max Fo Ce 27.14 27.14 - 25.13 22.94 22.94 70 2GPa
3GPa
# 0 Unused 8.0 PRIMC mante FeO* Pb 2.55 2.55 - 2.36 2.15 2.15 60 4GPa
# 0 Unused 38.0 PRIMC mantle MgO Pr 3.65 3.65 - 3.38 3.09 3.09 5GPa
50 2 6GPa
# Null PMC2_OLIVINE D(Ni) Sr 608 608 - 563 514 514 F=0.0
F0.0
DO NOT ALTER CELL POSITIONS!!! L Li, Metz, Bea, Wang Nd 16.43 16.43 - 15.21 13.88 13.88 P=1 F0.1
40 3 0.1
F0.2
Ni (wt.%) IN OLIVINE INCOMPATIBLE TRACE ELEMENT Sm 4.00 4.00 - 3.71 3.38 3.38 0.2
30 4 F0.3
0.7 100 Zr 95.98 95.98 - 88.87 81.16 81.16 F0.4
5 0.3
Hf 2.54 2.54 - 2.35 2.14 2.14 F0.5
20 6
F0.6
0.6 Eu 1.35 1.35 - 1.25 1.14 1.14 0.4

6
Hz
10 0.6 0.5
Gd 4.16 4.16 - 3.85 3.52 3.52 Gt

0.5 Tb 0.67 0.67 - 0.62 0.56 0.56 0


Dy 4.09 4.09 - 3.78 3.46 3.46 0 20 40 60 80 100

Y 23.01 23.01 - 21.31 19.47 19.47 OL QZ


0.4
10 Ho 0.86 0.86 - 0.79 0.72 0.72 SOURCE ADD/SUB P(OAQ) T(MgO) F%(OAQ) T wet (Kaz) F%(Kaz)
0.3 Er 2.48 2.48 - 2.30 2.10 2.10 Peri PB1 - 2.4 1436 10 1356 9
Tm 0.36 0.36 - 0.34 0.31 0.31 Peri PB2 0 2.4 1437 10 1370 9
0.2
PB1
PERI
8 Yb
Lu
2.42
0.36
2.42
0.36
-
-
2.24
0.34
2.05
0.31
2.05
0.31
Pxite PB2 2.5 <= P(CSMSA)
T(C) vs Mineral assemblage/ Composition/ H2O(melt)
0.1 PX Ni (ppm) 118 118 - 409 454 454 100
PCOT
Sample PBAS_1 Ol max. PMC_1 PBAS_2
PB2 Ni(ol)wt% - - 0.22 0.31 0.40 0.42 90 Oliv
0.0 1 Fo(ol)% - - 86.2 90.0 90.7 90.0 80 Plag
80 90 100 Rb Ba Th U Nb Ta K La CePb Pr Sr NdSm Zr Hf Eu GdTb Dy Y Ho Er TmYb Lu Mg# Bas - - 0.63 0.69 0.72 0.72 Cpx
70
MgO vs SiO2 MgO vs Al2O3 MgO vs CaO Fe 2+ /Fe(t) - - 0.83 0.86 0.86 0.86
60 Opx
58
56
25
20
12
10
H2O(wt%)
T C(COM)
-
-
-
-
2.30
1175
-
-
2.00
1312
2.00
- 6 50
40
Mt
Ilm
54 8 TWC(Katz) - - - - 1356 1370 FIT
15 30
52 6 TD C(Herz ) - - - - 1436 1437 Fo(ol)

4-5
10 PGPa - - 0.05 - 2.4 2.4 20
50 4 An(plag)
5 F%(Herz) - - - - 9 9 10
48 2 H2O*10
46 0 0 %Xfrac. - - - 8.0 17 17 0 %Xtal.
0 5 10 15 0 5 10 15 0 5 10 15 1400 1300 1200 1100 1000
MgO PM - - - - - 38

2. First back calculation 4. Forward calculations by 5. Find the closest match


1. Data input with PRIMACALC1 3. Parameter input for COMAGMAT3.72 with result with the observed
of observed and obtain primary COMAGMAT3.72 primary basalt comp- magma, calculate comp-
magma basalt composition
and run
osition ositional difference (DCx)
Subtract DCx from
8. Back calculate trace 7. Adjust T in hydrous mantle 6. Display crystallization primary basalt and Iteration
element composition source with Katz module sequence (panels) replace with new values
and display results in and adjust MgO in the and primary basalt
column and plot source mantle with olivine Fo composition (column) Iteration end

Figure 1. Schematic calculation flow and screenshot of PRIMACALC2. (1) Input natural magma composition (major element, incompatible trace element, and Ni). (2) First back calculation
to primary basalt at MgO 5 14 wt % by COMAGMAT3.72 fractionation template using a typical arc magma of Tatsumi and Suzuki [2009] calculated at 3 kb and a given H2O. (3) Forward
crystallization calculation conditions in COMAGMAT3.72, given H2O, P, and fO2 used with the calculated primary basalt in step 2. (4) Elemental differences found between the closest-
match calculation result and the natural magma, and the adjusted composition of primary basalt. (5) Steps 3 and 4 are automatically repeated until the solution is stabilized and the pri-
mary basalt composition is calculated. (6) Repeat steps 3–5 by checking that the NiO in the olivine within the primary magma satisfies the mantle-olivine Ni-Fo array of mantle equilib-
rium. (7) Calculate mantle P, T, and F in dry conditions using PRIMELT2; source mantle MgO is also obtained iteratively. (8) Recalculate T in the water-bearing mantle iteratively by Katz
et al. [2003] for a given H2O in the primary basalt, with T(dry), P, and F in the mantle. The calculations give chemical compositions of the primary basalt (10 major and 27 trace elements);
conditions of magma fractionation (given initial H2O, P, and fO2): % crystallization, mineral mode, mineral composition, major element, and H2O content in the fractionated magmas
with Fe21/Fe*; and conditions of mantle for the primary basalt (given H2O): P, T, and F (wet) conditions of mantle and MgO (fertility) in the source mantle.

which is believed to represent the typical condition of an arc magma chamber [Tatsumi and Suzuki, 2009].
We tested the versatility of the COMAGMAT3.72 thermodynamic model by comparing the model outputs
to the experimental results of Tatsumi and Suzuki [2009], performed under the same P-T-XH2O conditions,
but with a slightly lower fO2 at QFM12 rather than QFM13 used in Tatsumi and Suzuki [2009]. The calcu-
lated LLDs show excellent agreement with those by experiments as shown in Figure 2. The versatility of
COMAGMAT3.72 has also been shown for use with arc magmas in Eastern Kamchatka and for MORB in the
previous studies [Almeev et al., 2012, 2013a, 2013b, 2007, 2008].
Water saturation in a closed magma chamber is considered in COMAGMAT3.72. When the system is water
saturated, excess H2O is isolated as a separate phase and does not affect phase equilibria. Figure 3a shows
an example of 2.5 wt % water in the primary basalt, saturated at MgO 5 5 wt %. No further increase of
H2O is considered. Water content (Figure 3a), melt compositions (Figure 3b), mineral compositions (Figure
3c), proportions of the fractionated minerals (Figure 3e), and magmatic temperature T, fO2, and Fe21/Fe*
(total Fe as Fe*) (Figure 3f) are calculated for each 0.5 wt % incremental step of MgO in the host melt. The

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1496
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

3.5 16
FeO/MgO FeO
3 14
TH CA
12
2.5
10
2
8
1.5
6
1
4
0.5 Provisional “primary basalt” 2
SiO2 SiO2
0 0
45 50 55 60 65 70 75 45 50 55 60 65 70 75
1.8 16
1.6
TiO2 MgO Numbers are water content
14
COM_0.0
1.4 12 COM_0.5 TS09_0.5
1.2 COM_1.0 TS09_1.0
10
1 COM_1.5 TS09_1.5
8 COM_2.0
0.8
6 COM_2.5 TS09_2.5
0.6
TS09_3.0
0.4 4

0.2 2

0
SiO2 0
SiO2
45 50 55 60 65 70 75 45 50 55 60 65 70 75
20 16
18
Al2O3 CaO
14
16
12
14
12 10

10 8
Lines from COMAGMAT3.72
8 6
dots from experiments by
6 Tatsumi & Suziki (2009)
4
4
2 2

0
SiO2 0
SiO2
45 50 55 60 65 70 75 45 50 55 60 65 70 75

Figure 2. Comparisons of liquid lines of descent of a typical arc basalt with different water content between COMAGMAT3.72 model calcu-
lations and experimental results by Tatsumi and Suzuki [2009]. Numbers with COM and TS09 show H2O contents in the COMAGMAT3.72
and experimental systems, respectively. TH (tholeiitic) and CA (cacl-alkaline) boundary from Miyashiro [1974].

PRIMACALC1 Worksheet includes the calculated results of LLD with initial values of H2O 5 0–2.5 wt % in
3.0–14.5 wt % MgO. All the data are stored in lookup tables in the cell area $AL$788–$DC$1412 for minerals,
and in $CD$1–$DE$784 for melts in the PRIMACALC1 Worksheet. With the data table, back calculations of a
fractionated magma in the range 3.0–14.0 wt % MgO are available.
The first back calculation requires H2O wt % to be present in the primary magma because the initial water
content largely affects the LLDs, and thus the fractionation sequence (Figures 2 and 3a) [Grove and Baker,
1984; Tatsumi and Suzuki, 2009]. Cell $C$20 in the CONTROL_PANEL Worksheet defines initial H2O as user
input. The model LLD templates are for the H2O 5 0–2.5 wt % range, and any excess H2O is then forced to
be set at 2.5 wt % in PRIMACALC1. Note that PRIMACALC2 accepts up to 7 wt % water in the starting com-
position of a primary basalt, but the first back calculations only use values up to 2.5 wt %. More water is
valid in the second step iteration run, and the first step assumption is compensated after the second run
(see section 2.3).
With the given H2O, PRIMACALC1 chooses the designated LLD path and adds equilibrated minerals at a 0.5
wt % MgO step from the natural magma composition, until the bulk rock MgO reaches 14 wt %. Because of

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1497
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

PRIMARY MAGMA CALCULATOR ver. 1.00 [PRIMACALC_1: 2012/08/01]


[MgO vs. H2O RELATIONSHIP FOR SYSTEM XH2O DETERMINATION]
H2O 2.5 wt.% in sample (if not available, use XH2O = 1.5 wt.% below)
MgO 8.5 wt.% in model sample
XP 3 kbar FC system pressure (fix) and QFM+2 oxybuffer
XH2O 0.0 primary XH2O: read from diagram below and input [0-2.5 wt.%]

[DISPLAY MODEL PARAMETERS AND VARIABLES USED FOR CALCULATIONS] [PRIMACALC_1]


Element 1 SiO2 [1:SiO2, 2:TiO2, 3:Al2O3, 4:FeO*, 5:MnO, 6:CaO, 7:Na2O, 8:K2O]
Mode 3 Cpx [1:Ol, 2:Plag, 3:Cpx, 4:Opx, 5:Mt, 6:Melt]
Mineral 3 En(Cpx) [1:Fo,2:An,3:En(Cpx),4:Fs(Cpx),5:Wo(Cpx),6:En(Opx),7:Fs(Opx),8:Wo(Opx),9:Usp]
Variables 1 T(C) [1:T(C), 2:fO2, 3:Fe2+/Fe*]

 &0 
H2O &0
6DPSOH
&0
 &0 3ULPDU\\\%DV
%DV
Water saturation &0
&0 2/B0D[

&0
&0 
&0

&0
&0
&0

&0
&0 Sample/Primitive mantle
&0 

a d

*G
6P

7P
+I
3U
6U

=U

(U
'\
5E
%D
7K

1E
7D

/D
&H
3E

1G

(X

7E

+R

<E
/X
.

