Origin of arc-like continental basalts: Implications for ...

LITHOS-03785; No of Pages 41 Lithos xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Lithos journal homepage: www.elsevier.com/locate/lithos

Origin of arc-like continental basalts: Implications for deep-Earth fluid cycling and tectonic discrimination Xuan-Ce Wang a,⁎, Simon A. Wilde a, Bei Xu b, Chong-Jin Pang c a b c

The Institute for Geoscience Research (TIGeR), Department of Applied Geology, Curtin University, GPO Box U1987, Perth, WA 6845, Australia The Key Laboratory of Orogenic Belts and Crustal Evolution, Ministry of Education, School of Earth and Space Sciences, Peking University, Beijing 100871, China College of Earth Sciences, Guilin University of Technology, Guilin 541004, PR China

a r t i c l e

i n f o

Article history: Received 5 August 2015 Accepted 17 December 2015 Available online xxxx Keywords: Intra-continental basalts Arc basalts Fluid-mobile elements Arc-like geochemical signature Mantle transition zone Deep-Earth water cycling

a b s t r a c t Continental basalts generally display enrichment of fluid-mobile elements and depletion of high-field-strength elements, similar to those that evolved in the subduction environment, but different from oceanic basalts. Based on the continental flood basalt database for six large igneous provinces, together with rift-related basalt data from the Basin and Range Province, this study aimed to test the validity of geochemical tectonic discrimination diagrams in distinguishing arc-like intra-continental basalts from arc basalts and to further investigate the role of deep-Earth water cycling in producing arc-like signatures in large-scale intra-continental basalts. Our evaluation shows that arc-like intra-continental basalts can be distinguished from arc basalts by integrating the following factors: (1) the FeO, MgO, and Al2O3 concentrations of the primary melt; (2) Ti\\V, Zr\\Zr/Y, Zr\\Ti, and Ti/V\\Zr/Sm\\Sr/Nd discrimination diagrams; (3) the coexistence of arc-like and OIB-like subtype basalts within the same province; (4) primitive mantle-normalized trace element distribution patterns. The similarity of enrichment in fluid-mobile elements (Ba, Rb, Sr, U, and K) between arc-like and true arc basalts suggests the importance of water flux melting in producing arc-like signatures in continental basalts. Experimentally determined liquid lines of descent (LLD) imply high magma water concentrations for continental flood basalts (CFBs) and the Basin and Range basalts. Furthermore, estimates based on the Al2O3–LLD method indicates 4.0–5.0 wt% pre-eruptive magma H2O concentration for CFBs and the Basin and Range basalts. The tight relationships between H2O/Ce and Ba/La, Ba/Nb and Rb/Nb based on global arc basalt data were further used to estimate the primary H2O concentrations. With the exception of the Emeishan CFBs (mainly containing 4.0–5.6 wt% H2O), all other CFBs investigated have similar estimated primary H2O contents, with values ranging from 1.0 to 2.0 wt%. The estimated primary H2O content of the Basin and Range basalts is extremely high and up to 10.0 wt%. Thus, this study demonstrates that water flux melting played an important role in the generation of many intra-continental igneous provinces. This new finding was further employed to investigate the tectonic setting of 320–270 Ma basalts in Inner Mongolia, North China. Most basalts from three key rock units (i.e. Amushan, Benbatu, and Dashizhai formations) from the Central Asian Orogenic belt are classified as non-arc types. The estimated magma H2O concentrations suggest a strong link between H2O content and arc-like geochemical signatures. Together with established geological evidence, we proposed that these 320–270 Ma basaltic rocks were most likely produced in a post-orogenic extensional environment facilitated by subducted slab-driven deep-Earth fluid cycling. We propose a mantle transition zone water-filtering model that links deep-Earth fluid cycling, large-scale intra-continental basaltic magmatism, and supercontinent cycles into a self-organized system. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Continental basalts generally show a spectrum of elemental and isotopic compositions that range well beyond those observed in oceanic basalts. Compared with the smooth primitive mantle-normalized trace element distribution patterns (Hofmann, 1997; Sun and McDonough, 1989), the majority of continental basalts have complex trace element ⁎ Corresponding author at: Department of Applied Geology, Curtin University, GPO Box U1987, Perth, WA 6845, Australia. Tel.: +61 8 9266 4125; fax: +61 8 9266 3153. E-mail address: [email protected] (X.-C. Wang).

distribution patterns with pronounced negative and positive anomalies (Ivanov and Litasov, 2013; Jourdan et al., 2007; Puffer, 2001; Wang et al., 2008, 2009, 2014; Xia, 2014). The most intrinsic feature of continental basalts, especially those from large igneous provinces (LIPs), are the negative anomalies of the high-field-strength elements (HFSEs: Nb, Ta, Ti, Zr, and Hf), similar to island arc basalts. Such features will be called arc-like signatures hereafter (they are similar to the low-Ti continental basalts in the literature). Previous studies have demonstrated that continental flood basalts (CFBs), the most volumetrically abundant basalts found in intra-continental areas, plot within an arc tectonic setting in various discrimination diagrams (Duncan, 1987; Xia, 2014).

http://dx.doi.org/10.1016/j.lithos.2015.12.014 0024-4937/© 2015 Elsevier B.V. All rights reserved.

Please cite this article as: Wang, X.-C., et al., Origin of arc-like continental basalts: Implications for deep-Earth fluid cycling and tectonic discrimination, Lithos (2015), http://dx.doi.org/10.1016/j.lithos.2015.12.014

2

X.-C. Wang et al. / Lithos xxx (2015) xxx–xxx

This implies that the use of these diagrams, which are empirically derived, may lead to incorrectly characterizing tectonic setting (Duncan, 1987). Therefore, systemic investigation of how to discriminate arc-like continental flood basalts from island arc basalts is crucial for determining the tectonic setting of ancient continental basalts. The chemistry of basalt is dependent on the composition of its source reservoir, the melting conditions (pressure, temperature, hydrous, or anhydrous) and subsequent effects, including partial melting, melt migration and accumulation, and various assimilation and fractionation crystallization (AFC) processes that it may have undergone on its way to the surface (Duncan, 1987). Only when some, or all, of these are unique to a particular tectonic environment can basalt composition be used as a diagnostic indicator of tectonic setting. Thus, the question of how arc-like trace element distribution patterns are generated becomes crucial to discriminating arc-like continental basalts from true arc basalts generated in a subduction system. One point of general agreement is that the arc-like geochemical signatures of basaltic rocks indicate direct or indirect contributions from water (including other fluid phases) released from subducted slabs (Hawkesworth et al., 1995; Ivanov and Litasov, 2013; Ivanov et al., 2008; Jourdan et al., 2007; Merle et al., 2014; Murphy and Dostal, 2007; Puffer, 2001; Sprung et al., 2007; Ulmer, 2001; Wang et al., 2008, 2009, 2014, 2015; Wilson et al., 1995). Detailed geochemical analyses and petrogenetic studies suggest hydrated, long-term isolated mantle reservoirs account for both the arc-like chemical signature and enriched radiogenic isotope characteristics of arc-like continental basalts (Hawkesworth et al., 1995; Ivanov and Litasov, 2013; Ivanov et al., 2008; Jourdan et al., 2007; Merle et al., 2014; Wang et al., 2008, 2009, 2014). However, the question of how and where (asthenosphere or lithosphere) fluid was released from subducted slabs to influence or control the generation of arc-like signatures is still unclear. Attribution of arc-like geochemical signatures to sub-continental lithospheric mantle (SCLM) modified by paleo-subduction is a conventional interpretation (Hawkesworth et al., 1995; Murphy and Dostal, 2007; Sprung et al., 2007; Wang et al., 2008, 2009, 2014; Wilson et al., 1995). However, discovery of abundant hydrous phases (Pearson et al., 2014; Schmandt et al., 2014) and evidence for high water content in nominally anhydrous minerals (Fukao et al., 2009; Smyth et al., 2003), together with numerical modeling (Bercovici and Karato, 2003; Faccenda et al., 2012; Maruyama and Okamoto, 2007) and geophysical investigations (Fukao et al., 2009; Huang et al., 2005), show that at high rates of subduction, significant amounts of water can subduct as deep as the mantle transition zone (MTZ). Thus, the arc-like signatures may be linked to deep-Earth fluid cycling down to the MTZ (Ivanov and Litasov, 2013; Ivanov et al., 2008). However, arc-like geochemical signatures of continental basalts can also be simply attributed to contamination of asthenospheric mantle-derived melts by lithospheric components (Xia, 2014). Thus, the origin of arc-like geochemical signatures in intra-continental basalts remains unclear. Using available geochemical datasets of global continental flood basalts (Fig. 1a, b, and c), together with rift basalt examples from the Basin and Rang Province (Greater Basin) (GEOROC: http://georoc. mpch-mainz.gwdg.de/georoc/), we attempt to decipher the similarities and differences between island arc basalts and arc-like continental basalts, and to examine which of the most cited geochemical diagrams are successful in discriminating arc-like intra-continental basalts from true island arc basalts. Utilizing our new findings, we have applied them to a controversial area, in this case the Central Asian Orogenic Belt (CAOB; Fig. 1d), in order to test their veracity. 2. Database and methods The data used in this study were obtained from the GEOROC database (access date: December 13, 2014). The continental basalts included in this study are from the Siberian, Central Atlantic, Karoo, Deccan,

Emeishan, and Columbia River (Yellowstone) LIPs and the Basin and Range basalts (Greater Basin) as an example of typical post-orogenic extension-related basalts. Owing to their importance in characterizing water contribution in mantle melting, and in fractionation of light ion lithophile elements (LILEs, such as Ba, Sr, Rb, U, and Th) relative to HFSEs (such as Nb, Ta, Zr, Hf, and Ti), typical arc basalts are also examined. They include the well-studied intra-oceanic arcs (Izu-Bonin, Tonga, and Kermadec arcs) and other arcs that were emplaced within older continental materials or thick sequences of sediments (Andes and Luzon arcs). The data for these five arcs were also obtained from the GEOROC database. The average arc basalt data for global continental and oceanic arcs and 12 individual arcs (Kermadec, Marianas, Tonga, Aleutian, Andes, Central American, Greater Antilles, Luzon, Kamchatka, and Cascades arcs) from Kelemen et al. (2007) are also included in order to estimate the composition of primary arc basaltic melts on a global scale. Finally, for comparison, the estimated primary arc basaltic melts from the Marianas arc (Tamura et al., 2011, 2014) and the average of primary or nearly primary average arc basaltic melt compositions from the Japan, Cascades, Mexico, Indonesia, Izu-Bonin, and Aleutians arcs from Grove et al. (2012) are also included. There is always the issue of quality when using large compiled databases. The data quality depends on the analytical methods and the reputation of the various laboratories for maintaining high standards. In this study, we have not made any attempt to discriminate between “good” and “bad” data, since it is virtually impossible to quantify this. In some cases, where outliers are orders of magnitude different from the bulk of the data, we have ignored these by excluding them from the plots. We have applied three steps in selecting the sample data. Firstly, the major element data set only includes samples with SiO2 ranging from 43 wt% to 56 wt% and LOI values ≤5 wt% (or total summations ranging from 95 wt% to 102%, if LOI values are not available). Secondly, highly evolved samples with MgO ≤8 wt% are excluded in reconstructing the primary major element melt compositions from the data set. Thirdly, bivariate plots of Zr against selected trace elements were used for evaluating the mobility of such elements during alteration (Polat et al., 2002; Wang et al., 2008, 2010). Samples plotting off the main trend (defined by 80–90% of the data) are also excluded in the discussion involving fluid-mobile elements (Ba, Sr, Rb, U, Th, and K). The primary major melt compositions were estimated by adding equilibrium olivine to selected starting material at 1% increments until the resulting basaltic magma was in equilibrium with Fo91. The prerequisite of this method is to determine starting materials that only underwent olivine fractionation. However, fractionation of both clinopyroxene and plagioclase can modify FeO compositions of primary melts (Langmuir et al., 1992; Wang et al., 2012, 2014) and would cause an over- or underestimation of primary MgO contents (Langmuir et al., 1992; Wang et al., 2012). As magmas crystallize at depth, their major and trace element compositions evolve along lines determined by their phase equilibria. These are the so-called liquid lines of descent (LLDs). Fractional crystallization of plagioclase would significantly deplete Al2O3, whereas both olivine and clinopyroxene fractionation will increase Al2O3. Clinopyroxene fractionation would also quickly deplete CaO contents, but both olivine and plagioclase increase CaO contents. Thus, plagioclase and clinopyroxene together in the fractional mineral assemblage would change the slope of the Al2O3\\MgO and CaO\\MgO liquid lines of descent. This implies that the turning points on Al2O3\\MgO and CaO\\MgO liquid lines of descent can be used to determine the appearance of these two minerals in the fractional mineral assemblage. As shown in Fig. 2, Al2O3 contents correlate negatively with MgO at MgO = 3.5–4.0 wt% (Columbia River), 5 wt% (Karoo), 6 wt% (Siberia), 6–6.5 wt% (Basin and Range), and 8.0 wt% (CAMP) and then quickly decrease (become positively) or become flat with decreasing MgO. Al2O3 contents of basalts from the Emeishan (Fig. 2c) and Deccan (Fig. 2d) LIPs continuously correlate negatively with MgO without inflection on the Al2O3\\MgO liquid lines of descent. This indicates