<
8
&0
  MgO  
 
0W
SiO2 Cpx mode (%) 2OLY
6L2  3ODJ

([S &S[
([S  2S[
 ([S 0HOW
 ([S
([S ([S
 ([S ([S

([S
([S
 
  MgO   b   MgO   e
 
En(Cpx) Temperature (C)
 (Q &S[
([S
 
([S
 ([S 7 &
([S
 
([S


 
  MgO   c   MgO   f

Figure 3. View of the PRIMACALC1 Worksheet. (a) H2O versus MgO in melt, (b) SiO2 versus MgO in melt, (c) MgO(melt) versus En composition of clinopyroxene, (d) trace element composi-
tions, (e) mineral mode, and (f) magma temperature for both experimental [Tatsumi and Suzuki, 2009] (colored circles) and COMAGMAT3.72 results (lines).

the existing compositional difference between the model basalt and natural magma, it is possible that the
final MgO may not perfectly match MgO 5 14 wt % (see cell area $D$41–$AA$53 in the PRIMACALC1 Work-
sheet). Trace element compositions (Figure 3d) are also back calculated using the addition of minerals in
the cell area $D$55–$AA$80 of PRIMACALC1 Worksheet. For trace element partition coefficients (Ds)
between melts and minerals (see section 2.7). The first estimate of a primary basalt includes the trace ele-
ment composition, but this is only for reference and not used for further calculations. Only the major ele-
ment composition is used for further calculation steps. The PRIMACALC1 results are found in cell area
$D$43–$D$80 in the PRIMACALC1 Worksheet. The results also plot on the INCOMPATIBLE TRACE ELEMENT
panel in the CONTROL_PANEL Worksheet as PMC_1, along with other model results (Figure 1).

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1498
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

2.3. Problems in Back Calculations


In principle, the primary basalt composition given by PRIMACALC1 represents a primary basalt before frac-
tionation in a magma chamber with a fixed condition of 0.3 GPa, fO2 5 QFM12, and a given H2O following
a model template. There are two major uncertainties, one from the real primary melt composition that is
determined by which side of the cotectic the liquidus plain that the primary basalt first touches, and the
other from the values of P, T, and fO2, and H2O present in the real fractionation conditions that then deter-
mine the true cotectic.
In theory, the first uncertainty is not possible to solve using back calculations by any thermodynamic model
unless the primary melt composition is given. Therefore, PRIMACALC1 uses an ‘‘average arc magma LLD’’
template to track-back to the primary basalt. The template includes tholeiite to calc-alkaline series fractiona-
tion trends by controlling the values of H2O [Tatsumi and Suzuki, 2009] (Figure 2). The near-dry solution can
then be used for fractionated mid-ocean ridge basalt (MORB) or sub to mildly alkaline (tholeiitic) ocean
island basalt (OIB), as multiple saturation compositions are more or less similar to those of dry tholeiitic arc
basalts in terms of the major element contents. Good reproducibility of the phase equilibria in MORB has
been confirmed with COMAGMAT3.72 [e.g., Almeev et al., 2008].
It is noted that our template LLD approach is not the only way to deal with the first uncertainty and any vec-
tor (trend) extrapolation analyses of the fractionation path is able to do the same. Our PRIMACALC1 model
uses the averaged trend deduced from arc basalt data, but also uses back-ups using experimental and ther-
modynamic considerations for a wide range of magma compositions ranging from tholeiitic to calc-alkaline
(Figure 2). Note, however, that any sort of template/trend calculation does not warrant mantle equilibrium
of an estimated primary magma. This problem will be discussed below in section 2.5.
The second uncertainly, a variable cotectic, is basically related to the control of P, T, fO2, and H2O in a
magma chamber. However, although water content data are available from olivine melt inclusions, H2O in
natural magma is not well constrained due to degassing [Plank et al., 2013; Wallace, 2005]. Therefore, the
first step of the H2O assumption in PRIMACALC1 already contains large errors. The assumed magma cham-
ber pressure at P 5 0.3 GPa and fO2 5 QFM12 may be a good estimate for the condition of arc magma frac-
tionation [Tatsumi and Suzuki, 2009], but it also contains large unknowns in relation to any particular
magma of interest. How PRIMACALC2 works on this problem is given in section 2.4.

2.4. COMAGMAT3.72 Forward Calculations and Iterations


To minimize the second uncertainty, we simulate the fractionation process of the provisional primary
magma using COMAGMAT3.72 forward calculations. We use P, T, fO2, and H2O conditions as fitting parame-
ters. Among these, fO2 of QFM11 (or NNO) may be appropriate for many arc magmas [Kelley and Cottrell,
2012], alternatively, QFM (NNO-1) buffer may be appropriate for MORB or OIB sources [Cottrell and Kelley,
2011]. Versatility of the assumptions with the PRIMACALC2 model will be examined in section 4.1. Magmatic
T is calculated by the COMAGMAT3.72 thermodynamic model when the melt composition, P, fO2, and H2O
are given. Therefore, P and H2O in the magma chamber are the key variables that can be explored using
COMAGMAT3.72.
Inverse calculations along a plagioclase-olivine LLD can be achieved using other petrological models, such
as Petrolog3 [Danyushevsky and Plecov, 2011]. However, there are difficulties in using the inverse model
along a multiply saturated LLD, because Petrolog3 calculations usually halt at around the first multiphase
saturation. Alternatively, iterative forward calculations are possible using the alphaMELTS thermodynamic
model, which uses an algorism (1) after being given an initial estimate of the parental melt composition
and then (2) varies the melt composition until isobaric forward fractionation yields a specified target. This is
known as the Amoeba routine [Cooper et al., 2004; P. Antoshechkina and P. D. Asimow, AlphaMELTS Soft-
ware Manual, 2013, http://magmasource.caltech.edu/alphamelts/1/alphamelts_manual.pdf]. The PRIMA-
CALC2 model uses the same approach but employs the COMAGMAT3.72 thermodynamic model rather
than MELTS.
For the iteration, COMAGMAT3.72 first calculates the fractional crystallization sequence from the provisional
primary basalt at a given fO2, P, and H2O. MgO in the calculated magmas and that in the natural magma are
then compared, and the model magma which has the closest MgO content with the natural magma is cho-
sen. Differences in the other major elements between the model and the natural magma are calculated and

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1499
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

MgO vs SiO2 MgO vs Al2O3 MgO vs CaO


58 25 12
56
a Natural magma
(open circle) 10
20
54 8
15
52 Model magma 6
10
50 “Primary basalt” 4
48 5 2
46 0 0
0 5 10 15 0 5 10 15 0 5 10 15

100 0.7
b NiO (wt%) c
90 Oliv
0.6
80 Plag 2nd stage
olivine in
70 Cpx 0.5 pyroxenite
melting

60 Opx
0.4
Mt Mantle
50 olivine
Ilm 0.3 Ni array
40
FIT
30 0.2
Fo(ol)
PB1
20 PERI
An(pl)
0.1 PX
10 PCOT
H2O*10
T (C) Fo PB2
0 %Xtal. 0.0
1400 1300 1200 1100 1000 80 90 100

Figure 4. Calculation results in CONTROL_PANEL Worksheet of PRIMACALC2. (a) Forward fitting calculation results of COMAGMAT3.72
(blue dots) and the best fit result for the calculated primary basalt and magma compositions (red squares), showing a good fit to the target
magma composition (red circle). (b) Fractional crystallization sequence calculated by COMAGMAT3.72 showing mineral assemblage, min-
eral composition, H2O in the melt, and % crystallization (Ol: olivine, Plag: plagioclase, Cpx: clinopyroxene, Opx: orthopyroxne, Mt: magne-
tite, Ilm: ilmenite). (c) Calculated NiO content (wt %) and Fo in olivines in fractionation (yellow circles) and in the primary basalt (orange
square), showing a good fit with the olivine-mantle Ni array (orange parallelogram) [Takahashi, 1986]. Second stage clinopyroxene field
(green square) [Herzberg, 2011] is also shown for a pyroxenite source.

then added to or subtracted from the provisional primary basalt. The new basalt composition is then used
as the new input for the following iteration cycles. These iterations were continued until the natural and the
model magma compositions converge (Figure 4a).
In this iteration, the MgO content of the natural magma (therefore closest-match calculated magma) held
constant, and all the major element composition (including MgO) varied in the primitive basalt. In order to
obtain a better convergence, the subtracted/added element concentrations from/to the primary basalt are
given using half of the difference between the natural and the chosen calculated magmas. This simple
approach prevents an overshoot of the iteration and provides results that are precise enough (e.g., <0.37%
relative difference (%R.D.) in SiO2 and %R.D. < 5% in MgO; see major %R.D. in the CONTROL_PANEL). The
precision of iteration (given by [difference]/X, where X 5 2 defining added/subtracted amounts to 50% of
the R.Ds) and the number of the iteration are set in cells $E$20–$E$21 in the CONTROL_PANEL Worksheet.
Four iteration runs give sufficient convergence when X 5 2.
Figure 4b shows a fractional crystallization sequence including (1) mineral phases (olivine, plagioclase, clino-
pyroxene, orthopyroxene, magnetite, and ilmenite), (2) olivine Fo, (3) plagioclase An, (4) H2O content in
magma, and (5) the fractionated mass of the solid phases plotted against the magmatic temperature (T);
they were all calculated by the last COMAGMAT3.72 iteration run. Two red vertical lines delimit the range of
fractionation between the primary basalt and the target natural magma. This interval is used for back addi-
tion calculations in PRIMACALC2. The diagram is useful for observing which mineral phase(s) is incorporated
in the back calculations, defining whether or not H2O is saturated in the fractionated magma, and defining
the magma temperature in the shallow magma chamber.

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1500
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

A full data set of the calculated major element compositions and the mineral mode and compositions are
shown in the COM_Result and the PLOT Worksheets. The calculated major element composition in the
closest-match model magma is given in COM372 Cs/Condi. in cell area $L$2–$L$13 in the CONTROL_PANEL.
Residues between the model and observed magmas are found in $L$216–$L$25. The compositions of
Ni(olivine), Fo(olivine), Mg#(melt), Fe21/Fe*(melt), H2O(melt), and the conditions of T(melt) and P(melt) in the magma
chamber are also in the same column $L$42–$L$50. These provide information related to the equilibrium of
the observed natural magma in the magma chamber conditions. The data are refreshed during iterations
and finally becomes valid after iteration converges.

2.5. Identification of Mantle Equilibrium Using Ni in Olivine


The purpose of PRIMACALC2 is to estimate a reasonable primary basalt composition for an arc magma. We
therefore need to consider equilibration of the calculated primary basalt with the mantle peridotite, and
there are a number of possible ways to estimate this.
1. The use of Fe12/Mg ratios of magma, which should be in equilibrium with the mantle olivine (Fo86–90) by
Kd 5 (Fe12/Mg)ol/Fe12/Mg)melt 5 0.3 [Roeder and Emslie, 1970]. If the mantle olivine is Fo86–90, primary mag-
mas must be Mg# (Mg/Mg 1 Fe12) 5 0.68–0.75 or have a weight ratio of FeO/MgO 5 0.4–0.7 in this model.
2. More precisely, the use of partial melting (F)-dependent FeO21/MgO partitioning between basalt and
peridotite [Herzberg and O’Hara, 2002], in which estimated equilibrium is also dependent on the FeO and
MgO contents of the source peridotite.
3. The use of NiO wt % of olivine (0.3–0.42 wt % for Fo85–94) in equilibrium with the mantle peridotite
[Herzberg, 2011; Takahashi, 1986].
In PRIMACALC2, we use approach 3 because this model is independent from FeO/MgO in the source man-
tle. The compositional field of mantle olivine is shown by the orange parallelogram in the diagram showing
NiO (wt %) versus Fo in olivine in Figures 1 (6) and 4c. The NiO value in this field is regarded as the target
NiO content.
The partition coefficient of Ni between olivine and the melt D(Ni) is critical for the back calculation of NiO in
olivine. We examined various models ranging from T-independent/composition-dependent parameteriza-
tion of [Beattie et al., 1991] supported by [Jones, 1984] and [Herzberg et al., 2013] to the T and composition-
dependent parameterizations by Li and Ripley [2010], Matzen et al. [2013], and Wang and Gaetani [2008]. We
explored the relatively low T range (typically 1000–1400 C for fractional crystallization) with the Fe21-Mg of
melt, both calculated by the COMAGMAT3.72 model.
Beattie et al. [1991] model always shows a high D(Ni) in olivine, and therefore the back addition of olivine
does not satisfy mantle equilibrium in the PRIMACALC2 model. This could be related to the effect of the T
dependence of (Ni) [Matzen et al., 2013]. Wang and Gaetani [2008] model also shows a high D(Ni) in a low T
range (e.g., <1200 C), although T dependence is taken into account in the model. However, the D(Ni) of Li
and Ripley [2010] and Matzen et al. [2013] always delivers reasonable fits in the PRIMACALC2 model for a
wide range of target magma composition, and we therefore use the Li and Ripley [2010] model in this paper,
while providing the choice of using the D(Ni) model. Cell $E$27 in the CONTROL_PANEL Worksheet allows
the use of the D(Ni) models by typing; L: Li and Ripley [2010], M: Matzen et al. [2013], B: Beattie et al. [1991],
and W: Wang and Gaetani [2008] for the designated model.

2.6. Effect of P and H2O to D(Ni) in Fractional Crystallization


As shown in section 2.4, H2O and P are the key variables in the COMAGMAT3.72 forward calculations. These
parameters control not only the fractionation mineralogy by altering the liquidus temperatures of minerals
[Almeev et al., 2012, 2013a, 2007; Ariskin, 1999; Ariskin and Barmina, 2004] but also the magmatic tempera-
ture in crystallization [Almeev et al., 2007], which does affect D(Ni) because of the T and composition
dependence of Ni partitioning between olivine and the melt. This eventually results in a different NiO(olivine)
in the primary basalt.
Figure 5 shows examples of the fractionation sequences of a primary basalt at different values of P (0.3 and
1.0 GPa) and at a different initial H2O content (0 and 2 wt %). An increase in P increases the T of olivine crys-
tallization due to the increase in the liquidus temperature of olivine [Katz et al., 2003] (Figure 5a), and in
contrast an increase in H2O decreases the liquidus temperature of olivine [Almeev et al., 2007; Ariskin and

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1501
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

1450 Barmina, 2004; Falloon and Danyushevsky, 2000]


1400 T(C) (Figure 5a). Due to the change in the crystalliza-
tion temperature, the Fe-Mg partitioning between
1350 olivine and the melt also changes [Roeder and
1300 Emslie, 1970]. In PRIMACALC2, the above changes
1250 are in accordance with the thermodynamic model
of COMAGMAT3.72 [Ariskin et al., 1993] (Figure
1200
5b). The NiO content of olivine at equilibrium is
1150 low, either in relation to a lower P or to the higher
a H2O content of a magma, and this is reflected in
1100
94 the T and compositional dependence of D(Ni) in
Fo(olivine) the Li and Ripley [2010] model (Figure 5c). Note
92 again that both the models of Li and Ripley [2010]
and Matzen et al. [2013] are T and composition-
dependent, and we consider that these models
90
are internally consistent with COMAGMAT3.72 for
the temperature range of <1400 C in the PRIMA-
*3DGU\
88 *3D+2 CALC2 model.
*3DGU\
*3D+2 b By adjusting the P and H2O variables, we can
86 establish a reasonable equilibrium of the primary
0.5
NiO(olivine) basalt with the source mantle in terms of the Fo-
0.4 NiO contents of olivine (see Figures 4b and 5c).
rray
olivine a These calculations are achieved by repeating iter-
Mantle
0.3 ations with different P and H2O values given in
cells $C$4 and $C$20 in the CONTROL_PANEL. As
0.2 it would then be necessary to introduce further
constraints of P and H2O because they give an
0.1 opposite effect we have not considered any auto-
MgO(melt) c mated searches for these values, and this prob-
0 lem will be discussed in section 3 for application
8 9 10 11 12 13 14 15 to PRIMACALC2.
Figure 5. (a) Effects of olivine crystallization temperature by differ-
ent pressures and H2O; (b) Fo composition of equilibrated olivine;
2.7. Back Calculations for Trace Elements
and (c) NiO contents in olivine controlled by both temperature and
composition. Using the optimal crystallization sequence
obtained by the iteration calculations shown
above, back calculations of trace elements in the
primary basalt can be made using the stepwise addition of equilibrated minerals (e.g., magnetite, orthopyr-
oxene, clinopyroxene, plagioclase, and olivine). Both the PRIMACALC1 Worksheet for the first provisional pri-
mary basalt and the TRACECALC Worksheet for the final primary basalt use the same algorithm, which is
described as below.
Trace element back calculations, including NiO in olivine (see details in section 2.6), are available by using
the melt and mineral compositions calculated by COMAGMAT3.72 in the TRACECALC Worksheet. The parti-
tion coefficients of fractionated minerals change with the P, T, and mineral/melt composition, and we there-
fore use the compositional, P, and T-dependent partition coefficients whenever applicable. The partition
coefficients are derived from the following sources: the empirical MgO(melt)-dependent parameterization of
Bedard [2005] for olivine; the empirical T and An-dependent parameterization by Bindeman [2007] and
Bindeman et al. [1998] for plagioclase; the T, P, and melt/mineral composition-dependent lattice strain
model of Wood and Blundy [1997] for rare earth elements (REEs) and the experiment-based partitioning of
Pilet et al. [2011] for other incompatible elements of clinopyroxene; the orthopyroxene/clinopyroxene parti-
tioning derived from compilations of Pilet et al. [2011] and the clinopyroxene partitioning for orthopyrox-
ene; and the compilations by Pilet et al. [2011] and Rollinson [1993] for magnetite.
The back calculation result is shown as P.Bas_1 in cell area $N$3–$N$52 in the CONTROL_PANEL Worksheet.
The trace element composition also plots graphically on incompatible trace element and on the Ni (wt %)

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1502
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

IN OLIVINE panels (Figure 1 and supporting information S1). These panels show a primitive mantle (PM)
[Sun and McDonough, 1989] normalized multielement plots of the primary magma and MgO versus NiO
plots of the olivine composition, respectively.
In addition to the PRIMACALC2 results, the olivine maximum fractionation model of PRIMELT2 [Herzberg
and Asimow, 2008; Herzberg et al., 2007] is also applied to the natural magma by using the P, Fe21/Fe*, and
olivine Fo composition for the primary basalt from PRIMACALC2 (see the PRIMELT2_MOD Worksheet). This
also calculates the major and trace element compositions OL.max in the primary basalt for comparison with
those by PRIMACALC1 and P.BAS_1 by PRIMACALC2. The PRIMELT2 results are in cell area $M$3–$M$53
OL.max in the CONTROL_PANEL Worksheet. Ni partitioning in olivine is calculated by Li and Ripley [2010] for
PRIMACALC2, whereas the Beattie et al. [1991] model is used in the PRIMELT2_MOD Worksheet. The differ-
ences between the two models are shown as NiO in olivine calculated for the primary basalts of OL.max
and P.Bas_1.

2.8. Source Mantle Conditions (P-F-T)


The PRIMELT2_MOD2 Worksheet calculates the P, T, and F conditions of the mantle for the P.Bas_1 compo-
sition using PRIMELT2. P and F are both estimated by the Ol-An-Qz (projected from Di) CMAS projection
whereas T is estimated using the T-MgO(‘‘primary basalt’’) relationship [Herzberg et al., 2007]. The PRIMACOT
Worksheet shows certain CMAS plots, including Ol-CaTs-Qz, MgO-CaO, and FeO21 versus MgO, which are
key to the PRIMELT2 model. The TRIPLOT and CMAS_CALC Worksheets provide background calculations.
Only Cs-Ms-A (Ol) and Ol-An-Qz (Di) projections are shown in the CONTROL_PANEL.
Fractional and batch melting are two extreme cases of mantle melting, but fractional melting is more plausi-
ble mechanism. It is, however, hard to introduce this into the model, as CMAS petrogenetic grids are formed
based on the experimental results, which represent batch melting. Herzberg and O’Hara [2002] and Herzberg
and Asimow [2008] have introduced an accumulated fractional melt AFM (Accumulated Fractional Melt)
model in PRIMELT2 for F-Mg/Fe(melt) and T-MgO(melt) relations, but PRIMACALC2 uses the batch model of
CMAS.
The PRIMELT2 model [Herzberg and Asimow, 2008; Herzberg et al., 2007] is built on anhydrous experimental
data. We plotted the experimental results, including the water-bearing system from LEPR database [Hirsch-
mann et al., 2008] onto the Ol-An-Qz (Di) projection, and found that the P relationship is maintained (see
Figure 6a). There are not many water-bearing experiments that report F information in existence. We exam-
ined the experiments of Hirose and Kawamto [1995] at 1 GPa (results shown in Figures 6b and 6c) and found
that the estimated F by Ol-An-Qz (Di) projection reproduces the reported F well, and a near one-to-one cor-
relation line with a negligibly high zero intercept (y 5 0.9769x 1 0.0359; Figure 6b) is plotted.
The peridotite melting in water-bearing systems is known to provide higher F values at given T [Katz et al.,
2003]. In fact, T estimated by T-MgO(melt) by PRIMELT2 shows an erroneously 10% higher T than those in
the experiments by Hirose and Kawamto [1995] (y 5 1.096x; Figure 6c). Therefore, we considered that the
mantle T should be recalculated with an estimated P and F and the presence of H2O in the primary basalt,
and this is available using the parameterization of Katz et al. [2003]. The P and F estimates come from the
Ol-An-Qz (Di) projection, and H2O is initially given for the PRIMACALC2 calculations. Thus, the T during the
‘‘wet’’ melting at the postulated H2O in the primary basalt was iteratively calculated using the Katz Work-
sheet. This was achieved by altering T in the Katz Worksheet, which begins with T(dry), and then finding
when H2O in the primary basalt becomes unity to a given value of H2O. Figure 6d shows a graphical exam-
ple of this calculation (starting with 1340 C–dry–F 10%, to find 1240 C–3 wt % H2O–F 10%).
The calculations are associated with the VBA macro iteration calculations, and the results T(dry), P(Ol-An-
Qz), and F(Ol-An-Qz) from PRIMELT2, with T(wet), and F(wet) from the Katz Worksheet, are shown in the cell
area $Q$37–$W$37 in the CONTROL_PANEL Worksheet in the P.Bas_1 column. Since F is calculated inde-
pendently in the Katz et al. [2003] model, it is also given in $W$37 for comparison. We usually obtain coher-
ent F between the dry and the wet models that are within a 1% error, providing the basis of reliable wet
mantle T estimates in PRIMACALC2.