MgO = 8.5% using the method proposed by Kelley et al. (2006). The corrected and uncorrected samples with MgO N 8.5 wt% were therefore selected as starting materials by olivine addition modeling as described in Wang et al. (2012). A series of olivine and basalt compositions were then calculated from starting materials as follows: (1) the composition of equilibrium olivine was obtained using KD (Fe/Mg)oliv/liq = 0.32 (Putirka, 2005) and Doliv/liq (Beattie et al., 1991), assuming that Fe2+/ Ni 2+ 3+ (Fe + Fe ) = 0.90 for intraplate basaltic melts (Kelley and Cottrell, 2009; Lee et al., 2009) and 0.80 for arc basalt melts (Tamura et al., 2011), respectively; (2) a more primitive basalt composition was calculated as a mixture of the basalt and equilibrium olivine in a weight

that the magma systems of these two LIPs significantly depressed plagioclase fractionation. With the exception of basalts from the Deccan LIPs (inflection point at MgO = 8.5 wt%), all the others have inflection points at MgO = 6.0–8.0 wt%. These analyses demonstrate that olivine is the dominant fractionation phase in samples with MgO N 8.0 wt% and that the effect of fractionation of plagioclase and clinopyroxene on whole-rock major element compositions can be effectively ignored. Hence, the samples with MgO N8.5 wt% are only affected by olivine fractional crystallization or accumulation. To minimize the effect of clinopyroxene and plagioclase fractionation, the samples with MgO contents of 8–8.5% were firstly corrected to

180°

150°W

(a)

120°W

90°W

60°W

30°W

3

0°

30°E

60°E

90°E

120°E

150°E

Angayucham 210 Ma

80°N

High Arctic LIP 130 Ma Oslo-JutlandNE Germany 300 Ma

60°N

Uralo-Siberian (Siberian Traps) 250 Ma Tarim 280 Ma Emeishan Panjal Traps 260 Ma 280 Ma

NorthAtlantic LIP 62 Ma

Columbia River 16.5-1.5 Ma

Hess Rise 100 Ma Shatsky Rise 145 Ma

CentralAtlantic LIP (CAMP) 201 Ma Afro-Arabia 31 Ma

Manihiki 120 Ma

180°

High Arctic LIP 130 Ma

40°N

20°N

Ontong Java 122 Ma

Deccan 66 Ma

Caribbean-Colombian -Gorgona 90-88 Ma

0° Karoo 183 Ma

20°S Madagascar 90 Ma

Paraná-Etendeka 134 Ma

Hikurangi 120 Ma

Broken Ridge 90 Ma

40°S Tasman 190 Ma

Kerguelen 120 Ma

Maud Rise 100 Ma

60°S

Ferrar 183 Ma

Dronning Maud Land 182 Ma

Oceanic plateaux/Ocean basin flood basalt provinces

Continental flood basalt/volcanic rifted margins

(b)N ia

Arc-like Siberian CFBs

Nb/La

1

(c)

Deccan

Etendeka

0 0 500 1000 1500 2000 2500 Distance from Mongolia-Okhotsk subduction trench (km)

Rock/Primitive Mantle

im

r Ta

In South Ch do- China ina

Korea

100

No Mongo rth lia Ch ina Paleo-Tethys

Antarctica

Paraná

Sou

10

th A m

ice

Karoo

OIB

Primary arc melts

1

India

Africa

ATS TRM IAB

-alia

Si be r

Mongol- Okhotsk Ocean

OIB

TS

eric

geny e Oro wanid Gond a

0

cover

Austra-

hs

ak

M

ATS

n ta

N P

2

TR

N P TS

z Ka

80°S

2000 km

Primary melts of the Angara-Taseevskaya syncline Ba Th U Nb K

La Ce Sr Zr Hf Sm Ti

Y

Yb

Fig. 1. (a) Global distribution of 0–300 Ma large igneous provinces (LIPs) modified after Bryan and Ernst (2008). (b) Paleotectonic reconstructions of the Siberian Traps and subduction zones in Permo–Triassic time and the relationship between Nb/La variations in the Siberian CFBs and their relative distance to the Mongolia–Okhotsk suture zone. Also shown at the arc-like primary melt compositions of the Angara–Taseevskaya syncline (modified after Ivanov and Litasov, 2013). Green dots indicate localities: from present-day south to north, they are Transbaikalian-rifted margin (TRM), Angara–Taseevskaya syncline (ATS), Tunguska syncline (TS), Putorana (P), and Noril'sk (N). This feature indicates that fluids/melts released from subducted slabs via the Mongolia–Okhotsk subduction zone may play a key role in producing arc-like signatures (insets) in the Siberian CFBs. (c) Continental flood basalt provinces of Gondwanaland prior to breakup, modified after Cox (1978). The “southern” Karoo CFBs mainly display arc-like signatures, whereas the “northern” Karoo CFBs are composed of OIB-like types (Duncan, 1987; Puffer, 2001). This suggests the importance of deep-Earth water cycling by subducted slabs in producing arc-like signatures in the Karoo CFBs. (d) Geological sketch map of central Inner Mongolia showing distribution of Late Carboniferous to early Permian volcanic rocks widely distributed in the eastern segment of the CAOB. Inset shows location within the CAOB (modified after Xu et al., 2013). Locations of the Benbatu (Tang et al., 2011; Pang et al., this issue), Amushan (Liu et al., 2013; Pang et al.'s unpublished) and Dashizhai (Zhang et al., 2008, 2011) volcanic rocks are marked with pink boxes.


4


t o n 110°

111°

113°

mPa

lA

si

O

g

St

cB

t el

ar udy

ic

n

tra

an

ro

i en

cif

46°

Ce

112°

rcu

Si

Cra

Ci

109°

ian ber

114°

115°

116°

117°

118°

119°

Suture line

Dashizhai Formation (Lower Permian)

Sample

46°

Undivided units (Upper Carboniferous-Lower Permian)

Dong Ujimqin

Amushan Formation (Upper Carboniferous)

ea

120°

N

Elitu Formation (Middle Permian)

HolinGola

Benbatu Formation (Upper Carboniferous)

Hegenshan

North China Craton

Xi Ujimqin

Dashizhai volcanic rock Zhang et al. (2011)

100°

Dashizhai volcanic rock Zhang et al. (2008)

44° Sunizuoqi

NOB

Amushan basalt Liu et al. (2015) Pang et al. (unpub.)

Xilin Hot

Eren Hot

44°

Linxi

Benbatu basalt Tang et al. (2011) Pang et al. (2015) Suniyouqi

Solon Obo

SO B Chifeng 42°

42°

North China Craton 109°

110°

111°

112°

113°

114°

115°

116°

117°

118°

119°

(d)

Fig. 1 (continued).

ratio of 99.9:0.1; (3) steps (1) and (2) were repeated using the calculated primitive basalt to obtain more primitive basalt, until equilibrium with olivine Fo91 was attained. The composition of equilibrium olivine was calculated following the method proposed by Tamura et al. (2011). Major element compositions of primary basalt magmas were obtained by olivine addition modeling as described above. Incompatible trace elements were obtained with a Rayleigh crystal fractionation model by using the fraction of liquid (F) obtained from the olivine addition modeling (Tamura et al., 2011; Wang et al., 2012). The olivine-liquid partition coefficient (D) for any incompatible element is never zero; however, D = 0 was used for all our calculations because (1) the difference between trace element values calculated by using D = 0 and D = 0.1 at liquid fraction varying from 0.7 to 0.9 ranges from 3.5% (F = 0.7) to 2.2% (F = 0.9); (2) most F values are ≥0.8, which reduces the differences; (3) most reported D values for these incompatible elements are b0.1, which also makes the differences negligible (Tamura et al., 2011). To evaluate the significance of these results, it is important to consider the uncertainties involved. These mainly concern the composition of selected starting materials, the value of KD, the redox state (Fe3+/total Fe), and mantle Mg#. Some samples with MgO ranging from 8.0 to 8.5 may be modified to some extent by fractional crystallization of plagioclase (Plag) and/or clinopyroxene (Cpx) (Kelley et al., 2006; Wang et al., 2009, 2012). As noted above, the crystallization of Plag ± Cpx would increase FeO contents of the residual liquids, which would result in overestimation of MgO contents in primary melts and in the liquid fraction (Putirka, 2005; Wang et al., 2012). An important variable in fractionation correction is the proportion of Fe3 + relative to Fe2 + in the magma. The Fe3+/∑Fe ratio increases with oxygen fugacity fO2 (Kress and Carmichael, 1991). Typical intraplate magmas (OIB) and

mid-ocean ridge basalts (MORB) have fO2s near the fayalite–magnetite–quartz (FMQ) buffer and hence have Fe3+/∑Fe of ~0.07–0.1 (Lee et al., 2009). Arc magmas are generally thought to be more oxidized although by how much is still debated. Evidence from direct measurement of olivine-hosted melt inclusions suggest an oxygen fugacity with Fe3+/∑Fe ratio of 0.18–0.32 (Kelley and Cottrell, 2009) or 0.19– 0.26 (Brounce et al., 2014). However, recent studies of V/Sc systematics suggest that the mantle source regions of primitive arc magmas may only be slightly more oxidizing (up to 1–2 orders of magnitude greater than the FMQ buffer (Lee et al., 2005)) and that the oxidized nature of erupted magmas themselves may be due to self-oxidation imparted by water dissociation during magmatic differentiation or ascent (Holloway, 2004). Variations in Fe3+/∑Fe can introduce significant differences in estimated compositions of parental liquids, notably Mg#. For a given lava, increasing Fe3+/∑Fe from 0.1 to 0.2 would result in a decrease of about 1.5 wt% MgO for primary melts. Based on the average of less-evolved arc basalt data from Kelemen et al. (2007), increasing Fe3+/∑Fe from 0.2 to 0.3 would result in an increase of 0.3–0.5 wt% SiO2 and 0.6–0.8 wt% Al2O3, but a decrease of ~ 1.0 wt% FeO and 1.7–1.9 wt% MgO in the final estimated primary melts. KD values increase slightly with increasing pressure (Putirka, 2005, 2008) and melt composition (Herzberg and O'Hara, 2002; Lee et al., 2009), typically within the range of 0.30–0.35. For any given basalt, increasing KD from 0.30 to 0.33 would result in an increase of about 2 wt% MgO for the primary magma. Typical mantle Mg# (Mg# = Mg/[Mg + Fe2 +] molar ratio) values are 89–91, but the values do vary. Decreasing mantle Mg# values from 91 to 90 would reduce MgO by about 1.6 wt% for the calculated primary magma. The mantle source of arc basalts may be more highly refractive than intra-continental basalts, suggesting a higher Mg# value up to 93.8 (Tamura et al., 2011, 2014). This implies



25

25

Al2O3 CaO

Plag in MgO = 5.0 wt%

5


20

20 Cpx in MgO = 8.0 wt%

15

15

10

10

5

(a) Karoo

0 0

5

10

15

Cpx in MgO = 8.0 wt%

5


20

(e) Siberia

0 0

25

5

10

MgO (wt%)

15

25

25 20

20

15

15

10


(f) Columbia River

(b) CAMP 5

10

15


5 0

0 0

20

0

2

4

MgO (wt%)

6

8


15

10

10

5

5

(c) Emeishan 0

5

10

12

14

15

Plag in MgO = 6-6.5 wt%

20

15

0

10

MgO (wt%)

25 20

25

Plag in MgO = 3.5-4 wt%


5

20

MgO (wt%)

20

25

MgO (wt%)


(g) Basin and Range 0 0

2

4

6

8

10

12

14

16

MgO (wt%)


20 15 10 5

(d) Deccan 0 0

2

4

6

8

10

12

14

16

MgO (wt%) Fig. 2. Plots of Al2O3 and CaO versus MgO for (a) Karoo, (b) CAMP, (c) Emeishan, (d) Deccan, (e) Siberia, and (f) Columbia River CFBs, and (g) Basin and Range intra-continental basalts. Plag = plagioclase; Cpx = clinopyroxene. Data source: GEOROCK database (http://georoc.mpch-mainz.gwdg.de/georoc/).

that the effect of underestimates for equilibrium with olivine content (Fo91, this study) would result in underestimation of MgO and FeO, but overestimation of SiO2 and Al2O3 contents in the final estimated primary melt compositions, whereas increasing Fe3+/∑Fe from 0.2 to 0.3

can balance this. Thus, considering the compounded effects of the uncertainties of Fe3+/∑Fe ratios, KD values, and mantle Mg#, the final uncertainties are insignificant. For instance, we tested this by adjusting the Fe3+/∑Fe ratios from 0.2 to 0.3 and mantle source Mg# from 91 to


6


a

16

25

Karoo CFBs

12

20

Ar

cm

ag

Al2O3 (wt%)

FeO (wt%)

14

ma

10 8

Arc magma

15

10

5

6

Karoo CFBs 4

0 40

44

48

52

56

40

44

SiO2 (wt%)

48

52

56

SiO2 (wt%)

30

Karoo CFBs

MgO (wt%)

25

Global average of oceanic arc basalt (K07) Global average of continental arc basalt (K07)

Arc

ma

20

Average primary arc basalts from 12 individual arcs (K07)

gm

Primary Mariana arc basalts (T11&T14)

a

Near-primary arc basalts from 6 individual arcs (G12) Primary arc basalts (this study)

15

Primary Karoo continental flood basalts (this study)

10

5 40

42

44

46

48

50

52

54

56

SiO2 (wt%)

b

25

18

Central Atlantic

Central Atlantic

Al2O3 (wt%)

16

FeO (wt%)

14 12 10

20

15

10

8 5 6 0

4 40

44

48

52

56

40

44

48

52

56

SiO2 (wt%)

SiO2 (wt%) 30

Central Atlantic


25

MgO (wt%)

Average primary arc basalts from12 individual arcs (K07) Primary Mariana arc basalts (T11&T14)

20

Near-primary arc basalts from 6 individual arcs (G12) Primary arc basalts (thisstudy)

15

Primary Central A tlantic continental flood basalts (thisstudy)

10

5 40

44

48

52

56

SiO2 (wt%) Fig. 3. Comparing selected major element compositions (FeO, Al2O3, and MgO) of primary melts (in equilibrium with olivine Fo91, see discussion in text) of (a) Karoo, (b) CAMP, (c) Emeishan, (d) Deccan, (e) Siberia, (f) Columbia River (Yellowstone LIP), and (g) Basin and Range basalts with primary arc basalts. The primary melt compositions for global average continental, oceanic arc basalts, and average values for 12 individual arcs were estimated by the olivine-melt equilibrium method (see discussion in text) according to the database of Kelemen et al. (2007). The data for near-primary arc basalts from 6 individual arcs are from Grove et al. (2012). The primary Mariana arc basalts are from Tamura et al. (2011, 2014). The primary CFBs (green circles) and individual arc basalts (yellow circles) are estimated according to GEOROCK database (http://georoc.mpch-mainz.gwdg.de/georoc/). Data source: K07 = Kelemen et al., 2007; G12 = Grove et al., 2012; T11 and T14 = Tamura et al., 2011, 2014.



c

20

25

Emeishan

Emeishan

16

20

Al2O3 (wt%)

FeO (wt%)

7

12

8

4

15

10

5 40

44

48

52

56

40

44

SiO2 (wt%)

48

52

56

SiO2 (wt%)

35

Global average of oceanic arc basalt (K07)

Emeishan

Global average of continental arc basalt (K07) Average primary arc basalts from 12 individual arcs (K07)

MgO (wt%)

25

Primary Mariana arc basalts (T11&T14) Near-primary arc basalts from 6 individual arcs (G12) Primary arc basalts (this study) Primary Emeishan continental flood basalts (this study)

15

5 40

44

48

52

56

SiO2 (wt%)

d

25

Deccan

Deccan 16

Al2O3 (wt%)

FeO (wt%)

20 12

8

15

10

4

5 40

44

48

52

56

40

44

48

52

56

SiO2 (wt%)

SiO2 (wt%) 35

Deccan


MgO (wt%)

Average primary arc basalts from 12 individual arcs (K07)

25

Primary Mariana arc basalts (T11&T14) Near-primary arc basalts from 6 individual arcs (G12) Primary arc basalts (this study) Primary Deccan continental flood basalts (this study)

15

5 40

44

48

52

56

SiO2 (wt%) Fig. 3 (continued).


8


e

25

Siberian

Siberian

16

Al2O3 (wt%)

FeO (wt%)

20 12

8

15

10

4

5 40

44

48

52

56

40

44

SiO2 (wt%)

48

52

56

SiO2 (wt%)

35

Siberian

Global average of oceanic arc basalt (K07) Global average of continental arc basalt (K07) Average primary arc basalts from 12 individual arcs (K07)

MgO (wt%)

25

Primary Mariana arc basalts (T11&T14) Near-primary arc basalts from 6 individual arcs (G12) Primary arc basalts (this study) Primary Siberia continental flood basalts (this study)

15

5 40

44

48

52

56

SiO2 (wt%) 16

20

Columbia River Al2O3 (wt%)

FeO (wt%)

f

12

8

Columbia River

15

10

4

5 42

46

50

54

42

SiO2 (wt%) 25

46

50

54

SiO2 (wt%)

Columbia River


MgO (wt%)

20

Average primary arc basalts from 12 individual arcs (K07) Primary Mariana arc basalts (T11&T14) Near-primary arc basalts from 6 individual arcs (G12)

15

Primary arc basalts (this study) Primary ColumbiaRiver continental flood basalts (this study)

10

5 42

46

50

54




g

9

20

Basin and Range

Al2O3 (wt%)

FeO (wt%)

16

12

8

15

10

Basin and Range 5

4 40

44

48

52

56

40

SiO2 (wt%)

44

48

52

56

SiO2 (wt%)

30

Basin and Range


25

MgO (wt%)

Average primary arc basalts from 12 individual arcs (K07) Primary Mariana arc basalts (T11&T14)

20

Near-primary arc basalts from 6 individual arcs (G12) Primary arc basalts (this study)

15

Primary Basin and Range basalts (this study)

10

5 40

44

48

52

56


93. Based on the same data set, this adjustment resulted in a decrease of ~0.7 wt% Al2O3, 0.5 wt% SiO2, and 0.8 wt% FeO, but increase of 2.0 wt% MgO in the final estimated primary melt. Because the aim of this correction is to strip-off the effect of olivine crystallization and significant crustal contaminations on the major and trace element composition and then to be normalized to the same criteria, the difference between the estimated compositions and the true primary melts is insignificant.

3. Characteristics of typical continental basalt

The other estimated primary major element compositions, including CaO, TiO2, Na2O, and K2O, are undistinguished between CFBs, riftrelated basalts (Basin and Range), and arc basalts (figures not shown). In summary, estimated primary basaltic melts of CFBs generally displayed higher FeO and MgO contents but lower Al2O3 than those of arc basalts. The estimated primary basaltic melt compositions of rift-related basalts, such as the Basin and Range basalts, may be indistinguishable from those of arc basalts. However, although some CFBs, such as the Columbia River and Central Atlantic CFBs, show largely overlap the estimated arc basaltic melt compositions at SiO2 contents greater than about 46%, they may be distinguishable at lower SiO2 contents.

3.1. Continental basalts and arc basalts 3.1.1. Major elements We first examine the major element variation as a function of SiO2 content. As shown in Fig. 3, the estimated primary melts for the Karoo, Deccan, and Emeishan CFBs have higher FeO and MgO, but lower Al2O3 contents, than those of estimated primary arc melts at a given SiO2 content. The estimated primary melts for the Central Atlantic CFBs have FeO and MgO contents higher than those of primary arc melts at SiO2 b 46 wt%, but largely overlap with those of estimated primary arc melts at SiO2 N 46 wt% (Fig. 3b). The estimated primary Al2O3 for the Central Atlantic CFBs is significantly lower than that of primary arc melts (Fig. 3b). The Siberian CFBs have primary FeO and MgO contents higher than those of primary arc melts, whereas their Al2O3 contents largely overlap at SiO2 = 44–48 wt% (Fig. 3e). The Columbia River CFBs have primary FeO and MgO contents higher than primary arc melts at SiO2 b 46 wt% but are very similar to the latter at SiO2 N 46 wt% (Fig. 3f). More importantly, the estimated primary Al2O3 is very similar for the Columbia River CFBs and arc basalts (Fig. 3f). The Basin and Range basalts and arc basalts show the similar primary FeO, MgO, and Al2O3 contents (Fig. 3g).

3.1.2. Trace elements The primitive mantle-normalized trace element distribution patterns for primary basaltic melts (equilibrium with olivine Fo91) are presented in Figs. 4 and 5. The primary island-arc melts are generally characterized by enrichment in LILEs and light-rare earth elements (LREE), and depletion of HFSEs (Nb, Zr, Hf, and Ti) and heavy rare earth elements (HREE). The intra-oceanic arc basalts (Tonga, Kermadec, and Izu-Bonin; Fig. 4a, b, and d) have much lower incompatible trace element contents than those of continental arcs that were emplaced within older continental materials or thick sequences of sediments (Andes and Luzon arcs; Fig. 4c and e). Note that arc basalts with variable Nbenrichment and with or without slight HFSEs depletion are occasionally found in island-arc environments (Sorbadere et al., 2013 and references therein). For example, arc basalts from the Andes and Luzon arcs can be further divided into two subtypes: Nb-depleted and Nb-enriched (Fig. 4c and e). The Nb-enriched arc basalts display weak or no Nb anomalies with enriched MORB (E-MORB)- or even OIB-type trace element distribution patterns (Fig. 4c and e). The origin of Nb-enriched arc basalts is highly controversial (Sorbadere et al., 2013 and references therein). It may be attributed to a mixture of enriched OIB-like and


10


100

K

100

Pb Sr

K

Pb Sr

U

10

10

1

1

Zr Hf Zr Hf

Nb

Nb

(a) Tonga arc

0.1

(d) Izu-Bonin arc

0.1

Rb Th Nb La Pb Sr Zr Sm Ti Tb Y Er Yb Ba U K Ce Pr Nd Hf Eu Gd Dy Ho Tm Lu


100

100

Pb K

Sr

Pb Sr

10

10

Ba U

K

Nb

Zr Hf Ti

1

1

Zr Hf

Nb

(b) Kermadec arc

0.1

200

(e) Luzon arc

0.1



200 100

Pb

100

Individual primary arc basalt (K07)

Pb

Kermadec Mariana Tonga Aleutian Andes

Sr

Sr 10

Central American Greater Antilles Luzon Kamchatka Cascades

10 Zr Hf Nb

1

1

Zr Hf

Ti

Marian (G11)

Nb 0.1

(c) Andes arc Rb Th Nb La Pb Sr Zr Sm Ti Tb Y Er Yb Ba U K Ce Pr Nd Hf Eu Gd Dy Ho Tm Lu

0.1

(f) Average

Marian (G14) Global average continental (K07) Global average oceanic (K07)


Fig. 4. Primitive mantle-normalized trace element distribution patterns of primary (equilibrium with olivine Fo91, see discussion in text) arc basalts for (a) Tonga, (b) Kermadec, (c) Andes, (d) Izu-Bonin, (e) Luzon arcs, and (f) global average of continental, oceanic, and individual primary arc basalts. The yellow area in (f) outlines the field of primary arc basalts defined by (a)–(e). The data sources for arc basalts are the same as in Fig.2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

depleted MORB-type upper mantle (Hastie et al., 2011; Sorbadere et al., 2013), metasomatism of a mantle wedge by slab melts (Hastie et al., 2011), or decompression melting of a mantle source that has not undergone previous fluid-flux melting (Johnson et al., 2009). The hydrous and decompression (anhydrous) melting processes can occur within a single arc, as evidenced by the close spatial–temporal association of both hydrous and anhydrous arc magmas in several arc settings (Bartels et al., 1991; Tatsumi et al., 1983). Recent identification of the local coexistence of island-arc-type basalts with prominent negative HFSEs anomalies and Nb-enrichment without HFSEs depletion in the central Vanuatu arc demonstrates that these two distinct types of primitive arc basalts can co-exist at the scale of one volcanic island and within a relatively short-time span (Sorbadere et al., 2013). Geochemical analyses show that the agent that resulted in enriching the mantle source of these arc basalts played a key role in defining the final distinctive trace element distribution

patterns. The depletion of HFSEs was caused by hydrous enrichment of the mantle source, whereas the contribution of slab-derived fluid components in formation of E-MORB-type arc basalts can be negligible (Sorbadere et al., 2013). This implies that volatile flux from the subarc mantle is heterogeneous (Parman et al., 2011; Sorbadere et al., 2013). Although the E-MORB-type arc basalts exhibit no negative Nb anomalies with Nb* [Nb* = 2 × NbN/(ThN + LaN), where subscript N indicates primitive mantle-normalized value] ranging from 0.9 to 1.4, they do display visible negative Zr anomalies with Zr* [Zr* = 2 × ZrN/ (SmN + NdN)] varying from 0.6 to 0.9. This implies that fluid released from the subducted slabs was most likely a controlling factor to fractionation of Nb from La and Th and the depletion of Nb (Ta) relative to La and Th is therefore not a good diagnostic index to discriminate arc from non-arc basalts. The overall level of incompatible trace element content of arc basaltic melts is highly variable, but similar to those of typical continental



basalts (Fig. 5). This strongly argues against the Xia's (2014) conclusion that the overall level of incompatible elements, including Nb is higher for contaminated continental basalts (arc-like continental basalts) 100

Ba

Pb

Average continental arc basalt Average oceanic arc basalt

K Sr

11

than in arc basalts. This is because Xia's (2014) conclusion is only based on intra-oceanic arc basalts (Tonga, Kermadec, and Izu-Bonin arcs), which represent the low-concentration end-member (Fig. 4a, b, 100 Ba

Pb

U

Zr-Hf depletion of continental basalt

K

Sr

10

10

Zr Hf

1

Zr Hf

Ti

Nb

1

Th

(a) Karoo

0.1

(e) Siberian

0.1



100

100

Pb

Ba

Pb

U

Ba

Ti

Nb

K Nb

Sr

Sr

10

10

1

1

Zr Hf Th

Nb Zr Hf

Ti

(b) Central Atlantic

0.1

Th

Nb

(f) Columbia River

0.1



1000

1000

Ba U Pb

100

Ba

Pb K

Sr

100

K Sr

10

10

1

Th Nb

Zr Hf

Nb

Ti

Zr Hf Ti

1

Th

(g) Basin and Range (c ) Emeishan

0.1


Greater Basin

0.1 Rb Th Nb La Pb Sr Zr Sm Ti Tb Y Er Yb Ba U K Ce Pr Nd Hf Eu Gd Dy Ho Tm Lu

1000 Pb

100

Ba K Sr

10 Nb

Zr Hf

Ti

1

0.1

(d) Deccan Rb Th Nb La Pb Sr Zr Sm Ti Tb Y Er Yb Ba U K Ce Pr Nd Hf Eu Gd Dy Ho Tm Lu

Fig. 5. Primitive normalized trace element distribution patterns of estimated primary melt composition (equilibrium with olivine Fo91, see discussion in text) for (a) Karoo, (b) Central Atlantic, (c) Emeishan, (d) Deccan, (e) Siberian, (f) Columbia River CFBs, and (g) Basin and Range basalts. The yellow zones indicate the range of primary arc basalts according to Fig. 4. The global average of primary continental and oceanic arc basalts (Kelemen et al., 2007) are also shown.