2.9. Source Mantle Fertility (MgO)


The final exploration with PRIMACALC2 is the fertility of the source peridotite. PRIMACALC2 provides the pri-
mary basalt composition in equilibrium with the source mantle (see the P.Bas_1 column in the

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1503
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

a P estimate by Ol-An-Qz (Di) projection d Calculation for T in hydrous mantle melting


An from P-T-XH2O for basalt melt
100
Exp. 1.5 GPa Isobaric plane
0GPa 90 0.20
1GPa

Isotherm
2GPa 80
3GPa
70 0.15 1340 C O

Melt fraction (F)


4GPa
2
F = 0.1
5GPa 60 0% F dry
3 O
6GPa 1240 C
10
F0.0 50 F = 0.1
0.10
F0.1 4
1 GPa 3% H2O
F0.2 40 20
30
F0.3
30 5 0.05
F0.4
40
F0.5 20 6
F0.6
Harz- 10
0.00
burgite 0 1 2 3 4 5
Garnet 0
0 20 40 60 80 100
Water in the melt (wt%)
Ol Qz
Hydrous exp. 7.0 6.0 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 (GPa)

b F estimate by Ol-An-Qz (Di) projection for c T estimate by MgO for hydrous magma
hydrous magma
0.5 1600
F(Ol-An-Qz (Di)) y = 0.9769x + 0.0359 y = 1.096x
0.4 1500
F (PRIMELT2)

T (PRIMELT2)

1400
0.3
1300
0.2
1200
0.1 1100
F (experiment) T (experiment)
0.0 1000
0.0 0.1 0.2 0.3 0.4 0.5 1000 1100 1200 1300 1400 1500 1600

Figure 6. (a) Petrogenetic CMAS grid used for P and F estimates; (b) comparison of experimental F and calculated F values hydrous peridotite system; (c) comparison of experimental T
and T from thermometry for hydrous peridotite melts; and (d) the conversion scheme of T(dry) to T(wet) mantle melting. In Figure 6a, Ol-An-Qz (projected from Di) plot is originally from
Herzberg and O’Hara [2002] modified by the authors using additional experimental data from REPL database [Hirschmann et al., 2008], including dry and hydrous (red circles) experi-
ments. Correlations between experimental F values and those calculated by Ol-An-Qz (Di) projection for the hydrous experiments of Hirose and Kawamto [1995] are shown in Figure 6b.
Figure 6c shows T estimates by MgO(melt) by Herzberg and O’Hara [2002] for the same hydrous peridotite melts showing 10% higher T (C) in calculations. Figure 6d shows a schematic
example of T(dry) to T(wet) conversion. T(wet) 5 1240 C is obtained when T(dry) 5 1340 C, F 5 0.1, H2O 5 3 wt %, and P 5 1.5 GPa are known (figure modified from Katz et al. [2003]).

CONTROL_PANEL Worksheet). The estimated P.Bas_1 composition always displays slight discrepancies from
the equilibrium parameters given by the Fe21-Mg relationship at the given P by Ol-An-Qz (Di) (see section
2.8) in the PRIMELT2 algorithm [Herzberg et al., 2007] (see also FeO21 versus MgO plots in the PRIMACOT
Worksheet). PRIMELT2 cautions addition/subtraction of olivine to/from P.Bas_1 and the amounts shown
(see cell $C$11 in PREMELT2_MOD2 Worksheet). To accommodate this problem, PRIMACALC2 has an addi-
tional function that can modify the source mantle FeO and MgO compositions to take the mantle fertility
into account for equilibration [Herzberg and Asimow, 2008; Herzberg et al., 2007]. These mantle compositions
are given in cells $E$24 and $E$25 in the CONTROL_PANEL Worksheet.
FeO in the mantle peridotite is a canonical value and is fairly uniform at 8 wt %, whereas MgO varies from
25 to 50 wt % [Bodinier and Godard, 2003]. We therefore set FeO 5 8 wt % and change values of MgO to
make the equilibration conditions from PRIMELT2 consistent with the primary basalt estimated by PRIMA-
CALC2 P.Bas_1. The automated calculation forces a discrepancy in the Fe21O-MgO equilibria (shown in the
[FeO(wo_Fe31)-MgO] panel in the PRIMACOT Worksheet) during iteration of the VBA run. The renewed
MgO appears in cell $E$25 in the CONTROL_PANEL. The resultant MgO is usually between 25 and 50 wt %,
possibly indicating source heterogeneity. The recalculated primary basalt composition using the new

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1504
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

mantle MgO content with a further addition/subtraction of olivine with PRIMELT2 to create a new equilib-
rium is shown by P.Bas_2 in cell area $O$3–$O$53. The original olivine NiO(olivine) and Fo compositions cal-
culated primarily by PRIMELT2 are shown in cells $O$42 and $O$43 for comparison.
For this final calculation, Fo and NiO(olivine) are recalculated using the model of Beattie et al. [1991] and Jones
[1984] of Ni partitioning with the P.Bas_2 composition. The Fo versus NiO composition of the finally equilibrated
olivine for P.Bas_2 and the fractionation-accumulation trajectories of olivine by PRIMELT2 are in the Ni (wt %) in
olivine panel (see the big orange square and the small gray crosses, respectively, in Figure 4c). In many cases,
the Fo and NiO in olivine from the PRIMACALC2 (COMAGMAT3.72 and Li and Ripley [2010] Ni partitioning) in
P.Bas_1 and the PRIMELT2 P.Bas_2 results agree with each other (see the example in Figure 4c).
However, a large discrepancy occurs when, for example, Hawaiian basalts are calculated. The trials show
higher values of NiO (plot in the field of the ‘‘second stage olivine in pyroxenite melting’’ in Figure 4b) by
PRIMELT2, even though the NiO in olivine from PRIMACALC2 falls into the Fo-NiO mantle array. The high
NiO in olivine reproduces the pyroxenite source model proposed by Herzberg [2011]. However, any prob-
lems associated with this are not discussed here because the results for the low-K tholeiitic to medium-K
calc-alkaline arc magmas examined here did not show the problem. A caveat here is that P.Bas_2 uses Fo
and NiO(olivine) from the PRIMELT2 [Beattie et al., 1991; Jones, 1984] model. If a large discrepancy is found in
relation to the results of P.Bas_1 and P.Bas_2 in PRIMALCALC2, the user should note which values are used.
We use the P.Bas_2 model results throughout this paper. Validity of the estimation of source mantle fertility
is examined in section 4.

3. Calculations of Arc Magmas With PRIMACALC2


In this section, we give a brief outline of how to operate PRIMACALC2. First, users should download support-
ing information S1. PRIMACALC2.zip folder, unzip this on the desktop, then copy the whole COMAGMAT
folder to the C: root directory of the hard disk. Next, copy PRIMACALC_2.00(COM3.72).xls on the desktop
and double click it, thereby opening PRIMACALC2 with Excel which is ready for use. Please note that any
changes made to the file names or cell formats will destroy the program links. We therefore recommend
keeping the original (zip) files and not changing the file names.
In PRIMACALC2, the cells color-coded in yellow with red bold text are the input areas (CONTROL_PANEL
Worksheet; Figure 1). Those in blue with red bold text are the output regions. It is possible to alter these
input values manually. Please note that any changes in other cells will cause fatal failure of the data links
and calculations. The same applies to other Worksheets, and thus it is advisable not to alter any cells. The
output data can be copied using the COPY function of Excel. Copies via Excel do not alter the cells, so that
it is possible to copy any part of PRIMACALC2 and export these numbers using the PASTE-SPECIAL function.

3.1. Data Input


Data input, calculations, and data output are managed in cell area $J$53–$O$53 in the CONTROL_PANEL.
The geochemical data of an observed magma should include 10 major elements, 26 incompatible trace ele-
ments (including K in ppm), and Ni (see (1) in Figure 1).

3.2. Back Calculation and Iteration Run


To run the VBA macros, including the COMAGMAT3.72 shell program, the predefined calculation conditions
of COMAGMAT3.72 should be set in cell area $C$2–$C$26 in the CONTROL_PANEL (see (3) in Figure 1). In
this paper, we use pressure mode 1: [ISO]baric (value is input in cell $3$C), with a magma chamber pressure
of 0.1–10 kb (cell $C$4), and various oxygen buffers in the magma chamber 1: QFM or 2: NNO (cell $C$12)
plus/minus X log unit (cell $C$13), and a water content in the primary basalt of 0–7 wt % (cell $C$20). The
option to calculate a polybaric (decompression) crystallization sequence with a constantly decreasing pres-
sure increment is also available [Almeev et al., 2013a, 2013b; Ariskin and Barmina, 2004], but we do not use it
in this paper for the basic model test. Detailed documentation of the COMAGMAT3.72 code settings are in
the INPUT1_file Worksheet. Note that some function codes (e.g., EQU: equilibrium crystallization) do not
work with PRIMACALC2. Changeable codes are shown by yellow cells in the CONTROL_PANEL; otherwise,
codes should be set by a default function.
Once the assumed magma chamber conditions are set, it is possible to run the COMAGMAT3.72 FORTRAN
shell program by pressing the COMAGMAT CALC. button at cell $F$16. The PRIMACALC2 iteration

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1505
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

125˚ 130˚ 135˚ 140˚ 145˚ 150˚


rc
ile a
Kur
20 45 Osore

20 37 Hakkoda
Eurasia Plate Sannomegata 15 39 26 32 Akitakoma
40˚ 40˚

rc
Kampu 13 50 25 33 Iwate

pa n a
Chokai 9 38
27 Funagata
Japan Sea 20
9 cm/y

NE Ja
17 46 Nasu
14 42 Takahara
arc 20
Pacific Plate
a p an 41 Asama

J 10 35
SW Fuji
35˚ 14 35
22 37
Izu-Oshima 35˚
22 29 Toshima
Niijima 10 32 Udonejima
Subduction vector

Izu-Bonin
arc

Pacific Plate slab


gh
ou 4 cm/y depth contour
i Tr
hu

nka (50 km interval)


Na
us

Philippine Sea slab


-Ky

depth contour

arc
(10km interval)
yu

30˚ Phillipine Sea Plate Plate boundary 30˚


uk

Volcanic front
Ry

Figure 7. Locations of examined volcanoes with depth contours of the Pacific Plate slab and the Philippine Sea Plate slab. Red text indi-
cates volcanic front volcanoes while blue text indicates rear-arc volcanoes. Numbers to the right of volcanoes show MgO in the source
mantle (wt %) and numbers to the left show degree of melting of the mantle (F%).

calculations include (1) saving the COMAGMAT3.72 control codes and the major element compositions
from the CONTROL_PANEL in supporting information; (2) running COMAGMAT3.72, which reads the text
files, calculates, and saves the output results in another text file; and (3) reading the output text file and
storing in the COM_Result Worksheet. All processes are coded in the Excel VBA macro. Additional conver-
gent calculations for T(wet) using the Katz Worksheet, and MgO(peridotite) in the mantle using the PRIMELT2_-
MOD2 and PRIMACOT Worksheets (see section 2.8) run with the iteration calculations. The calculation
precision (see section 2.4) and the number of the iteration run are set in cells $E$20 and $E$21 usually two
and four, respectively (see above section 2.4).

3.3. Output
The MgO versus SiO2, MgO versus Al2O3, and MgO versus CaO panels graphically show the fitting results
(see (4) and (5) in Figure 1 and Figure 4a). The mantle equilibrium of the olivines is tested by Ni (wt. %) IN
OLIVINE panel in the CONTROL_PANEL (Figures 1 and 4c), where the users can test the controlling parame-
ters of COMAGMAT3.72 iteratively. The fractionation sequence including the mineral phases and composi-
tions and the fractionated solid and H2O content in the magma plot are shown in one panel (T(C) versus
mineral assemblage/composition/H2O(melt)) (Figures 1 and 4b). Two CMAS plots ((7) in Figure 1) and the
trace multielement plots ((8) in Figure 1) show the calculation results. Finally, users can extract the results
from PRIMACALC2 using the copy function of Excel from the field $I$2–$O$53 of the CONTROL_PANEL.

4. Applications
We evaluate PRIMACALC2 by applying the software program to the Quaternary arc magmas from the Japa-
nese Islands. Figure 7 shows the locations of volcanoes from which the basalts to basaltic andesites exam-
ined in this study were reported. From the Izu collision zone, we examined the N-Izu volcanic front (VF) Izu-
Oshima, Toshima; Udonejima and rear-arc (RA) Niijima lavas [Kimura et al., 2010]; and Fuji [Watanabe et al.,
2006]. From the southern NE Japan arc, we examined Asama [Gust et al., 1997; Kaneko, 1995; Okamoto,
1979], Takahara, and Nasu [Ban et al., 2013] lavas. From the northern NE Japan arc, we examined Funagata

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1506
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

100 100

90 a b
80
Oliv
70 Plag

Sample / PM
Cpx 10
60
Opx
50 Mt
Ilm
40 FIT
Fo(ol)
30 An(plag) 1
20 H2O x10 Rb Ba Th U Nb Ta K La CePb Pr Sr NdSm Zr Hf EuGd Tb Dy Y Ho Er TmYb Lu
%Xtal.
10
0 Sample Ol max. P.Bas_1 P.Bas_2
1300 1250 1200 1150 1100 1050 1000
0
100 100
90 c d
80
Oliv
70

Sample / PM
Plag
60 Cpx
Opx
50 Mt
Ilm 10
40 FIT
Fo(ol)
30 An(plag)
20 H2O x10
%Xtal.
10
0 Sample Ol max. P.Bas_1 P.Bas_2
1350 1300 1250 1200 1150 1100 1050 1000 Rb Ba Th U Nb Ta K La CePb Pr Sr NdSm Zr Hf EuGd Tb Dy Y Ho Er TmYb Lu
1

Figure 8. (a, b) Fractional crystallization sequence and (b, d) trace element compositions of Izu-Oshima and Chokai basalts for observed (sample) and primary basalts estimated by oli-
vine maximum fractionation model (Ol max.) and PRIMACALC2 (P.Bas_1 and P.Bas_2). Note that the effects of plagioclase and clinopyroxene fractionation along with olivine fractionation
are shown by lower REE and higher Sr contents in P.Bas_2 in the Izu-Oshima basalt.

[Kimura and Yoshida, 2006], Iwate [Kuritani et al., 2014a], Hachmantai [Ban et al., 2013], Akitakoma [Ban et al.,
2013], and Hakkoda and Osore [Ban et al., 2013] lavas. From the NE Japan rear-arc, we examined Sannome-
gata [Kuritani et al., 2014b], Kampu [Ban et al., 2013], and Chokai [Kimura and Yoshida, 2006]. Low-K tholeiitic
basalt to basaltic andesite were inferred from lower crustal amphibolite melts at Zao [Tatsumi et al., 2008]
and Azuma [Takahashi et al., 2012] volcanoes, and lavas with significant crustal assimilation such as Haruna
volcano [Kobayashi and Nakamura, 2001] were omitted from the examinations.

4.1. Preset and Explored Variables for Calculations


COMAGMAT3.72, and thus PRIMACALC2, requires P, fO2, and H2O to be preset values. For the fO2 in the
magma chamber, we assumed QFM to QFM10.5. This may be slightly reduced condition for an arc magma
which has Fe21/Fe(total) 5 0.78 of QFM11 equivalent [Kelley and Cottrell, 2012] at H2O 5 3 wt % [Kelley and
Cottrell, 2009]. However, the use of QFM11 oxygen buffer led to early crystallization of magnetite after oli-
vine in COMAGMAT3.72, which is inconsistent with the observed crystallization sequence. Measured Fe21/
Fe* in the natural basalt melt in olivine melt inclusions showed a wide scatter of 0.78–0.75 (QFM-QFM11.5
equivalent) [Kelley and Cottrell, 2012]. Given these values, our assumption is still within the observed range.
It is noted, however, that the spinel thermodynamic model formulated by the atmospheric pressure experi-
ments [Ariskin and Barmina, 1999] may have errors in the high pressure range. The second author is working
on this issue, and the results will be presented elsewhere.
Using NiO-Fo in olivine in the primary basalt, we can explore mantle equilibrium by altering P and H2O in
the magma chamber (see section 2.6). We began calculations with H2O 5 4 wt % or more (which provided
greater NiO in olivine; see Figure 4c) and found the associated magma chamber P (lower P also provided
greater NiO) by observing the Ni-Fo constraint. If the shallowest P did not satisfy the NiO-Fo constraint, H2O
was reduced accordingly. The origin of using 4 wt % H2O in the primary arc magma at the start has been
discussed elsewhere [Almeev et al., 2013a, 2013b; Hamada and Fujii, 2007; Kimura et al., 2010; Kuritani et al.,
2014a, 2014b; Plank et al., 2013]. The results always show 2.0–4.0 wt % H2O when the Li and Ripley [2010]
model D(Ni) is used, and water saturation was noted in some cases (Figure 8a), but always with a shallow
magma chamber (e.g., 0.001 GPa for Izu-Oshima; see supporting information S2 and Figure 8a). In these

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1507
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

cases, the initial water content was considered not to be a significant constraint, and the magma chamber
P was considered to be the fundamental control.

4.2. Back Calculations for Fractionated Basalt


We first investigate the role of olivine, plagioclase, and clinopyroxene fractionation. Figures 8a and 8b
shows an example of the calculation for IO-6 basalt from the older stage volcanic edifice of Izu-Oshima
[Kimura et al., 2010] (see supporting information S2).
In order to satisfy the NiO-Fo constraint for olivine, a total of 65 wt % fractionation of crystals is required
(supporting information S2). In the order of onset of crystallization, the fractionated minerals are olivine, pla-
gioclase, and clinopyroxene (see Figure 8a). This mineral assemblage is consistent with that observed in the
basalt lava. The depth of the magma chamber needs to be very shallow at 0.001 GPa, and although the
magma at this depth accommodates only 0.5 wt % H2O, the initial setting of water was >4 wt % according
to the previous estimates by experiments and the ABS3 model calculations [Hamada and Fujii, 2007; Kimura
et al., 2010]. This means that crystallization in the shallow magma chamber should proceed at H2O-satu-
rated conditions, with the excess of the initial water released as a result of decompression. The COMAG-
MAT3.72 model allows the user to account for such a possibility in the modelling fractionation process
[Almeev et al., 2008], so that the initial water content in the primary magma is not reflected in the shallow
magma chamber differentiation.
Mineral compositions were calculated and resulted in Fo72 olivine and An79 plagioclase when the calcu-
lated magma had the same chemical composition as the observed IO-6 (at 1140 C in Figure 8a). The older
stage basalt lavas had Fo75 olivine [Hamada and Fujii, 2007], close to the calculated value. In contrast, pla-
gioclase inclusions in the olivine had >An90, indicating disequilibrium crystallization with early crystallized
plagioclase (Hamada M. pers. com.). Crystallization of >An90 plagioclase was estimated in an early crystalli-
zation phase by COMAGMAT3.72 (at 1200 C; see Figure 8a). This discrepancy could be derived from the
crystallization mode between the perfect fractional crystallization used in the calculations. This examination
shows that the crustal magma chamber processes should be explored individually. PRIMACALC2 may pro-
vide some predictions, but petrographical and geochemical observations in the natural magma are the pri-
mary constraint.
The trace element compositions in the estimated primary basalt of IO-6 are compared in Figure 8b, using
various methods in PRIMACALC2 (Ol.max, P.Bas_1, and P.Bas_2). When comparing a simple olivine maxi-
mum fractionation calculation by PRIMELT2 Ol.max to that of PRIMACALC2 P.Bas_2 (supporting information
S2 and Figure 8b), PRIMELT2 resulted in a 35 wt % olivine fractionation whereas PRIMACALC2 requires 65
wt % fractionation. The differences in the calculated incompatible trace elements were 40%–46% lower for
PRIMACALC2 for large ion lithophile elements (LILE) and high field strength elements (HFSE), and 10%–30%
lower for REE. The differences in LILE, HFSE, and REE are related to the back addition of clinopyroxene and
plagioclase in PRIMACALC2, whereas PRIMELT2 uses Ol-monosaturated modelling [Herzberg and Asimow,
2008]. Sr is, in contrast, 24% higher by PRIMACALC2, reflecting the back addition of Sr-rich plagioclase (sup-
porting information S2 and Figure 8b). As shown in Figure 8a, the back addition region is shown in the [T(C)
versus mineral assemblage/composition/H2O(melt)] panel. Users can identify whether or not mineral phases
other than olivine are added. This provides a caution for the results of PRIMELT2 calculations in which the
olivine maximum fractionation is only considered.
The difference between the models may not be significant in relation to the trace element mass balance
calculation for arc basalts. However, in these types of calculations using, for example, ABS models, there are
large uncertainties in the source slab and mantle compositions and in the elemental partitioning in slab
dehydration/melting [Kimura et al., 2010], but improving the estimation of the primary arc basalt composi-
tion will improve this problem.