12


and d). Consequently, the overall level of incompatible elements cannot be used as a diagnostic feature to discriminate arc-like continental basalts and arc basalts, as proposed by Xia (2014). Most arc basalts are characterized by prominent negative Zr\\Hf and positive Sr anomalies in primitive normalized trace element patterns, whereas most arc-like continental basalts do not display these features. Thus, the presence of coupled prominent negative Zr and Hf and positive Sr anomalies can be used to discriminate arc-like continental basalts from true arc basalts. It should be pointed out that a few basalts from the Karoo (Fig. 5a), Central Atlantic (Fig. 5b), and Emeishan (Fig. 5c) LIPs, and most Basin and Range basalts (Fig. 5g), display negative Zr\\Hf anomalies. This may be attributed to water flux melting in the intra-continental setting. 3.2. Arc-like versus true arc basalts: Re-evaluation of geochemical discrimination diagrams Continental basalts are mainly of arc-like type. For example, arc-like basalts are the dominant component in the Siberian LIP, one of the most voluminous volcanic provinces on Earth (Ivanov et al., 2008), and make up 80%, 99%, and 50% by volume of the classical Noril'sk lava sequence, the west Siberian basin, and Maymecha–Kotuy units, respectively. The Central Atlantic CFBs are also largely composed of arc-like type basalt, characterized by low titanium content, enrichment in the most incompatible elements, and showing negative Nb anomalies (Merle et al., 2014). The southern Karoo LIP (such as the main Lesotho sequence) is mainly composed of arc-like basalts (Duncan, 1987; Puffer, 2001). Furthermore, geochemical analysis of basaltic lava flows, sills, and dykes from the Karoo LIP show that both high- and low-Ti basalts display enrichment in LREE relative to HREE and HFSEs and have prominent negative Nb anomalies, with some depletion of Zr\\Hf (Jourdan et al., 2007). As shown in Fig. 5, the global CFBs from the Karoo, Central Atlantic, Emeishan, Deccan, Siberian, and Columbia River LIPs and the Basin and Range basalts display arc-like geochemical signatures (depletion of Nb\\Ta\\Ti). It has been demonstrated that some often-cited geochemical discrimination diagrams result in misidentification of arclike intra-continental basaltic rocks as arc-related ones (Duncan, 1987; Wang et al., 2009; Xia, 2014), especially for ancient remnants of CFB provinces (Wang et al., 2008, 2009). Thus, the question of how to distinguish arc-like from true arc basalts using geochemical discrimination diagrams is crucial for investigating ancient continental basalts. Because continental crustal material is generally characterized by depletion of HFSEs relative to LILEs (Rudnick and Gao, 2003) and crustal contamination is an unavoidable process in the formation of continental basalts, we need to examine the effect of crustal contamination on parental magma compositions. Incompatible trace element ratios, such as Sm/Nd and Nb/La and radiogenic isotope compositions, are sensitive to both source heterogeneity and crustal contamination but insensitive to fractional crystallization. Major element compositions are sensitive to fractional crystallization and assimilation (AFC) processes and thus provide key constraints on the mass balance of input of crustal materials. The average compositions of alkali OIB and N-MORB were chosen to represent depleted and enriched types of mantle melts. Binary mixing between primary melts and upper continental crust was performed to examine the potential effect of crustal contamination on chemical diversity of intra-continental basalts. Crustal materials are characterized by low Nb/La, Sm/Nd, and MgO and high SiO2 and unradiogenic Nd isotopes, whereas asthenospheric mantle melts have high Nb/La, Sm/Nd, and MgO, and low SiO2 and radiogenic Nd isotopes. Thus, significant crustal contamination would produce positive correlations between incompatible trace element ratios (Nb/La and Sm/Nd) and major elements (MgO and SiO2) and εNd(t) values. As shown for the Basin and Range basalts in Fig. 6, the maximum values of Nb/La and Sm/Nd ratios appear at MgO = 7.0–8.0 wt%, and these two ratios correlate positively with εNd(t). This implies that samples with MgO b 8.0 wt% most likely underwent significant crustal contamination. Similar trends

are observed in six LIPs where the Nb/La and Sm/Nd ratios reach the maximum values at MgO = 5.0–8.0 wt% and correlate positively with εNd(t) (Appendix Figs. 1–6). Along with fractional crystallization analyses (Fig. 2), basalts with MgO ≥ 8 wt% were less contaminated, whereas samples with MgO b 8 wt% were highly evolved. Thus, the distribution patterns on geochemical discrimination diagrams defined by less- and more-contaminated basalts would provide a first-order constraint on the role of crustal contamination in classifying arc-like intra-continental basalts (Figs. 6-13). Xia (2014) used the Nb/La ratio as a key factor to discriminate contaminated (Nb/La b 1) and uncontaminated (Nb/La N 1) continental basalts and further applied the geochemical discrimination diagrams to uncontaminated basalts. Based on this, the author proposed that “uncontaminated” plume-derived continental basaltic rocks (CFBs) are normally characterized by high Nb/La ratios and “hump-shaped” OIB-like mantle-normalized multi-element patterns. This led him to attribute the arc-like geochemical features of continental basalts to lithosphere contamination. However, the interpretation of “contamination” is inconsistent with the following observations. Firstly, both high- and low-Ti CFBs possess significant Nb\\Ta and to some extent Zr\\Hf negative anomalies in the Karoo LIP (Fig. 5a). More importantly, the high-magnesium basalts (picrites) from the Karoo and Siberian LIPs also display pronounced Nb\\Zr\\Hf negative anomalies (Ivanov et al., 2008; Jourdan et al., 2007). This implies that arc-like geochemical signatures indeed reflect their source rather than the result of contamination (Hawkesworth et al., 1995; Jourdan et al., 2007; Merle et al., 2014; Murphy and Dostal, 2007; Puffer, 2001; Sprung et al., 2007; Wang et al., 2008, 2009, 2014; Wilson et al., 1995). Secondly, the contamination model is also inconsistent with the observations that arc-like and OIB-like CFBs have distinct geographic distributions in the Parana (Campbell, 2007), Karoo (Duncan, 1987), and Siberia (Ivanov and Litasov, 2013) (Fig. 1b) LIPs. Thirdly, the ratios of HFSE/ LILE, such as Nb/La, are more sensitive to hydrous enrichment of the mantle source rather than crustal contamination processes. Hydrous enrichment of the mantle source would elevate LILE content but not affect the HFSE contents, which would result in fractionation between HFSEs and LILEs. Thus, the ratio of Nb/La cannot be used as a diagnostic index to discriminate lithosphere contamination (Wang et al., 2008). Fourthly, both the Basin and Range basalts (Fig. 6) and CFBs from the six LIPs (appendix Figs. 1–6) define broad positive correlations and no inflection points are observed on the Nb/La − εNd(t) trends. This implies that the Nb/La ratio cannot be used to discriminate contaminated (Nb/La b 1) and uncontaminated (Nb/La N 1) continental basalts as Xia (2014) suggested. In addition, if low Nb/La ratios were indeed the result of crustal contamination, a tight correlation between Nb/La and other geochemical indices of crustal contamination, such as Th/Ta, La/Sm, and SiO2, should be observed. However, only the Deccan CFBs define a correlation between Nb/La and SiO2 and La/Sm (Fig. 10g and h). Fifthly, if the crustal contamination is a key factor in generating the observed arc-like geochemical signatures, the less-evolved and evolved samples should have distinctive patterns on geochemical discrimination diagrams, which they do not as our analyses will show. With the exception of the Deccan CFBs (Fig. 10), as shown in Figs. 7–9 and 11–13, the whole database and the less-evolved samples (MgO N 8 wt%) display very similar distribution patterns on the Nb/ 16\\Hf/3\\Th (Wood, 1980), Ti/1000\\V [modified by Rollinson (1993) based on Shervais (1982)], Zr\\Ti (Pearce, 1996), and Zr\\Zr/Y (Pearce and Norry, 1979) discrimination diagrams. Together with the lack of significant correlations between Nb/La ratios and other geochemical indices of crustal contamination (SiO2 and La/Sm; figures not shown), we emphasize that although continental crustal contamination can impart arc-like signatures to asthenospheric melts, it does not play a key role in generating the arc-like signatures in intra-continental basalts. In contrast with most studied CFBs, the arc-like geochemical signatures of the Deccan CFBs were most likely caused by crustal contamination, as evidenced by the following observations. Firstly, the less-evolved



0.40

13

0.40

Basin and Range

N - MORB

N - MORB 0.30

OIB

LC

Sm/Nd

Sm/Nd

0.30

UC

0.20

(a)

0.10 0

5

10

15

LC

OIB

UC

0.20

(d)

0.10

20

-30

-20

-10

0

10

20

εNd(t)

MgO (wt%) 2.0

3.0

1.5

Nb/La

Nb/La

2.0

OIB 1.0

N - MORB

LC

0

5

10

1.0

15

LC

UC

0.5

(b)

UC

0.0

OIB N - MORB

(e)

0.0

20

-30

-20

-10

0

10

20

εNd(t)

MgO (wt%) 20

70

OIB

UC

65

55

LC

MgO (wt%)

SiO2 (wt%)

15 60

N - MORB

50 45

OIB

N - MORB 10

5

LC

(c)

-30

-20

-10 0 εNd(t)

10

20

(f)

UC

0

40

-30

-20

-10

0

10

20

εNd(t)

Fig. 6. Plots of (a) Sm/Nd and (b) Nb/La versus MgO and (c) SiO2, (d) Sm/Nd, (e) Nb/La, (f) MgO versus εNd(t) for the Basin and Range basalts. The average values for normal mid-ocean ridge basalts (N-MORB) and alkali ocean-island basalt (OIB) are from Sun and McDonough (1989). The data for average upper continental crust (UC) and lower continental crust (LC) are from Rudnick and Gao (2003). The two-end-member mixing between OIB and N-MORB with upper continental crust is also shown. Each cross on the binary mixing lines indicates 0.1 increments of upper continental crust. The data for the Basin and Range basalts are from GEOROCK database (http://georoc.mpch-mainz.gwdg.de/georoc/).

Deccan CFBs (MgO ≥ 8 wt%) are dominantly intraplate types, whereas the highly evolved samples (MgO b 8 wt%) plot mainly in the arc basalt field (Fig. 10a and b). Secondly, the Nb/La ratios of the Deccan CFBs correlate with SiO2 and La/Sm (Fig. 10g and h). This implies that magma evolution resulted from crustal contamination, shifting from the intraplate field to the arc field. However, all other studied CFBs and the Basin and Range basalts display no such correlation between Nb/La and other indices of crustal contamination. We also examined the success rate of geochemical discrimination diagrams in classifying arc-like continental basalts and arc basalts. The global CFBs and the Basin and Range basalts are dominantly classified as non-arc types according to Ti/1000\\V and Zr\\Zr/Y diagrams (Figs. 7–13). These discrimination diagrams misclassify only about 10–15% of the continental basalts. Based on the Zr\\Zr/Y discrimination diagram, the Emeishan (87%), Deccan (92%), and Columbia River (85%) CFBs and the Basin and Range basalts (88%) also plot mainly in the intraplate basalt field or are offset from the volcanic arc basalt field (VAB) (Figs. 9e, 10e, 12d, and 13e). However, a high proportion of the other three CFBs (33% for Karoo, 40% for Central Atlantic Magmatic Province (CAMP), and 30% for Siberia) are misclassified as VAB (Figs. 7e, 8e, and 11e). In contrast, the Nb/16–Hf/3–Th discrimination diagram misclassified virtually all of the global CFBs and the Basin and Range

basalts as arc basalts. Thus, we suggest that the best discrimination diagrams for classifying arc-like intra-continental basalts are Ti\\V and Zr\\Zr/Y, whereas discrimination diagrams based on Nb and/or Ta cannot be used to classify the tectonic setting of ancient continental basalts, especially for remnants of ancient LIPs. The primary melts show distinctive Zr, Sr, and Ti compositions between arc-like continental and arc basalts (Figs. 4 and 5). We therefore designed a ternary discrimination diagram using the trace element ratios of Ti/V, Zr/Sm, and Sr/Nd (Fig. 7f). The arc-like continental basalts define a trend mostly along the Ti/V\\Zr/Sm boundary, whereas the arc basalts define a trend toward Sr enrichment that is nearly perpendicular to the trend of arc-like continental basalts (Figs. 7–13). Because the effect of crystal fractionation/accumulation of olivine (plus clinopyroxene) in basaltic magma systems on these three trace element ratios can be ignored, the advantage of this diagram is that it strips off the effect of early magma evolution caused by fractionation or accumulation of olivine (spinel) and clinopyroxene on discrimination diagrams. The early fractionation of spinel and clinopyroxene may significantly affect the Ti contents of the liquid, which could result in misclassification in Ti\\V and Zr\\Ti diagrams. Because the ratios of Ti/V, Zr/Sm, and Sr/Nd in basaltic magma systems are insensitive to magma evolution caused by crystal fractionation, but


14


(a)

(d)

Hf/3

50000 Within Plate Basalt

N

-M

O

E-

R

M

B

O

A

W

IC

Ti (ppm)

IA T

All

R

B

10000

33%

PT Volcanic Arc Basalt

1000 10

W PA

Th

Nb/16

(b)

(e)

Hf/3

20 10

N

-M

O

E-

R

M

B

O

IC

A

W

R

B

WPB

PT

MORB

IAB W PA

1 10

100

1000

Zr (ppm)

Th

Nb/16

Zr/Sm

(f)

600 Arc tholeiite

MORB

CFBs

500

V (ppm)

1000

Zr/Y

IA T

MgO>8%

(c)

100

Zr (ppm)

400

Karoo all data MgO>8.0 wt% Arc

300 200 OIB

100

Calc-alkaline basalts

Arc

0 0

5

10

15

20

25

Sr/Nd

Tr e

nd

Ti/V

Ti/1000 Fig. 7. Plots of (a)–(b) Nb/16\ \Hf/3\ \Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr\ \Zr/Y (Pearce and Norry, 1979), and (f) Ti/V\ \Zr/Sm\ \Sr/Nd for the Karoo CFBs. The full dataset and less-evolved samples with MgO N 8.0 wt% show similar distribution features in (a) - (f). This strongly argues against the arc-like signatures of CFBs being mainly caused by crustalcontamination as proposed by Xia (2014).