4.3. Back Calculations for Near-Primary Basalt


The PRIMACALC2 result is also given for a Sannomegata basalt [Kuritani et al., 2014b] (supporting informa-
tion S2 and Figures 8c and 8d). The results of PRIMELT2 (fractionated olivine 4 wt %) and PRIMACALC2 (frac-
tionated olivine 10 wt %) (see supporting information S2 and Figures 8c and 8d) are almost identical. This is
consistent with the near-primary nature of the Sannomegata basalt, as shown by the presence of skeletal
olivine alone in its phenocryst phase [Kuritani et al., 2014b], and thereby suggests a small role for olivine

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1508
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

fractionation. In the case of the Sannomegata basalt, the water content in the primary basalt is set at 4 wt
% and the magma chamber pressure at 1 GPa in PRIMACALC2 (supporting information S2) which is required
to satisfy NiO-Fo mantle equilibration of olivine.
Although water was estimated at 7 wt % in the primary Sannomegata basalt using the thermodynamic
phase equilibria [Kuritani et al., 2014b] and at 4–8 wt % using the trace element mass balance by ABS4
[Kimura and Nakajima, 2013], the water content estimated by PRIMACALC2 was not more than 5 wt %.
Water content in the fractionated magmas increased dramatically to >10 wt % after the onset of clinopyr-
oxene in a deep closed magma chamber for the Sannomegata basalt (Figure 8c). There are, however, large
errors in the COMAGMAT3.72 calculations, and thus the effects of water on crystal fractionation should be
investigated further, but it is worth noting that the fractional crystallization sequence in the crust should
control water by means of the emplacement depth of the magma chamber. COMAGMAT3.72 could be used
to explore such an effect by comparison of a crystal fractionation sequence to the observed magmas
[Almeev et al., 2013a, 2013b, 2007]. PRIMACALC2 output reflected the model characteristics and enabled
using these to constrain the mantle to crustal processes. But, as has been shown in previous work, individ-
ual magma chamber processes should be examined in detail using petrography, petrology, and geochemis-
try [Kobayashi and Nakamura, 2001; Takahashi et al., 2012; Tatsumi and Suzuki, 2009].

4.4. Source Mantle Conditions


We calculated source mantle conditions using PRIMACALC2, and the results for volcanoes in N-Izu, southern
and northern NE Japan arcs are given in supporting information S3 and Figure 9. Important factors are the
T, F, P conditions, and the source fertility expressed by MgO in the source peridotite (Figures 9a–9c). Water
content in the primary basalt, which affects the mantle melting T, was determined using the NiO-Fo con-
straint and resulted in 2–4 wt % in all primary basalts (supporting information S3). The overall agreement of
the water content for an arc magma (4 wt %) [Kimura and Nakajima, 2013; Plank et al., 2013] was achieved,
although 70% of the calculation results gave rise to 2 wt % initial H2O in the primary basalts in the magma
chamber (supporting information S3).
The estimated source conditions demonstrate that the primary basalts in VF volcanoes had greater values
of F (15–26%) over the temperature range 1190–1340 C, and in contrast lower F values (4–17%) were seen
with the same T range for the RA volcanoes (Figure 9a). Although the T ranges overlap, the RA basalts show
a higher P at a given T (Figure 9c). In the experiments, an increase in P leads to an increase in the solidus-
liquidus temperature [e.g., Katz et al., 2003; Tatsumi et al., 1983], so that the calculated relation in the across
the arc setting is consistent with the petrological rule.
Source fertility shown by MgO ranges from 27–45 wt % up to 53 wt %, covering almost the entire composi-
tional range of mantle peridotites [Bodinier and Godard, 2003]. This suggests a pyroxene-rich lherzorite to
harzburgite-dunite lithology of the source mantle (Figure 9b). Fo in the olivine in mantle equilibrium and
Mg# (Mg# 5 Mg/Mg1Fe molar ratio calculated by the estimated mantle MgO and fixed FeO 5 8 wt %) of
the source mantle showed a clear correlation Mg# 5 0.9906 3 Fo (r2 5 0.937) (not shown). This indicates
that the MgO estimate by our method is internally consistent.
It is also consistent that a fertile mantle source shows a greater F within the designated pressure range for VF
and RA (Figures 9a and 9b), but T shows a clearer correlation with F (Figure 9a). MgO inversely correlates with F,
irrespective of the similar T range for VF and RA basalts. Therefore, the systematic difference in F between VF
and RA is likely to be due to the difference in melting pressure (Figure 9c) by which increases in P decrease F at
the same temperature [Katz et al., 2003] rather than seeing the effects from changes in source fertility or T.
Pressure is the important factor in the chemical difference between the VF and RA primary basalts, with a
lower P 5 0.8–1.6 GPa in VF and a higher P 5 1.4–2.3 GPa in RA (Figure 9d). F is the secondary control in
both VF and RA mantle, which is basically controlled by the difference in temperature [Falloon et al., 2007],
i.e., lower at shallow depths with T 5 1200 C and higher at greater depths with T 5 1350 C (Figures 9a and
9d). An increase in T with depth increases F, although an increase in P can cause a counter effect. This com-
mon feature for both VF and RA suggests a strong normal thermal gradient in the upper half of the mantle
wedge asthenosphere (Figure 9d).
This thermal gradient further suggests that convection in the mantle wedge plays a key role in the thermal
structure. The P, F, and T estimates developed here consider the effect of water. The water content in the

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1509
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

30 30 N-Izu VF
N-Izu RA
a b Fuji
NE Japan NVF
25 High 25 NE Japan SVF
NE Japan RA

20 20

F (%)

F (%)
15 15

10 10

5 5
DMM
Fertile Depleted
Low
0 0
1180 1200 1220 1240 1260 1280 1300 1320 1340 25 30 35 40 45 50 55
Temperature (C) MgO in mantle (wt%)
2.4
RA VF
2.2
c High
Depth
d Arc crust

(km) Moho
2 Partial melting
50

3% 5%
1.8 1200 Partial
(C) melting
P (GPa)

ss
n

ce
tio
Slab fluid

Ex
ra
1.6

yd

n
tio
H

ra
1350

ss
Ex ion

yd
ce

eh
100

t
ra
1.4

D
yd
Slab melt

eh
D
1.2

n
tio
g
tin

ra
el

yd
m

eh
ab
1

D
Sl
Low 150 t
us

0.8
cr

1180 1200 1220 1240 1260 1280 1300 1320 1340


an
ce

Temperature (C)
O

Figure 9. Intensive-extensive parameters estimated by PRIMACALC2 for N-Izu, southern NE Japan, and northern NE Japan arcs. (a) Mantle
melting temperature (T) versus degree of partial melting (F). Thick lines in light red and blue show the average of volcanic front and rear-
arc primary basalts, respectively. (b) MgO in the source mantle versus degree of melting. (c) Melting pressure (P) versus temperature. (d)
Across arc schematic cross section of mantle wedge. Black rectangles show P and T ranges of primary basalt segregation estimated by PRI-
MACALC2. Overall positive correlation between P and T suggests geotherm of the melting zone shown as 1200 and 1350 C contour lines.
For slab dehydration and melting conditions, see Kimura and Nakajima [2013].

source most significantly affects the T estimate, with subordinate effects to P and F (Figure 6). Therefore,
the conclusions given here are robust. The mantle P-T structure has been inferred by ABS models [Kimura
and Yoshida, 2006; Kimura and Nakajima, 2013; Kimura et al., 2009, 2010] and thermodynamic phase equili-
bria [Kuritani et al., 2014a, 2014b]. PRIMACALC2 shows similar results, with a completely different model
scheme.

4.5. Along-Arc Variation of the Source Mantle


Source mantle conditions beneath VF and RA in the N-Izu arc are characterized by a greater F 5 14–22%
beneath VF and a lesser F 5 10% beneath RA with a relatively fertile mantle composition (MgO 5 28–35 wt
%; Figures 9a and 9b). The Fuji volcano in the N-Izu collision zone has RA characteristics with lower F 5 5–
16% and a wide range of T 5 1220–1320 C (Figure 9a). Mantle fertility beneath Fuji varied with MgO 5 27–
42 wt %, suggesting a heterogeneous source (Figure 9b) located in an intermediate depth range P 5 1.4–
1.5 GPa (Figure 9c). These are different from the RA basalts in N-Izu.
The results from the Fuji volcano raise a question as to whether or not the source mantle beneath a volcano
is heterogeneous. Each sample that has passed through the whole melting process, melt migration system,

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1510
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

and crustal plumbing system must represent a mixture of melts derived from a range of sources. The appa-
rent source composition derived from PRIMACALC2 must therefore represent some kind of mixture. This
implies that the spectrum of heterogeneity is even larger than that inferred, since mixing presumably
reduces the range of extremes. If so, the wide variation found between T-F-MgO would relate to the thermal
gradient in the mantle that forms greater F in the high T region, leaving a high MgO residue in the source.
The P estimate for Fuji is fairly constant at P 5 1.4–1.6 GPa (Figure 9c), and the Sr-Nd-Pb isotope composi-
tions are fairly uniform [Watanabe et al., 2006]. Thus the thermal heterogeneity, if persistent, is significant
even in the root of one single volcano.
The basalts from the southern NE Japan VF, including Asama, showed relatively low F and T, with depleted
sources (Figures 9a and 9b). There was a considerable overlap between the southern and northern VF in NE
Japan. However, the northern VF basalts tended to show higher F and T in comparison to N-Izu VF but with
similar depletion in the source (Figures 9a and 9b). The RA basalts from northern NE Japan has a higher
mantle T and F, with an intermediate mantle depletion similar to the depleted MORB source mantle [Work-
man and Hart, 2005] (Figure 9c).
Spatial distributions of average values of MgO and F in the source mantle are shown in Figure 6. The upper
mantle beneath the northern NE Japan VF is relatively fertile (27–37 wt %), but in southern NE Japan it is
depleted (41–46 wt % MgO). The mantle becomes fertile again in N-Izu VF (29–35 wt % MgO) (Figure 6). In
contrast, the RA mantle consistently shows intermediate depletion (32–39 wt % MgO) close to the DMM
value (35 wt % MgO), with the exception of the very depleted mantle (50 wt % MgO) beneath Kampu vol-
cano (Figure 6). F relates to T and P and may represent the production rate of the mantle. F is high (20%–
26%) in northern NE Japan, low (14%–20%) in the southern NE Japan VF, and remains lower (14%–22%) in
the VF of the N-Izu arc (Figure 6). F is low (9%–15%) in the NE Japan RA and remains low and constant
(10%) through Fuji to the N-Izu RA. All of these observations suggest RA characteristics of Fuji primary
basalt similar to those noted in a previous report [Tani et al., 2011].
The greater mantle depletion, and the lower degree of melting beneath the southern NE Japan VF, could
be related to the overlap of the Philippine Sea Plate slab onto the Pacific Plate slab that promoted depletion
of the mantle by a greater water supply to the source mantle [Nakamura and Iwamori, 2009; Nakamura
et al., 2008] and a poor development of the mantle wedge corner flow. Alternatively, intensive magma pro-
duction in the middle Miocene fore arc, due to an extremely high temperature mantle during the opening
of the Sea of Japan, depleted the subarc mantle beneath the southern NE Japan VF [Yamamoto and Hoang,
2009].
As shown above, PRIMACALC2 is a useful petrological tool for estimating primary basalt compositions,
including those of major/trace elements and water contents, as well as the P, T, F, and MgO source mantle
conditions, which generate the tholeiitic low-K to calc-alkaline medium-K primary basalts. Moreover, a com-
bined use with ABS models [Kimura and Nakajima, 2013; Kimura et al., 2009, 2010] provides (1) a robust esti-
mate of the primary basalt composition which is immediately used for ABS and (2) a unique test for the
estimates of the source mantle conditions beneath arcs by back calculation modelling. Because COMAG-
MAT3.72, and thus PRIMACALC2, can be applied to dry basalt systems with tholeiitic to mildly alkalic com-
positions, PRIMACALC2 is useful for MORB and some OIB, including those with tholeiitic to transitional
compositions.

5. Summary
We developed a numerical simulation model PRIMACALC2 for estimation of hydrous primary magma com-
positions and their genetic conditions in subduction zones. PRIMACALC2 uses COMAGMAT3.72 for the for-
ward calculation of crystal fractionation of hydrous magma in the intracrustal depth magma chamber at
given P, T, fO2, and H2O conditions, and back calculates major and trace element compositions to a primary
basalt in mantle equilibrium using the NiO-Fo relationship. PRIMACALC2 also calculates the P, T, and F con-
ditions of the source mantle, and the source mantle fertility expressed by MgO. Calculations were applied
to VF and RA basalts and to basaltic andesite magmas from N-Izu, southern NE Japan, and northern NE
Japan arcs. The results argue that the mantle beneath VF is melted at shallow depths (0.8–1.6 GPa), whereas
melting occurs deeper (1.4–2.3 GPa) beneath RA. Although a similar range of melting temperatures was
found beneath VF and RA, higher temperatures were observed from the deeper mantle. These results

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1511
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

suggest that the isotherm in the mantle is inclined toward RA, and that a normal mantle geotherm is main-
tained in the upper half of the mantle beneath VF and RA. Water contents of the primary magmas were
constrained by relationships between crystallization conditions and NiO contents in olivine in mantle equi-
librium. Initial H2O in the primary basalt positively correlated with NiO in olivine, whereas the magma cham-
ber P was negatively correlated, thus providing the petrologically useful constraints. Estimated water
contents in the primary magmas were 2–4 wt %, consistent with (or somewhat lower than) previous esti-
mates. Thus, PRIMACALC2 is a new powerful tool for estimation of the conditions and compositions of pri-
mary magmas generated in the arcs, as well as for decoding P, T, F, H2O parameters of their fractionation.

Notation
Important acronyms used in PRIMACALC2 are listed below.
Ni(ol)wt % NiO in olivine.
Fo(ol)% Fo content in olivine.
Mg# Bas Mg# 5 Mg/(Mg1Fe21) molar ratio in basalt.
Fe21/Fe* ferric iron content over total iron in basalt.
H2O(wt %) water content in basalt.
P(COM) magma chamber pressure set for COMAGMAT3.72.
TWC(KAZ) hydrous mantle temperature estimated by Katz et al. [2003].
TDC(Herz) dry mantle temperature estimated by Herzberg et al. [2007].
PGPa(COM/Herz) pressure at magma chamber given for COMAGMAT3.72 and that estimated for source
mantle by Herzberg et al. [2007].
F%(Herz) degree of partial melting estimated by Herzberg et al. [2007].
%Xfrac. total fractionated mineral weight in magma chamber.
MgO PM estimated MgO in the source mantle by Herzberg et al. [2007].

Acknowledgments References
This project was greatly improved
Almeev, R. R., F. Holtz, J. Koepke, F. Parat, and R. E. Botcharnikov (2007), The effect of H2O on olivine crystallization in MORB: Experimental
through discussions with G. Barmina of
calibration at 200 MPa, Am. Mineral., 92, 670–674.
Vernadsky Institute, H. Kawabata, M.
Almeev, R. R., F. Holtz, J. Koepke, K. M. Haase, and C. W. Devey (2008), Depths of partial crystallization of H2O-bearing MORB: Phase equili-
Hamada, and T. Sakyuama of IFREE/
bria simulations of basalts at the MAR near Ascension Island (7 -11 S), J. Petrol., 49, 25–45.
JAMSTEC, J. B. Gill of the University of
Almeev, R. R., F. Holtz, J. Koepke, and F. Parat (2012), Experimental calibration of the effect of H2O on plagioclase crystallization in basaltic
California Santa Cruz, R. J. Stern of the
melt at 200 MPa, Am. Mineral., 97, 1234–1240.
University of Texas at Dallas, and M.
Almeev, R. R., A. A. Ariskin, J.-I. Kimura, and G. S. Barmina (2013a), The role of polybaric crystallization in genesis of andesitic magmas: Phase
Carr and M. Feigenson of Rutgers
equilibria simulations of the Bezymianny volcanic subseries, J. Volcanol. Geotherm. Energy, 263, 182–192, doi:10.1016/
University. We thank M. Ban, T.
j.jvolgeores.2013.1001.1004.
Takahashi, Y. Hirahara, T. Ohba, A.
Almeev, R. R., J.-I. Kimura, A. A. Ariskin, and A. Y. Ozerov (2013b), Decoding crystal fractionation in calc-alkaline magmas from the Bezy-
Fujinawa, S. Hayashi, T. Yoshida, T.
mianny Volcano (Kamchatka, Russia) using mineral and bulk rock compositions, J. Volcanol. Geotherm. Res., 263, 141–171, doi:10.1016/
Miyazaki, Q. Chang, R. Senda, and Y.
j.jvolgeores.2013.1001.1003.
Tatsumi for allowing us to use the
Ariskin, A. A. (1999), Phase equilibria modeling in igneous petrology: Use of COMAGMAT model for simulating fractionation of ferro-
unpublished major element and Ni
basaltic magmas and the genesis of high-alumina basalt, J. Volcanol. Geotherm. Res., 90, 115–162.
data for Takahara, Nasu, Akitakoma,
Ariskin, A. A., and G. S. Barmina (1999), An empirical model for the calculations of spinel-melt equilibrium in mafic igneous system at
Hakkoda, Osore, and Kampu
atomospheric pressure: II. Fe-Ti oxides, Contrib. Mineral. Petrol., 134, 251–263.
volcanoes. Thanks are also extended
Ariskin, A. A., and G. S. Barmina (2004), COMAGMAT: Development of a magma crystallization model and its petrological applications, Geo-
to Kirill Bychkov for providing
chem. Int., 42, S1–S157.
descriptions of the interface and the
Ariskin, A. A., M. Y. Frenkel, G. S. Barmina, and R. L. Nielsen (1993), COMAGMAT: A Fortran program to model magma differentiation proc-
details of the FORTRAN program of
esses, Comput. Geosci., 19, 1155–1170.
COMAGMAT3.72. Review comments
Asimow, P. D., J. E. Dixon, and C. H. Langmuir (2004), A hydrous melting and fractionation model for mid-ocean ridge basalts: Application
by C. Herzberg, an anonymous
to the Mid-Atlantic Ridge near the Azores, Geochem. Geophys. Geosyst., 5, Q01E16, doi:10.1029/2003GC000568.
reviewer, and the Associate Editor C.-T.
Ban, M., et al. (2013), Sr-Nd-Pb isotope compositions of frontal arc stratovolcanoes in Northeast Japan arc, in IAVCEI Scientific Assembly
Lee greatly improved this paper.
Kagoshima, Japan, p. 15, The Volcanological Society of Japan, Tokyo.
Beattie, P., C. Ford, and D. Russell (1991), Partition coefficients for olivine-melt and orthopyroxene-melt systems, Contrib. Mineral. Petrol.,
109, 212–224.
B
edard, J. H. (2005), Partitioning coefficients between olivine and silicate melts, Lithos, 83, 394–419.
Bindeman, I. N. (2007), Erratum to I. N. Bindeman, A. M. Davis, and M. J. Drake (1998), ‘‘Ion microprobe study of plagioclase-basalt partition
experiments at natural concentration levels of trace elements’’ Geochimica Cosmochimica Acta 62, 1175–1193, Geochim. Cosmochim.
Acta, 71, 2414.
Bindeman, I. N., A. M. Davis, and M. J. Drake (1998), Ion microprobe study of plagioclase-basalt partition experiments at natural concentra-
tion levels of trace elements, Geochim. Cosmochim. Acta, 62, 1175–1193.
Bodinier, J.-L., and M. Godard (2003), Orogenic, ophiolitic, and abyssal peridotites, in Mantle and Core, Triatise on Geochemisty, edited by
R. W. Carlson, chap. 2.04, pp. 103–170, Elsevier, Amsterdam.

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1512
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