sensitive to arc or intraplate origin, we therefore suggest that it is better to integrate the diagrams based on element concentration (Ti\\V and Zr\\Ti diagrams) and those using element ratios (such as the Ti/V\\Zr/Sm\\Sr/Nd diagram). Arc-like intra-continental basalts can be distinguished from arc basalts by the following factors: (1) Ti\\V, Zr\\Zr/Y, Zr\\Ti, and Ti/ V\\Zr/Sm\\Sr/Nd discrimination diagrams, which can distinguish arclike continental basalts from arc basalts; (2) the coexistence of arc-like and OIB-like subtypes within a single unit; and (3) primitive mantlenormalized trace element distribution patterns with or without prominent Zr\\Hf depletion and Sr enrichment; (4) primary melts of intra-continental arc-like basalts have higher concentrations of FeO, MgO, and Al2O3 than those of primary arc basaltic melts at the same silica contents. The coexistence of arc-like and OIB-like continental basalts within the same province can provide an important constraint on their intraplate tectonic affinity (Wang et al., 2008, 2014), regardless the relative proportion of arc- and OIB-like phases. The size of the database is also crucial for evaluating the validity of the results. Note that most of the currently cited geochemical discrimination diagrams are only based on a very small dataset. This study found

that when the size of the database was enlarged, the boundary between the arc and non-arc basalt fields (MORB, OIB, and intraplate types) becomes less distinctive. The larger the size of the data set, the higher the proportion of samples that overlap. Some arc basalts that may be induced by slab-tearing and ridge subduction are indistinguishable from typical intraplate types (such as some from the Andes and Luzon arcs). Thus, results based on limited data may result in misinterpretation of the tectonic setting. For instance, Xia (2014) proposed that the overall level of incompatible elements, including Nb, is higher in arclike continental basalts than in true arc basalts and can thus be used as a key indicator to distinguish them. However, this study has found that this is true only for intra-oceanic arc basalts. Other arc basalts, such as those from the Andes and Luzon arcs, are similar or comparable to CFBs. Therefore, there are two conditions for applying any geochemical discrimination diagram to constrain the tectonic setting of ancient rocks. One requires examination of the controlling factors (e.g. source composition, melting conditions and processes, and magma evolution) that generate the geochemical index that is employed in the tectonic discrimination diagram. The other is that the data set must characterize the entire magmatic province/rock units. Finally, the results based on



(a)

(d) 50000

Hf/3

Within Plate Basalts

N

E-

M

B

O

W

A IC

-M

R

Ti (ppm)

IA T

All O

R

15

B

10000

45%

PT

D

Volcanic Arc Basalts

1000 10

W PA

Th

100

Nb/16

(e) 20

Hf/3

(b)

1000

Zr (ppm)

MgO>8.0 wt%

10

N

-M

O

R

M E-

B

O

R

W

B

MORB

PT

IC

A

Zr/Y

IA T

WPB

IAB W PA

1 10

Th

(c) 600

Nb/16

(f)

MORB

Arc tholeiite

100

1000

Zr (ppm)

Zr/Sm CAMP all data MgO>8.0 wt% Arc

V (ppm)

500 400

CFBs

300

Ka

200

OIB

ro

o

Calc-alkaline basalts

100 0 0

5

10

15

20

Arc

25

Ti/1000

Sr/Nd

Tr e n

d

Ti/V

Fig. 8. Plots of (a)–(b) Nb/16—Hf/3—Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr\ \Zr/Y (Pearce and Norry, 1979), and (f) Ti/V\ \Zr/Sm\ \Sr/Nd for the Central Atlantic CFBs. The diamonds in (b) to (f) indicate less-evolved samples with MgO N 8 wt%.

any geochemical discrimination diagrams are only complementary evidence for constraining tectonic setting and should always be carried out in association with petrology, structural geology, and other independent geological evidence. 3.3. Arc-like geochemical signatures in intra-continental basalts: a result of deep-Earth fluid cycling? 3.3.1. Enrichment in fluid-mobile elements and Th relative to HFSE and LREE: A common feature for both intra-continental and arc basalts The geochemical fingerprint of arc magmas is considered to originate from a mixture of subducted slab-derived fluids/melts, the subducted sedimentary veneer and depleted MORB-type mantle (Ellam and Hawkesworth, 1988; Elliott et al., 1997; Grove et al., 2012; Kelemen et al., 2007; Kelley et al., 2005; Marschall and Schumacher, 2012; Plank and Langmuir, 1998; Sorbadere et al., 2013). This geochemical fingerprint not only led to the establishment of three-component mixing models to explain the composition of magmatic rocks produced at continental-arc and island-arc settings (Ellam and Hawkesworth, 1988; Kelemen et al., 2007) but also provided an important reference

framework to examine the origin of arc-like geochemical signatures in an intra-continental setting. Water plays an important role in generating arc basalts because fluids and/or hydrous melts released from the subducted slab rise into the mantle wedge, lowering its melting temperature and thus triggering magma generation (Bebout, 2007; Grove et al., 2012; Johnson et al., 2009; Kelemen et al., 2007; Plank et al., 2013). The enrichment of fluid-mobile elements (Ba, Rb, Sr, U, K, and Pb) is generally attributed to aqueous fluids released from the slabs (Deschamps et al., 2012; Shervais and Jean, 2012; Tonarini et al., 2001; You et al., 1996). The contribution of partial melting of subducted sediments is generally recognized by enrichment of Th relative to LREE or Nb (e.g., Th/La, Th/Nb) and enrichment of light/middle REEs (e.g., La/Sm) (Bebout, 2007; Plank, 2005; Plank and Langmuir, 1998). Some high thorium primitive arc magmas display MORB-like Pb, Nd, and Sr isotope compositions, such as primitive andesites from the western Aleutians (Kelemen et al., 2007). This implies that the enrichment of Th in these arc magmas may not simply be attributable to incorporation of sedimentary Th. Considering the highly incompatible behavior of thorium during partial melting, Kelemen et al. (2007) attributed Th enrichment


16


Hf/3

(a)

(d) 50000 Within Plate Basalt

N

-M

O

E-

R

M

B

O

IC

A

W

R

Ti (ppm)

IA T

All B

PT

10000

13%

VAB

W PA

1000 10

Th

100

Nb/16 Hf/3

(b)

1000

Zr (ppm)

(e)

20

MgO>8.0 wt%

N

-M

O

R

M E-

B

O

IC

A

W

R

WPB

Zr/Y

IA T

10 B

PT

MORB

IAB

W PA

1 10

Th

100

Nb/16

(c) 600

Arc tholeiite

MORB

(f)

Zr/Sm

CFBs

Emeishan all data MgO>8.0 wt% Arc

500 400

V (ppm)

1000

Zr (ppm)

300 200

OIBs

100 Calc-alkaline basalts

Arc

0 0

5

10

15

20

25

Ti/1000

Sr/Nd

Tr e n

d

Ti/V

Fig. 9. Plots of (a)–(b) Nb/16\ \Hf/3\ \Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr/Y\ \Zr (Pearce and Norry, 1979), and (f) Ti/V\ \Zr/Sm\ \Sr/Nd for the Emeishan CFBs. The full data set and less-evolved samples show similar distribution features in (a)–(f). This strongly argued against that the arc-like signatures of CFBs are mainly caused by crustal contamination proposed by Xia (2014). The diamond signs in (b) to (f) indicate less-evolved samples with MgO N 8 wt%.

in arc magmas to partial melting of subducted basalts with or without sedimentary input. The contribution of MORB-type mantle and Nbenriched components (called the anhydrous-enriched mantle reservoir hereafter) in the formation of arc magmas can be recognized by enrichment of Nb relative to LREE. In this study, the trace element ratios with similar incompatibility (in order to remove the effects of crystal fractionation) during mantle melting were employed to examine the contributions of a fluid component (component-1), a low-degree partial melt from subducted oceanic crust (component-2), and other mantle reservoirs (MORB-source-type or anhydrous-enriched mantle reservoir, component-3). The tight correlations defined by typical arc basalts (with MgO N 8 wt%) from Tonga, Kermadec, Izu-Bonin, and Luzon arcs (Fig. 14) demonstrate that the above-mentioned three end-member components can be recognized by ratios of fluid-mobile elements and Th/Nb and Nb/La ratios. As shown in Fig. 14, the low-Nb/La end-member melt displays high Ba/Nb, Sr/Nb, Rb/Nb, U/Nb, Th/Nb, and K/Nb ratios, a typical characteristic of hydrated mantle reservoirs. Furthermore, because the effects of partial melting processes on fractionation of U/Nb are

negligible, the primary U/Nb ratio reflects the nature of source reservoir (Hofmann, 1997). The coupling of high U/Nb and low Nb/La ratios in low-Nb/La end-member melts most likely reflects the contribution of a fluid component in fractionating U/Nb and Nb/La. Thus, hydration of the mantle source may result in HFSE depletion and fluid-mobile element enrichment (arc-like geochemical signatures). We then examined the geochemical composition of continental basalts in the above-mentioned diagrams. Considering mantle source heterogeneity, the average of the end-member OIB mantle reservoir is also included. As shown in Figs. 15–21, low-Nb/La continental basalts are generally characterized by enrichment in fluid-mobile elements and Th, plotting beyond the OIB end-member mantle reservoir. Although the magnitude of enrichment of fluid-mobile elements and Th is variable in different basalt provinces, the trends presented in Figs. 15–21 are very similar to those of arc basalts (Fig. 14). Thus, both intra-continental basalts and arc basalts are characterized by coupling between HFSE depletion (characterizing by low Nb/La, arc-like signatures) and fluid-mobile elements and Th enrichment (high Ba/Nb, Sr/ Nb, Rb/Nb, U/Nb, K2O/Nb, and Th/Nb) (Figs. 14–21). These ratios are



Hf/3

(a)

17

20

(e)

All

WPB

N

-M E

O

R

-M

B

O

A

W

IC

Zr/Y

I AT

10

R

B

PT

MORB

IAB

W PA

1 10

Th

100

Nb/16 Hf/3

(b)

I AT

-M E

O

R

Deccan all data MgO>8.0 wt% Arc

-M

B

O

IC

A

W

R

Zr/Sm

(f)

MgO>8.0wt%

N

B

PT

W PA

Arc

Th

(c)

Nb/16

600 Arc tholeiite

(g)

CFBs

Tr e n

Ti/V

58

C

54

on

SiO2 (wt%)

400 300 200

MORB

OIBs

d

Sr/Nd

500

V (ppm)

1000

Zr (ppm)

ta

m

in

at

50

io

n

46


42

0

0

5

10

15

20

25

0. 0

0.5

1. 0

1.5 Nb/La

Ti/1000

(h)

(d) 50000

2. 0

8

La/Sm

4

S

Co

Ti (ppm)

6 10000

2.5

rc

e

er

o

ei

ty

nt

8%

ou

t he

n ge

am

2

in at io n

0

1000 10

100

Zr (ppm)

1000

0. 0

0.5

1. 0

1.5 Nb/La

2. 0

2.5

Fig. 10. Plots of (a)–(b) Nb/16 versus Hf/3 versus Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr\ \Zr/Y (Pearce and Norry, 1979), (f) Ti/V\ \Zr/Sm\ \Sr/ Nd, (f) Zr/Sm-Sr/Nd-Ti/V, (g) SiO2-Nb/La, and (h) La/Sm-Nb/La for the Deccan CFBs. The less-evolved Deccan CFBs mainly plot within the intraplate tectonic setting. The tight correlation of SiO2\ \Nb/La and La/Sm\ \Nb/La defined by the Deccan CFBs suggest crustal contamination played an important role in their generation.


18


(a)

(d)

Hf/3

50000

All

N

-M

O

E-

R

M

B

O

A

W

IC

Ti (ppm)

IA T

WTP

R

B

10000

30% VAB

PT

W PA

1000 10

100

Nb/16

Th

(b)

(e)

Hf/3 MgO>8.0 wt%

20

N

-M

O

R

-M

B

O

A

W

IC

Zr/Y

IA T

10

E

R

B

WPB

PT

MORB

IAB W PA

1 10

Th

(c)

(f)

600

MORB

1000

Zr/Sm

CFBs

500

V (ppm)

100

Zr (ppm)

Nb/16

Arc tholeiite

1000

Zr (ppm)

Siberian all data MgO>8.0 wt% Arc

400 300 200

OIBs 100 Calc-alkaline basalts

0 0

5

10

15

20

Arc

25

Ti/1000

Sr/Nd

Tr e n

d

Ti/V

Fig. 11. Plots of (a)–(b) Nb/16\ \Hf/3\ \Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr\ \Zr/Y (Pearce and Norry, 1979), and (f) Ti/V\ \Zr/Sm\ \Sr/Nd for the Siberian CFBs.

ultimately attributed to water flux melting. Because H2O behaves like a highly incompatible trace element, such as Ce and K2O, during mantle melting (Dixon et al., 2002; Hauri et al., 2006), H2O/Ce ratios of primary melts (without the effect of degassing in the crust, which would drive H2O loss and decouple any primary relationship) can therefore be used to constrain the water contribution in arc water flux melting. Based on the global primitive arc basalt data (with the exception of one rock unit, all other rock units have SiO2 in the range of 44%–53%) compiled by Ruscitto et al. (2012), ratios of fluid-mobile elements (Ba, Rb, Sr, U, and K) versus fluid-immobile elements (La and Nb) are highly correlated with H2O/Ce (Fig. 22). This implies that the enrichment of fluid-mobile elements (Ba, Sr, Rb, K2O, and U) relative to HFSEs (Nb) and REE (La) can be used to constrain the primary H2O/Ce ratios. There is some scatter evident at low H2O/Ce ratios, perhaps attributable to the effects of degassing. Degassing in the crust drives H2O loss, with little effect on trace elements, and would decrease the H2O/Ce ratio and decouple any primary relationships (Plank et al., 2013). The

similarity of fluid-mobile element enrichment and HFSE depletion (Figs. 4–5 and 14–21) between arc-like intra-continental basalts and arc basalts suggests the importance of deep-Earth water cycling in the formation of at least some of Earth's large igneous provinces. 3.3.2. Water effect on magma evolution paths: Independent constraint from experimentally determined LLDs for hydrous and anhydrous basaltic rocks Numerous processes have been identified that modify the composition of mantle-derived basaltic magmas at different depths during their ascent from the source region toward the surface. Although assimilation of country rocks, trapping of interstitial liquids in cumulates, mixing of different magma types, and replenishment of magma chambers with less differentiated magmas can occur, crystal fractionation is generally identified as the predominant process operating in crustal magma reservoirs. Petrological and geochemical (Cox, 1980; Lightfoot et al., 1990; Thompson et al., 1980) and experimental (Botcharnikov et al., 2008; Freise et al., 2009; Toplis and Carroll, 1995; Villiger et al., 2004,


Hf/3

(a)

(d)

50000

Ti (ppm)


10000

19

WTP

IA T

All

N

-M

O

E-

R

M

B

O

IC

A

W

R

B

15% VAB

PT 1000

W PA

10

100

1000

Zr (ppm)

Th

Nb/16

(b)

(e)

Hf/3 MgO>8.0 wt%

N

-M

O

E-

R

M

O

A

W

IC

10

B R

B

WPB

Zr/Y

IA T

20

PT

MORB

W PA

IAB

Th

Nb/16

1 10

100

1000

Zr (ppm)

(f)

(c) 600

Arc tholeiite

MORB

Zr/Sm

CFBs

Columbia River all data MgO>8.0 wt% Arc

500

V (ppm)

400 300 200 OIBs


0 0

5

10

15

20

Arc

25

Ti/1000

Sr/Nd

Tr e n

d

Ti/V

Fig. 12. Plots of (a)–(b) Nb/16\ \Hf/3\ \Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr\ \Zr/Y (Pearce and Norry, 1979), and (f) Ti/V\ \Zr/Sm\ \Sr/Nd for the Columbia River CFBs.