Cooper, K. M., J. M. Eiler, P. D. Asimow, and C. H. Langmuir (2004), Oxygen isotope evidence for the origin of enriched mantle beneath the
mid-Atlantic ridge, Earth Planet. Sci. Lett., 220, 297–316.
Cottrell, E., and K. A. Kelley (2011), The oxidation state of Fe in MORB glasses and the oxygen fugacity of the upper mantle, Earth Planet.
Sci. Lett., 305, 270–282.
Danyushevsky, L. V., and P. Plecov (2011), Petrolog3: Integrated software for modeling crystallization processes, Geochem. Geophys. Geo-
syst., 12, Q07021, doi:10.1029/2011GC003516.
Danyushevsky, L. V., F. N. Della-Pasqua, and S. Sokolov (2000), Re-equilibration of melt inclusions trapped by magnesian olivine phenoc-
rysts from subduction-related magmas: Petrological implications, Contrib. Mineral. Petrol., 131, 68–83.
Falloon, T. J., and L. V. Danyushevsky (2000), Melting of reflactory mantle at 1.5, 2.0 and 2.5 GPa under anhydrous and H2O under-
saturated conditions: Implications for petrogenesis of high-Ca Boninites and influence of subduction components on mantle melting, J.
Petrol., 41, 257–283.
Falloon, T. J., L. V. Danyushevsky, A. A. Ariskin, D. H. Green, and C. E. Ford (2007), The application of olivine geothermometry to infer crystal-
lization temperatures of parental liquids: Implications for the temperature of MORB magmas, Chem. Geol., 241, 207–233.
Ghiorso, M. S., M. M. Hirschmann, P. W. Reiners, and V. C. Kress (2002), The pMELTS: A revision of MELTS for improved calculation of phase
relations and major element partitioning related to partial melting of the mantle to 3 GPa, Geochem. Geophys. Geosyst., 3(5), 1030, doi:
10.1029/2001GC000217.
Grove, T. L., and M. B. Baker (1984), Phase equilibrium controls on the tholeiitic versus calcalkaline differentiation trends, J. Geophys. Res.,
89, 3253–3274.
Grove, T. L., and R. J. Kinzler (1986), Petrogenesis of andesites, Annu. Rev. Earth Planet. Sci., 14, 417–454.
Gust, D. A., R. J. Arculus, and A. B. Kersting (1997), Aspects of magma sources and processes in the Honshu arc, Can. Mineral., 35, 347–365.
Hamada, M., and T. Fujii (2007), H2O-rich island arc low-K tholeiitic magma inferred from Ca-rich plagioclase-melt includion equilibria, Geo-
chem. J., 41, 437–461.
Herzberg, C. (2011), Identification of source lithology in the Hawaiian and Canary Islands: Implications for origins, J. Petrol., 52, 113–146.
Herzberg, C., and P. D. Asimow (2008), Petrology of some oceanic island basalts: PRIMELT2.XLS software for primary magma calculation,
Geochem. Geophys. Geosyst., 9, Q09001, doi:10.1029/2008GC002057.
Herzberg, C., and M. J. O’Hara (2002), Plume-associated ultramafic magmas of Phanerozoic age, J. Petrol., 43, 1857–1883.
Herzberg, C., P. D. Asimow, N. Arndt, Y. Niu, C. M. Lesher, J. G. Fitton, M. J. Cheadle, and A. D. Saunders (2007), Temperatures in ambient
mantle and plumes: Constraints from basalts, picrites, and komatiites, Geochem. Geophys. Geosyst., 2, Q02006, doi:10.1029/
2006GC001390.
Herzberg, C., P. D. Asimow, D. A. Ionov, C. Vidito, M. G. Jackson, and D. Geist (2013), Nickel and helium evidence for melt above the core-
mantle boundary, Nature, 493, 393–397.
Hirose, K., and T. Kawamto (1995), Hydrous partial melting of lherzorite at 1 GPa: The effect of H2O on the genesis of basaltic magmas,
Earth Planet. Sci. Lett., 133, 463–473.
Hirschmann, M. M., M. S. Ghiorso, F. A. Davis, S. M. Gordon, S. Mukherjee, T. L. Grove, M. Krawczynski, E. Medard, and C. B. Till (2008), Library
of experimental phase relations (LEPR): A database and web portal for experimental magmatic phase equilibria data, Geochem. Geophys.
Geosyst., 9, Q03011, doi:10.1029/2007GC001894.
Jones, J. H. (1984), Temperature and pressure-independent correlations of olivine-liquid partition coefficients and their application to trace
element partitioning, Contrib. Mineral. Petrol., 88, 126–132.
Kaneko, T. (1995), A kinematic subduction modl for the genesis of back-arc low-K volcanoes at a two-overlaping subduction zone, central
Japan: Another volcanic front originated from the Philippine Sea Plate subduction, J. Volcanol. Geotherm. Res., 66, 9–26.
Katz, R. F., M. Spiegelman, and C. H. Langmuir (2003), A new parameterization of hydrous mantle melting, Geochem. Geophys. Geosyst., 4(9),
1073, doi:10.1029/2002GC000433.
Kelley, K. A., and E. Cottrell (2009), Water and the oxidation state of subduction zone magmas, Science, 325, 605–607.
Kelley, K. A., and E. Cottrell (2012), The influence of magmatic differentiation on the oxidation state of Fe in a basaltic arc magma, Earth
Planet. Sci. Lett., 329–330, 109–121.
Kimura, J.-I., and J. Nakajima (2013), Behavior of subducted water and its role on the arc magma genesis in the NE Japan arc: A combined
geophysical and geochemical approach, in IAVCEI Scientific Assembly, Kagoshima, Japan, p. 15, The Volcanological Society of Japan,
Tokyo.
Kimura, J.-I., and T. Yoshida (2006), Contributions of slab fluid, mantle wedge and crust to the origin of Quaternary lavas in the NE Japan
arc, J. Petrol., 47, 2185–2232.
Kimura, J.-I., M. Katakuse, Y. Shimoshioiri, T. Nagao, S. Kakubuchi, K. Furuyama, and N. Tsuchiya (2006), Various slab fluids and melts from a
common slab: Sources for intra-plate and arc basalts, high-Mg andesites, and adakites in the SW Japan arc, Eos Trans. AGU, 87(52), Fall
Meet. Suppl., Abstract V41B-1726.
Kimura, J.-I., P. van Keken, B. R. Hacker, H. Kawabata, T. Yoshida, and R. J. Stern (2009), Arc Basalt Simulator (ABS) version 2, a simulation
model for slab dehydration, fluid-mantle reaction, and fluid-fluxed mantle melting for arc basalts: Modeling scheme and application,
Geochem. Geophys. Geosyst., 10, Q09004, doi:10.1029/2008GC002217.
Kimura, J.-I., J. R. K. Adam, M. Rowe, N. Nakano, M. Katakuse, P. van Keken, B. Hacker, and R. J. Stern (2010), Origin of cross-chain geochemi-
cal variation in Quaternary lavas from northern Izu arc: A quantitative mass balance approach on source identification and mantle
wedge processes, Geochem. Geophys. Geosyst., 11, Q10011, doi:10.1029/2010GC003050.
Kobayashi, K., and E. Nakamura (2001), Geochemical evolution of Akagi volcano, NE Japan: Implications for interaction between island-arc
magma and lower crust, and generation of isotopically various magmas, J. Petrol., 42, 2303–2331.
Kuritani, T., T. Yoshida, J.-I. Kimura, Y. Hirahara, and T. Takahashi (2014a), Water content of primitive low-K tholeiitic basalt magma from
Iwate volcano, NE Japan arc: Implications for differentiation mechanism of frontal-arc basalt magmas, Mineral. Petrol., 108, 1–11.
Kuritani, T., T. Yoshida, J.-I. Kimura, T. Takahashi, Y. Hirahara, T. Miyazaki, R. Senda, Q. Chang, and Y. Ito (2014b), Primary melt from
Sannome-gata volcano, NE Japan arc: Constraints on generation conditions of rear-arc magmas, Contrib. Mineral. Petrol., 167, 969, doi:
10.1007/s00410-00014-00969-00417.
Li, C., and E. M. Ripley (2010), The relative effects of composition and temperature on olivine-liquid Ni partitioning: Statistical deconvolu-
tion and implications for petrologic modeling, Chem. Geol., 275, 99–104.
Matzen, A. K., M. B. Baker, R. B. John, and E. M. Stolper (2013), The temperature and pressure dependence of nickel partitioning between
olivine and silicate melt, J. Petrol., 54, 2521–2545.
Miyashiro, A. (1974), Volcanic rock series in island arcs and active continental margins, Am. J. Sci., 274, 321–355.
Nakamura, H., and H. Iwamori (2009), Contribution of slab-fluid in arc magmas beneath the Japan arcs, Gondwana Res., 16, 431–445.

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1513
Geochemistry, Geophysics, Geosystems 10.1002/2014GC005329

Nakamura, H., H. Iwamori, and J.-I. Kimura (2008), Geochemical evidence for enhanced fluid flux due to overlapping subducting plates,
Nat. Geosci., 1, 380–384.
Niu, Y. (1997), Mantle melting and melt extraction processes beneath ocean ridges: Evidence from sbyssal peridotites, J. Petrol., 38, 1047–
1074.
Okamoto, K. (1979), Geochemical study on magmatic differentiation of Asama volcano, central Japan, J. Geol. Soc. Jpn., 85, 525–535.
Pilet, S., M. L. Baker, O. M€ untener, and E. M. Stolper (2011), Monte Carlo simulations of metamorphic enrichment in the lithosphere and
implicaitons for the source of alkaline basalts, J. Petrol., 52, 1415–1442.
Plank, T., K. A. Kelley, M. M. Zimmer, E. H. Hauri, and P. J. Wallace (2013), Why do mafic arc magmas contain 4 wt % water on average?,
Earth Planet. Sci. Lett., 364, 168–179.
Roeder, P. L., and R. F. Emslie (1970), Olivine-liquid equilibrium, Contrib. Mineral. Petrol., 29, 275–289.
Rollinson, H. (1993), Using Geochemical Data: Evaluation, Presentation, Interpretation, Addison-Wesley, Essex, U. K.
Sisson, T. W., and T. L. Grove (1993), Experimental investigations of the role of H2O in calc-alkaline differentiation and subduction zone
magmatism, Contrib. Mineral. Petrol., 113, 143–166.
Sun, S., and W. F. McDonough (1989), Chemical and isotopic systematics of oceanic basalts: Implications for mantle composition and proc-
esses, in Magmatism in the Ocean Basins, edited by A. D. Saunders and M. J. Norry, pp. 313–345, Spec. Publ., Geol. Soc., London.
Takahashi, E. (1986), Orgin of basaltic magmas—Implications from peridotite melting experiments and an olivine fractionation model, J.
Volcanol. Soc. Jpn., 30, S17–S40.
Takahashi, T., Y. Hirahara, T. Miyazaki, R. Senda, Q. Chang, J.-I. Kimura, and Y. Tatsumi (2012), Primary magmas at the volcanic front of NE
Japan arc: Coeval eruption of crustal low-K tholeiitic and mantle-derived medium-K calc-alkaline basalts at Azuma volcano, J. Petrol., 54,
103–148, 10.1093/petrology/egs1065.
Tani, K., R. S. Fiske, D. J. Dunkley, O. Ishizuka, T. Oikawa, I. Isobe, and Y. Tatsumi (2011), The Izu Peninsula, Japan: Zircon geochronology
reveals a record of intra-oceanic rear-arc magmatism in an accreted block of Izu-Bonion upper crust, Earth Planet. Sci. Lett., 303, 225–
239.
Tatsumi, Y., and T. Suzuki (2009), Tholeiitic vs calc-alkalic differentiation and evolution of arc crust: Constraints from melting experiments
on a basalt from the Izu-Bonin-Mariana arc, J. Petrol., 50, 1575–1603.
Tatsumi, Y., M. Sakuyama, H. Fukuyama, and I. Kushiro (1983), Generation of basaltic magmas and thermal structure of the mantle wedge
in subduction zone, J. Geophys. Res., 88, 5815–5825.
Tatsumi, Y., T. Takahashi, Y. Hirahara, Q. Chang, T. Miyazaki, J.-I. Kimura, M. Ban, and A. Sakayori (2008), New insights into andesite genesis:
The role of mantle-derived calc-alkalic and crust-derived tholeiitic melts in magma differentiation beneath Zao Volcano, NE Japan, J.
Petrol., 49, 1971–2008, doi:10.1093/prtrology/egn1055.
Till, C. B., T. L. Grove, and M. Krawczynski (2012), A melting model for variably depleted and enriched lherzolite in the plagioclase and spi-
nel stability fields, J. Geophys. Res., 117, B06206, doi:10.1029/2011JB009044.
Wallace, P. J. (2005), Volatiles in subduction zone magmas: Concentrations and fluxes based on melt inclusion and volcanic gas data, J. Vol-
canol. Geotherm. Res., 140, 217–240.
Wang, Z., and G. A. Gaetani (2008), Partitioning of Ni between olivine and siliceous eclogite partial melt: Experimental constraints on the
mantle source of Hawaiian basalts, Contrib. Mineral. Petrol., 156, 661–678.
Watanabe, S., E. Widom, T. Ui, N. Miyaji, and A. M. Roberts (2006), The evolution of a chemically zoned magma chamber: The 1707 eruption
of Fuji volcano, Japan, J. Volcanol. Geotherm. Res., 152, 1–19.
Wood, B. J., and J. D. Blundy (1997), A predictive model for rare earthe element partitioning between clinopyroxene and anhydrous silicate
melt, Contrib. Mineral. Petrol., 129, 166–181.
Workman, R. K., and S. R. Hart (2005), Major and trace element composition of the depleted MORB mantle (DMM), Earth Planet. Sci. Lett.,
231, 53–72.
Yamamoto, T., and N. Hoang (2009), Synchronous Japan Sea opening Miocene fore-arc volcanism in the Abukuma Mountains, NE Japan:
An advancing hot asthenosphere flow versus Pacific slab melting, Lithos, 112, 575–590.

KIMURA AND ARISKIN C 2014. American Geophysical Union. All Rights Reserved.
V 1514

You might also like