2007) studies on basalts clearly support the view that crystal fractionation is the predominant process involved in the evolution of basalts once they cross the crust-mantle boundary. Experimental studies of crystal fractionation demonstrate that water contents have a dramatic influence on LLDS in basalts from both oceanic LIPs, such as the Kerguelen LIP (Freise et al., 2009), and their continental counterparts, such as the Yellowstone LIP (Columbia River CFBs; Botcharnikov et al., 2008). Thus, comparing magma evolution paths for global CFBs and other continental basalts with experimentally determined LLDs by crystal fractionation of anhydrous and hydrous basaltic rocks can provide independent and important constraints on the role of water in formation of continental basalts. The magma evolution paths defined by CFBs and other typical continental basalts are not strictly LLDs, because of contamination and magma mixing, but the petrological, geochemical, and experimental studies outlined above show that major element variation is mainly controlled by crystal fractionation. This enables us to compare

the magma evolution paths of continental basalts with experimentally determined LLDs. Both hydrous crystal fractionation experimental data (Almeev et al., 2013; Berndt et al., 2005; Blatter et al., 2013; Botcharnikov et al., 2008; Freise et al., 2009; Mandler et al., 2014; Nandedkar et al., 2014; Pichavant and Macdonald, 2007; Rader and Larsen, 2013; Sisson et al., 2005) and anhydrous crystal fractionation experimental data (Thy et al., 2006; Toplis and Carroll, 1995; Villiger et al., 2004, 2007) were compiled to examine the effect of water on the LLDs of basaltic rocks. The starting materials for hydrous experiments range from highly evolved basaltic rocks (SiO2 = 54.63 wt%, MgO = 3.29 wt% in Rader and Larsen, 2013) to nearly primitive basalts (SiO2 = 48.4 wt%, MgO = 10.1 wt% in Nandedkar et al., 2014; SiO2 = 49.6 wt%, MgO = 9.8 wt% in Berndt et al., 2005; and SiO2 = 47.0 − 48.4 wt%, MgO = 10.6 − 12.5 wt% in Pichavant and Macdonald, 2007). The hydrous experiments were conducted with H2O = 1.0–9.0 wt% (mostly between


20


(a)

(d)

Hf/3

50000

WTP

N

-M

O

E-

R

M

B

O

A

W

IC

Ti (ppm)

IA T

All

R

B

10000

12% VAB

PT

1000

W PA

10

Th

100

Nb/16

(b)

20

(e)

Hf/3

1000

Zr (ppm)

10

N

-M

O

R

M E-

B

O

A

W

IC

Zr/Y

IA T

MgO>8.0 wt% R

B

WPB

PT

MORB

IAB W PA

1 10

Nb/16

Th

(c) 600

100

MORB

Arc tholeiite

(f)

CFBs

Zr/Sm Basin and Range all data MgO>8.0 wt% Arc

500

OIBs

V (ppm)

400

1000

Zr (ppm)

300 200 100 Calc-alkaline basalts

0 0

5

10

15

20

Arc

25

Ti/1000

Sr/Nd

Tr e n

d

Ti/V

Fig. 13. Plots of (a)–(b) Nb/16 versus Hf/3 versus Th (Wood, 1980), (c) V\ \Ti/1000 (Shervais, 1982), (d) Ti\ \Zr (Pearce, 1996), (e) Zr\ \Zr/Y (Pearce and Norry, 1979), and (f) Ti/V\ \Zr/ Sm\ \Sr/Nd for the Basin and Range basalts. The diamonds in (b) to (f) indicate less-evolved samples with MgO N 8 wt%.

3 wt% and 6 wt%), and experimental temperatures of 825–1200 °C (mostly concentrated between 950–1200 °C). The starting materials for anhydrous crystal fractionation experiments ranged from primitive basalts (SiO2 = 48.7–51.6 wt%, MgO = 10.8–14.8 wt% in Villiger et al., 2004, 2007) to highly evolved basalts (SiO2 = 48.0–54.3 wt%, MgO = 8.9–1.8 wt% in Thy et al., 2006; Toplis and Carroll, 1995; Villiger et al., 2004, 2007). The anhydrous experiments were conducted at atmospheric pressure (Thy et al., 2006; Toplis and Carroll, 1995) up to lower middle-crust level (1.0 GPa in Villiger et al., 2004 and 0.7 GPa in Villiger et al., 2007) and temperatures of 1050–1330 °C. The ca. 60 Ma old Baffin Bay picrites between Baffin Island and West Greenland are among the earliest manifestations of the ancestral Iceland mantle plume. The high 3He/4He end-member of the mantle composition ranges up to 50RA (where RA is the atmospheric value of 1.39 × 10− 6; Starkey et al., 2009) in the picrites and may signify an undegassed primitive mantle source or isolated primordial He-rich reservoir that is a residue of ancient mantle depletion (Heber et al., 2007).

The Baffin Bay lavas are therefore the best candidate to characterize a lower mantle reservoir, since they probably originated from the core– mantle boundary (Jackson et al., 2010; Wang et al., 2013). The Baffin Bay lavas were therefore used as a reference frame in this study for hot and deep-mantle plume components. The data used here were compiled by Wang et al. (2013). As shown in Fig. 23, the LLD of the Baffin Bay lavas is consistent with the trend defined by anhydrous crystal fractionation in experimentally determined liquids. The crystal fractionation experimentally determined liquid compositions and the Baffin Bay lava data are summarized in Fig. 23. The anhydrous experimental liquids generally display relative higher FeOt, TiO2, CaO/Al2O3, but lower Al2O3, Na2O + K2O, and SiO2 at given lower MgO contents. For example, the crystal fractionation of anhydrous basalts results in significant enrichment in total iron starting from about MgO = 8 to 4 wt% in residual magmas, whereas the crystal fractionation of hydrous basalts slightly increases total iron between MgO = 8 to 6 wt% but this quickly decreases from MgO ≤ 6 wt% in



0.8

300 Subduction ad d ition (Hydrous fluid)

Andes Arc Tonga Luzon Kermadec Izu-Bonin

200

Subduction ad d ition (Hydrous fluid)

0.6

U/Nb

250

Ba/Nb

21

150 100

0.4

0.2 Anhydrous enrichment

50

(a)

0.0 Anhydrous enrichment

0 0.0

0.5

1.0

1.5

2.0

0.0

2.5

1.0

1.5

2.0

2.5

1.6

1600 Subduction ad d ition

Subduction ad d ition

(Hydrous fluid)

1200

(sediment melting)

1.2

Th/Nb

Sr/Nb

0.5

(d)

800

400

0.8

0.4

Anhydrous enrichment


(b) 0

(e)

0.0 0.0

0.5

1.0

1.5

2.0

0.0

2.5

0.5

1.0

1.5

2.0

2.5

10

40

Subduction ad d ition Subduction ad d ition

(K/Nb)/(100,ppm)

(Hydrous fluid)

Rb/Nb

30

20

10

(Hydrous fluid)

8 6 4 2


(c) 0

0 Anhydrous enrichment

(f)

-2 0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 14. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for less-evolved arc basalts (MgO N 8.0 wt%) from the Andes, Tonga, Luzon, Kermadec, and Izu-Bonin arcs. The estimates are calculated from the GEOROCK database (http://georoc.mpch-mainz.gwdg.de/georoc/). The trends suggest the source of arc basalts contains both hydrous and anhydrous enrich end-member components. The high Th/Nb end-member may reflect the contribution of recycled sedimentary components or a metasomatized mantle reservoir formed by slab-derived melt-peridotite interaction.

residual liquids (Fig. 23). The trend defined by crystal fractionation of hydrous basalts is characterized by slight enrichment of Al2O3 at MgO b6 wt% and slight depletion of Al2O3 at MgO ≥ 6 wt% (Fig. 23). In contrast, the trend defined by the crystal fractionation of anhydrous basalts shows Al2O3 depletion at MgO b 8 wt% and they are generally lower than 13% Al2O3 at MgO b6 wt%. Liquids produced by crystal fractionation of anhydrous basalts are characterized by high TiO2 contents (mostly N 2 wt%) and high CaO/Al2O3 ratios (mostly N 0.6) at MgO b 6 wt%, whereas liquids resulting from the crystal fractionation of hydrous basalts display low TiO2 contents (mostly ≤2 wt%) and low CaO/Al2O3 ratios (mostly b0.6) at MgO = 2–6 wt%. Thus, LLDs of hydrous basalts are systematic different from those of anhydrous basalts. Because the starting materials and pressure–temperature conditions are comparable between hydrous and anhydrous experiments, these differences are mainly attributed to water content. Fig. 24 compares Karoo CFBs with the experimentally determined LLDs. The comparisons for the other CFBs (Central Atlantic, Emeishan, Deccan, Siberia, and Columbia River) and rift basalts (Basin and

Range) are presented in Appendix Figs. 7–12. CFBs all show iron depletion trends starting at about MgO = 6–5 wt%, a typical characteristic of hydrous basalts. CFBs generally display relatively high Al2O3 contents and low CaO/Al2O3, plotting mainly within the hydrous LLD field. The Karoo, Central Atlantic, Deccan, and Columbia River CFBs have low TiO2 at MgO b 5 wt%, mostly plotting within the field of hydrous LLDs (Fig. 24 and Appendix Figs. 7, 9, and 11). The highly evolved Siberia (Appendix Fig. 10) and Emeishan (Appendix Fig. 8) CFBs (MgO b 6 wt%) display large ranges of TiO2 contents, plotting in both the hydrous and anhydrous fields. CFBs also evolve toward high Na2O + K2O and SiO2 contents, plotting predominately within the field defined by hydrous LLDs (Fig. 24 and Appendix Figs. 7–11). Thus, the evolution paths of CFBs are generally similar to those defined by hydrous LLDs, suggesting that H2O played an important role in their evolution. 3.3.3. Estimated magma H2O concentrations of typical continental basalts As magmas crystallize at depth, their LLDs are determined by their phase equilibria. Because the stabilities of many minerals, such as


22


120

0.25

Karoo 100

Hydrous fluid

0.20

Hydrous fluid 90

U/Nb

Ba/Nb

80 60 40

0.15

0.10

70

90 50


70

20

0.05

50


(d)

(a) 0

0.00 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

1.0

160

Sediment addtion Hydrous fluid

0.8

Th/Nb

Sr/Nb

120

80

90

0.6

0.4


40

50

0.2

(b)

50

70

90


(e) 0.0

0 0.0

0.5

1.0

1.5

2.0

0.0

2.5

0.5

1.5

2.0

2.5

3.0

10 Hydrous fluid

Hydrous fluid

2.5

K/Nb(/100,ppm)

8

Rb/Nb

1.0

6 90


70 50

2

2.0 1.5

90 70

1.0


50

0.5

(f)

(c) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

Nb/La N-MORB

OIB

E-MORB

EM1

1.0

1.5

2.0

2.5

Nb/La Upper continental crust EM2

HIMU

Arc basalts with MgO > 8 wt% Karoo CFBs with MgO>8 wt%

Two-end-member mixing between average OIB and upper continental crust

Fig. 15. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Karoo CFB with MgO N 8.0 wt%. CFBs data are from the GEOROCK database (http://georoc.mpch-mainz.gwdg.de/georoc/). The average values for normal and enriched mid-ocean ridge basalts (N-MORB and E-MORB), alkali ocean-island basalt (OIB) are from Sun and McDonough (1989). The fields for enriched (EM-1 and EM-2) and high 238U/204Pb (HIMU) mantle end-members are from Weaver (1991). The data for average upper continental crust are from Rudnick and Gao (2003). The two-end-member mixing between OIB and upper continental crust is also shown. Each cross on the binary mixing lines indicates 0.1 increments of upper continental crust.

plagioclase, are highly sensitive to H2O concentration, LLD may record the pre-eruptive volatile concentration of the magma (Parman et al., 2011 and references therein). The Al2O3–LLD method was therefore applied to constrain the pre-eruptive H2O contents using the experimentally determined function of H2O (wt %) = 1.34 × Al2Omax (wt%) − 21.05, 3 where Al2Omax is the maximum concentration in a magma before 3 plagioclase saturation (Parman et al., 2011). We use three criteria to identify LLDs: (1) Eu anomalies (Eu* = 2 × EuN/(SmN + GdN), where subscript N indicates chondrite normalized value); (2) Sr anomalies relative to the REE (Sr* = 2 × SrN/(SmN + NdN), where subscript N indicates primitive mantle-normalized values); (3) Nb/La ratios. Firstly, the diagram of Al2O3 versus FeOt/MgO can be

used to examine the crystal fractionation or accumulation of olivine, clinopyroxene, and plagioclase based on the Al2O3 contents of liquids. Crystal fractionation or accumulation of plagioclase does not change the FeOt/MgO ratio but could result in a significant decrease or increase in the Al2O3 content. A nearly vertical trend on the Al2O3\\FeOt/MgO diagram therefore is predicted for plagioclase fractionation or accumulation. Contrasting with plagioclase fractionation, fractionation of olivine and clinopyroxene would increase both the Al2O3 content and FeOt/ MgO ratio. Thus, by stripping-off plagioclase accumulation using the Eu* and Sr* anomaly values, fractionation of olivine and clinopyroxene can be identified by positive correlation between Al2O3 and FeOt/MgO. The advantage of this diagram is that it is sensitive to the site of



23

0.25

160

CAMP Hydrous fluid

Hydrous fluid

0.20

120

U/Nb

Ba/Nb

90

80

90

40

0.10

70 50



0.05

70 50

0.15

(a)

(d)

0

0.00 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

1.0

200 Hydrous fluid

Sediment addtion

0.8

Th/Nb

Sr/Nb

150

100

50

0.5


0.2

(e) 0.0

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

10 Hydrous fluid

Hydrous fluid

K/Nb(/100,ppm)

8

Rb/Nb

70

(b)

50

70

90

0.0

0.4

50


0

90

0.6

6 90

4 70 Anhydrous enrichment

2.0 90 70

1.0

50


50

2

(f)

(c) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 16. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Central Atlantic CFBs with MgO N 8.0 wt%. The data sources and symbols are the same as in Fig. 15.

plagioclase-in on LLDs. We can then exclude samples that underwent plagioclase accumulation using Eu* N 1.02 and then use Sr* to check whether or not the Eu* criterion works (see details in Parman et al., 2011). Due to the lack of Sr\\Nd isotope data for most samples, the alterative criterion, the Nb/La ratio, was used to examine the contribution of magma or source mixing to produce the maximum Al2O3. If magma/source mixing is a dominant factor to produce the maximum Al2O3, correlations between Al2O3 and Nb/La are expected. Both filtered and unfiltered data are shown in Fig. 25 and Appendix Figs. 13–18. CFBs and the Basin and Range basalts show very similar correlations between Al2O3 and FeOt/MgO (Fig. 25 and Appendix Figs. 13–18). Their Al2O3 contents quickly increase with increasing FeOt/MgO (first stage) and then become nearly constant (second stage), and finally significantly decrease (last stage). The first transition point suggests the occurrence of plagioclase in the fractionating mineral assemblage, whereas the late-stage negative correlation between Al2O3 and FeOt/MgO indicates that plagioclase should be saturated (becoming a dominant

phase in the fractionated mineral assemblage). Thus, the point that marks the start of Al2O3 depletion was chosen to constrain the maximum concentration in a magma before plagioclase saturation. With the exception of the Emeishan CFBs, the difference of Al2Omax 3 values between filtered and unfiltered data varies by only 0–1 wt%. This suggests minor plagioclase accumulation in the LLDs of these CFBs. The unfiltered data from the Emeishan LIP define a much higher Al2Omax value than that obtained from the filtered data, suggesting the 3 importance of plagioclase accumulation in some Emeishan basalts. Thus, only filtered data were used to constrain Al2Omax and subsequent 3 H2O estimate. The H2O concentrations obtained by the Al2O3–LLD method were as follows: Central Atlantic = 4.41 wt% (filtered data) to 5.75 wt% (unfiltered data); Columbia River CFBs = 3.74 wt% (filtered data) to 4.41 wt% (unfiltered data); Deccan and Karoo = 5.75 wt% (both filtered and unfiltered data); Emeishan = 4.41 wt% (filtered data); Basin and Range and Siberia = 4.41 wt% (both filtered and unfiltered data).


24


0.4

100

Emeishan Hydrous fluid

Hydrous fluid

0.3

60

U/Nb

Ba/Nb

80

40

0.2 90

90

90 70

20

70

0.1


50

50


(a)

(d) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

1.2

120 Hydrous fluid

Sediment addtion

100 0.8

Th/Nb

Sr/Nb

80 60 40

90

0.4


70

Anhydrous enrichment 50

(b)

50

70

90

20

(e) 0.0

0 0.0

0.5

1.0

1.5

2.0

0.0

2.5

10

0.5

1.0

1.5

2.0

2.5

3.0 Hydrous fluid

K/Nb(/100,ppm)

Rb/Nb

Hydrous fluid

2.5

8

6 90


70

2

50

2.0 90

1.5

70

1.0

50


0.5

(c) 0

(f) 0.0

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 17. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Emeishan CFBs with MgO N 8.0 wt%. The data sources and symbols are the same as in Fig. 15.

Considering the error in the Al2O3–LLD method (1–2 wt%; Parman et al., 2011), the estimated pre-eruptive magma H2O concentrations for typical continental basalts are very similar at 4.0–5.0 wt%. Arc-like continental basalts and arc basalts share the similar characteristic of coupling between fluid-mobile element enrichment and HFSE depletion (Figs. 4–5 and 14–21). This suggests that water flux melting played a key role in formation of these two basalt types. Thus, the tight correlations between Ba/La, Ba/Nb, Rb/Nb, and H2O/Ce defined by global primitive arc basalts can be used to estimate the H2O/Ce ratios of the primary continental basaltic melts. In order to minimize the effect of source heterogeneity and degree of partial melting, primary melt H2O/ Ce ratio (wt%/ppm) was calculated by the relationship of (1) H2O/Ce = 0.1092 − 0.0008 × (Ba/La) + 0.0001 × (Ba/La)2, R2 = 0.80 (Fig. 22a); (2) H2O/Ce = 0.1210 + 0.001 × (Ba/Nb) + 0.000000779 × (Ba/Nb)2, R2 = 0.77 (Fig. 22b); (3) H2O/Ce = 0.1221 + 0.0073 × (Rb/ Nb) + 0.0011 × (Rb/Nb)2, R2 = 0.76 (Fig. 22c). The primary melt H2O concentration was then determined by primary Ce concentration (CeFo91, equilibrium with olivine Fo91) and the H2O/Ce ratios were

obtained. The final primary melt H2O concentrations are the average of those determined by the above relationships. The estimated primary melt H2O concentrations for typical continental basalts are presented in Fig. 26. The estimates for the Karoo, Central Atlantic, and Siberian CFBs are nearly constant (mostly ranging from 1.0 to 3.0 wt%), whereas the estimates for the Deccan, Columbia River, and Emeishan CFBs and the Basin and Range basalts show a wide range, up to 10 wt%. The estimates for the first three LIPs give weighted mean primary melt H2O concentrations of 1.31 ± 0.05 wt% (Siberian), 1.41 ± 0.12 wt% (CAMP), and 1.92 ± 0.12 wt% (Karoo). The large range of estimated H2O for the Deccan CFBs may be attributed to crustal contamination because of the tight correlation between SiO2 and Nb/La (Fig. 10g). Input of crustal materials would likely increase the ratios of Ba/Nb, Ba/La, and Rb/Nb and the Ce content, resulting in an overestimation of the H2O concentration. The estimates for the Emeishan CFBs are generally higher than the others, mostly varying from 4.0 to 5.6 wt%. With the exception of the Emeishan CFBs, the estimated primary H2O contents are otherwise similar, with weighted means ranging from 1.0



60

0.25

Deccan

Hydrous fluid

Hydrous fluid

50

0.20

40

90

90

U/Nb

Ba/Nb

25

30 70

20

0.15 0.10

70

Anhydrous enrichment Anhydrous enrichment

50

50

0.05

10

(a)

(d)

0

0.00 0.0

0.5

1.0

1.5

2.0

2.5

0.0

1.0

2.0

2.5

0.8

Th/Nb

80 60 40

90

0.6 0.4



50

0.2

(b)

50

70

90

20

(e) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

2.5

8 Hydrous fluid

K/Nb(/100, ppm)

Hydrous fluid

6

Rb/Nb

1.5

Sediment addtion

Hydrous fluid

100

Sr/Nb

0.5

1.0

120

90


2

50

2.0 1.5

90

70

1.0

50


0.5

(c)

(f) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 18. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Deccan CFBs with MgO N 8.0 wt%. The data sources are the as in Fig. 15.

to 3.0 wt% (Fig. 26). The estimates for the Basin and Range basalts are high, up to 10.0 wt%, mostly ranging from 3.8 to 6.6 wt%. This is consistent with similarities between the geochemical characteristics of the Basin and Range and arc basalts, as evidenced by similar primitive mantle-normalized trace element distribution patterns (Fig. 5g) and estimated primary major element compositions (Fig. 3g). This is also consistent with the evolution trend of the Basin and Range basalts, which plot mainly in the field defined by hydrous LLDs (Appendix Fig. 12). A fundamental tenet of the classical thermal plume model is its prediction of pre-eruption uplifting, with 100 °C of plume excess temperature predicted to produce about 0.8–1 km pre-volcanic uplifting (Campbell and Griffiths, 1990). However, no evidence of uplift was found in the Columbia River CFBs, Siberian Traps, CAMP, or Ontong Java Plateau (Wang et al., 2015 and references therein). The Permian Emeishan LIP in Southwest China is often cited as a typical example of crustal domal uplift caused by thermal mantle plume upwelling prior to the onset of volcanism. However, this model has been questioned by the discovery of hydromagmatic volcaniclastic deposits formed in a marine environment, located near the central Emeishan LIP area, which are inferred to be the zone of maximum uplift (Ukstins Peate and Bryan, 2008). Recent detailed volcanology and sedimentology

studies demonstrated no pre-eruption crustal uplifting in the Emeishan LIP (Zhu et al., 2014). This strongly argues against the classical thermal plume model. The high primary melt H2O contents calculated in this study confirm that the CFBs were likely the result of wet upwelling of mantle without any excess temperature (Wang et al., 2015), thus predicting no pre-volcanic eruption surface uplift. Therefore, the role of thermal plumes in generation of CFBs may be overestimated or else the behavior of lower mantle plumes is not as proposed by the classical model (Campbell and Griffiths, 1990). Both magma and primary melt water concentrations in the Columbia River CFBs are well constrained by density numerical modeling, crystallization experiments, and olivine-hosted melt inclusion and were used to validate our calculation. Numerical modeling based on the effect of dissolved H2O or exsolved CO2 on the density of various Columbia River CFBs as a function of depth indicates the minimum total volatile (H2O + CO2) concentration ranges to N4 wt% and dissolved H2O up to 4.0 wt% (Lange, 2002). Crystallization experiments using a hydrous ferrobasalt as starting material indicated that differentiation of some Columbia River CFBs occurred in magmatic systems containing 0.5– 3.0 wt% H2O (Botcharnikov et al., 2008). Recently, studies of olivinehosted melt inclusions from the Columbia River flood basalt province


26


0.25

160

Siberia

Hydrous fluid

Hydrous fluid

0.20

120

U/Nb

Ba/Nb

90

80

0.15 0.10

70

90



70

40

50

0

0.00 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.5

1.0

2.0

2.5

1.0

200

Sediment addtion

Hydrous fluid

0.8

Th/Nb

150

Sr/Nb

(d)

0.05

(a)

100

90

0.6 0.4



50

50

50

90

70

0 0.0

0.2

(b)

0.5

(e)

0.0

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

10

Hydrous fluid Hydrous fluid

K/Nb(/100, ppm)

8

Rb/Nb

6 90

4 70

2


2.0 90 70

1.0


(c)

(f) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 19. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Siberian CFBs with MgO N 8.0 wt%. The data sources are the as the Fig. 15.

in eastern Oregon and Washington show a wide range in H2O concentrations, varying from 0.5 to 4.24 wt% (Cabato et al., 2015). Thus, the highest water concentrations constrained by crystallization experiments (3.0 wt%) and olivine-hosted melt inclusions (4.24 wt%) are consistent with that determined by the Al2O3–LLD method in this study: 3.74 wt% (Al2Omax = 18.5 wt%, filtered data) or 4.41 wt% (Al2Omax = 18.5 wt%, un3 3 filtered data). The primary water concentrations obtained by correlations between Ba/La, Ba/Nb and Rb/Nb and H2O/Ce in this study plot mainly within the H2O range constrained by crystal fractionation experiments (grey band in Fig. 26a) and olivine-hosted melt inclusion studies (yellow bins in Fig. 26a). Only a few estimates have higher H2O concentrations than the highest H2O concentration recorded by the melt inclusions. Whether this is due to higher H2O melt inclusions not being identified or the H2O/Ce ratio in the primary melt was overestimated in some samples is an area for future study. Regardless of the few exceptions, our estimates match well with the independent constraints from crystal fractionation experiments, melt inclusions, and thermodynamical modeling. Thus, our estimates appear reasonable. This implies that water flux melting most likely played an important role in the generation of some large-scale intra-continental igneous provinces.

4. Case study of 313–273 Ma arc-like continental basalts from the Xing'an–Mongolia orogenic belt, North China: New constraints on post-orogenic extension and water flux melting The Central Asian Orogenic Belt (CAOB) extends eastward from the Urals in the west, through Mongolia, to Far East Russia and southward from Siberia to northern China (Jahn, 2004; Jahn et al., 2000; Windley et al., 2007). Late Carboniferous to early Permian mafic volcanic rocks are widely distributed in the eastern segment of the CAOB (Fig. 1d) and are crucial for understating the evolution of CAOB. However, the petrogenesis of these mafic rocks is highly debated (Liu et al., 2013; Xiao et al., 2003, 2009; Xu et al., 2013, 2014; Zhang et al., 2008). For example, based on SHRIMP zircon data from ultramafic rocks in Inner Mongolia, Xiao et al. (2003, 2009) proposed an island arc setting for the volcanic rocks. Liu et al. (2013) proposed that the Late Carboniferous to early Permian volcanic rocks may have been produced within a subduction environment based on evidence of arc-like geochemical signatures in the highly deformed Amushan basalts in the Daqing area. However, the geochemical characteristics of Permian bimodal volcanic rocks from central Inner Mongolia led Zhang et al. (2008) to propose a post-orogenic extensional environment for these volcanic rocks. This



160

27

1.2

Columbia River

Hydrous fluid

Hydrous fluid

120

U/Nb

Ba/Nb

0.8 80

0.4


70

90


(a)

50

0

(d)

50

90

70

40

0.0 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

2.5

250

Sediment addtion Hydrous fluid

2.0

150

Th/Nb

Sr/Nb

200

100

1.5 1.0 Anhydrous enrichment


90

50

0.5

(e)

50

70

(b)

90 70 50

0

0.0 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

2.5

10

Hydrous fluid

Rb/Nb

K/Nb(/100, ppm)

Hydrous fluid

8 6

90


2

50

2.0 1.5

90

1.0


50

0.5

(c)

(f) 0.0

0 0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 20. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Columbia River basalts with MgO N 8.0 wt%. The data sources are the as the Fig. 15.

is consistent with evolution of the eastern segment of the CAOB based on sedimentology, field geology, and geochronology studies by Xu and his co-workers (Xu et al., 2013, 2014, 2015; Zhao et al., in press). Volcanic rocks from three typical units in the Xing'an–Mongolia segment of the CAOB (the Benbatu, Amushan, and Dashizhai formations; Fig. 1d) were selected to examine whether they represent arc-like intracontinental basalts or were developed in the subduction environment. The Late Carboniferous Benbatu Formation is widely distributed and composed of shallow-marine sandstone, mudstone, limestone, and volcanic rocks. The Benbatu volcanic rocks (also named as the Chagannuoer volcanic rocks in some geological reports) mainly consist of basalt, basaltic andesite, and andesite, along with a small amount of dacite, showing a bimodal feature (e.g. Tang et al., 2011) and formed between 313 and 300 Ma (Pan et al., 2012; Tang et al., 2011). The basalts (named the Benbatu basalts hereafter) used in this study show both pillow and massive structures in outcrop in the Sunid Youqi area (Fig. 1d). The Late Carboniferous Amushan Formation mainly consists of marine limestone, bioclastic limestone, and basalts, together with subordinate sandstone and siltstone (Li, 1996; Liu et al., 2013). Recent studies show that the Amushan basalts in the Daqing area of the Xi Ujimqin

County (Fig. 1d) were formed at 314–318 Ma and can be divided into foliated and pillow types (Liu et al., 2013). The foliated basalts show high LOI (6.3–9.7 wt%) and are not considered in this study, whereas the pillow basalts (LOI = 2.0–4.1 wt% (Liu et al., 2013) were investigated to determine their petrogenesis. The early Permian volcanic rocks are generally assigned to the Dashizhai Formation and its equivalent in Inner Mongolia (Fig. 1d; Zhang et al., 2008, 2011) and are widely distributed in central and eastern Inner Mongolia. They are mainly composed of basalt, basaltic andesite, and andesite, with interbedded muddy siltstone in the lower part and more felsic volcanic rocks (andesite, dacite, and rhyolite) in the upper part (Bao et al., 2005). The Dashizhai Formation is typical of the lower Permian marine volcano–sedimentary sequence in Inner Mongolia (Bao et al., 2005; Li, 1996; Zhang et al., 2008). SHRIMP and LA-ICPMS zircon U\\Pb dating demonstrate that the Dashizhai volcanic rocks and their equivalents were erupted at ca. 289–273 Ma (Chen et al., 2012; Zhang et al., 2008, 2011). Based on our argument that V\\Ti/1000, Zr/Y\\Zr, Zr\\Ti, and Sr/ Nd\\Zr/Sm\\Ti/V diagrams can be used to distinguish arc-like intracontinental basalts from arc basalts, we then applied these to examine the tectonic setting of the Xing'an–Mongolia volcanic rocks. The rocks


28


0.4

300

Hydrous fluid

Basin and Range

Hydrous fluid

250 0.3

U/Nb

Ba/Nb

200 150

0.2 90 Anhydrous enrichment


100

70 50

(a)

50

90

50

70

0.1

(d)

0

0.0 0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

1.2

250 Hydrous fluid

Sediment addtion

200

100

Th/Nb

Sr/Nb

0.8 150


90

0.4

70 Anhydrous enrichment 50

0 0.0

50

(b)

70

90

50

0.5

(e) 0.0

1.0

1.5

2.0

2.5

0.0

0.5

1.0

1.5

2.0

2.5

4

12 Hydrous fluid

Hydrous fluid

K/Nb(/100, ppm)

10

Rb/Nb

8 6 90 Anhydrous enrichment

4 70

2

3


70

1

50

50

(c)

(f) 0

0 0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

0.0

0.5

1.0

1.5

2.0

2.5

Nb/La

Fig. 21. Plots of (a) Ba/Nb, (b) Sr/Nb, (c) Rb/Nb, (d) U/Nb, (e) Th/Nb, and (f) (K/Nb)/100 (ppm) versus Nb/La for the Basin and Range basalts with MgO N 8.0 wt%. The data sources are the as the Fig. 15.

plot mainly within the intraplate basalt field and show geochemical affinity with typical rift-related basalts as defined by the Basin and Range basalts (Fig. 27). Thus, these volcanic rocks were most likely produced in a post-orogenic environment. Because most of the studied samples have low MgO (b 8.0 wt%), crustal contamination has the potential to affect the whole-rock compositions. As we have shown, the crustal contamination results in a shift from plotting within the intraplate field to plotting in the arc field (Fig. 10). Therefore, considering the effect of crustal contamination, these basalts were most likely produced within an intraplate tectonic setting, rather than in a subduction environment. The intra-continental geochemical signature is also consistent with geological evidence. Firstly, two orogenic belts, namely the Late Devonian Northern Orogenic Belt (NOB) and Late Silurian–Early Devonian Southern Orogenic Belt (SOB) were identified in central Inner Mongolia (Xu et al., 2013; Zhao et al., in press and references therein). High-pressure blueschists of ca. 380 Ma in the Upper Devonian mélange belt of the NOB and the development of a molasse basin in both the NOB and SOB indicate the closure of the Paleo-Asian Ocean and the formation of the eastern part of CAOB (Xu et al., 2013). Secondly, Carboniferous to Permian

volcano–sedimentary rocks were widely deposited in central Inner Mongolia (Fig. 1d), unconformably overlying the basement and Early Paleozoic ophiolitic mélange (IMBGMR, 1991; Zhao et al., in press). Basal deposits of these strata packages are diachronous but commonly characterized by terrestrial conglomerates, sandstones, and siltstones, with plant fossils, representing fluvial to lacustrine facies (Zhao et al., in press). These are succeeded by widespread shallow-marine limestones of the Benbatu Formation (IMBGMR, 1991), representing the first-stage carbonate platform deposits formed in an inland sea (Zhao et al., in press). The following Benbatu basalts were thus erupted within an intra-continental shallow-marine basin, whereas the occurrence of fossil plant-bearing conglomerates, sandstones, siltstones, and mudstones in the lower part of the Amushan Formation indicates terrestrial deposition again in most regions. Limestones and overlying conglomerates, sandstone, and siltstones in the middle and upper part of the Amushan Formation represent the second-stage carbonate platform and postcarbonate terrestrial deposits (Zhao et al., in press). There are abundant shallow-marine fossils (e.g., brachiopods, corals, and pelecypods) in the limestones (IMBGMR, 1991). These lines of evidence imply a predominant shallow-marine depositional environment during the



2.0

1.0

0.5

Primitive arc basalt

R 2 = 0.60

H 2O/Ce (wt%/ppm)

H 2O/Ce (wt%/ppm)

H 2 O/Ce = 0.1092 -0.0008(Ba/La) 2 +0.0001(Ba/La)

1.5

2.0


2

R = 0.80

29

1.5

H 2 O/Ce = 0.0717+0.0008(Sr/Nb)

1.0

0.5

(a)

(d)

0.0

0.0 20

0

40

60

80

100

120

140

0

200

400

Ba/La 2.0

800

1200

H 2O/Ce (wt%/ppm)

H 2 O/Ce = 0.1210+0.001(Ba/Nb) +7.79E-07(Ba/Nb)2

1.5


R 2 = 0.65

1.0

0.5

H 2 O/Ce = 0.1557+0.1227(K 2 O/Nb) +0.3362(K 2 O/Nb) 2

1.5

1.0

0.5

(e)

(b) 0.0

0.0 0

200

400

600

800

0.0

1000

0.4

H 2 O/Ce = 0.1221+0.0073(Rb/Nb) 2 +0.0011(Rb/Nb)

1.5

2.0


2

R = 0.76

1.2

1.6

1.0

0.5


2

R = 0.52

H 2O/Ce (wt%/ppm)

2.0

0.8

K 2O/Nb (wt%/ppm)

Ba/La

H 2O/Ce (wt%/ppm)

1000

2.0


2

R = 0.77

H 2O/Ce (wt%/ppm)

600

Sr/Nb

1.5

H 2 O/Ce =0.1491-0.0187(U/Nb) 2 +2.5863(U/Nb)

1.0

0.5

(f)

(c) 0.0

0.0 0

5

10

15

20

25

30

Rb/Nb

0.0

0.2

0.4

0.6

U/Nb

Fig. 22. Plots of H2O/Ce versus (a) Ba/La, (b) Ba/Nb, (c) Rb/Nb, (d) Sr/Nb, (e) K2O/Nb (wt%/ppm), and (f) U/Nb for global primitive or near-primary arc basalts (with the exception of one rock unit, all other rock units have SiO2 ranging from 44% to 53%; data from Ruscitto et al., 2012). These tight correlations of fluid-mobile elements/Nb ratios with H2O/Ce imply that fluid-mobile element enrichment was the results of water flux melting.

Late Carboniferous in central Inner Mongolia (IMBGMR, 1991; Xu et al., 2014). Finally, detrital zircon ages demonstrate that the Late Carboniferous strata in central Inner Mongolia have a similar provenance and were mainly sourced from local, exposed Early Paleozoic arc-related magmatic rocks with subordinate input of Precambrian basement (Zhao et al., in press). This implies that the Late Carboniferous strata were formed on top of the Precambrian basement and above both the NOB and SOB (Zhao et al., in press). This indicates that there was an inland sea rather than a wide ocean during the Late carboniferous in central Inner Mongolia (Xu et al., 2014). The Dashizhai Formation is widely distributed in central Inner Mongolia, which is also consistent with an intra-continental extensional setting rather than a linear arc belt. However, each of the three rock units in central Inner Mongolia has distinctive trace element distribution patterns (Fig. 28). With the exception of Rb, U, and Sr, the Amushan basalts show smooth trace element distribution patterns with depletion of highly incompatible elements and unfractionated HREEs. This pattern is very similar to

N-MORB but significantly different from arc basalts (Fig. 28a). The Benbatu basalts display prominent negative Nb and Ti anomalies and strong enrichments of U, Zr, and Hf (Fig. 28b). The basalts also show variations in Sr and HREEs. The most pronounced feature of the Benbatu basalts is the significant enrichment of Zr and Hf relative to Sm and Nd. Such a feature is rarely observed in basalts formed in arc environments (Fig. 4), but intra-continental basalts, such as some of the Karoo CFBs, also display prominent positive Zr\\Hf anomalies (Fig. 5a). The Dashizhai basalts show a diversity of chemical compositions (Fig. 28c). Their trace element distribution patterns can be divided into two subtypes. One type displays relative smooth trace element distribution patterns similar to alkali OIB. The other type has relatively flat trace element distribution patterns and strong enrichment in U and depletion of Nb (arc-like). The remainder exhibits high contents of incompatible trace element contents with distribution patterns similar to OIB-type, but with significant depletion of Nb. The lack of depletion of Zr and Hf relative to adjacent REE and the coexistence of arc-like


30


20

12 10

Na2O+K2O (wt%)

FeOt (wt%)

15 FeOt = 13wt%

10

5

6 4 2

(a) 0

(d)

0 0

2

4

6

8

10

0

12

2

4

6

8

10

60

SiO2 (wt%)

20

16

12

55 50 45

(b)

(e) 40

8 0

2

4

6

8

10

12

0

2

4

6

8

10

12

1.2

CaO/Al2O3 (wt%)

6

TiO2 (wt%)

12

65

24

Al2O3 (wt%)

8

4

2

0.8

0.4

(c)

(f) 0.0

0 0

2

4

6

8

10

12

0

2

4

MgO (wt%) crystal crystallisation experiments glass Dry: Starting material Hydrous: glass Starting material

6

8

10

12

MgO (wt%) Nature unhydrous basalts ca. 60 Ma Baffin picrites

Fig. 23. Hydrous (Almeev et al., 2013; Berndt et al., 2005; Blatter et al., 2013; Botcharnikov et al., 2008; Freise et al., 2009; Mandler et al., 2014; Nandedkar et al., 2014; Pichavant and Macdonald, 2007; Rader and Larsen, 2013; Sisson et al., 2005) and anhydrous (Thy et al., 2006; Toplis and Carroll, 1995; Villiger et al., 2004, 2007) experimental data compiled to examine the effect of water on the liquid lines of descent (LLDs) of basaltic rocks. The ca. 60 Ma old Baffin Bay picrites with primitive He, Os, Nd, and Pb isotope compositions (Heber et al., 2007; Jackson et al., 2010; Starkey et al., 2009; Wang et al., 2013) are also shown to reflect the possible contribution from a lower mantle reservoir within a mantle plume (Jackson et al., 2010; Wang et al., 2013). The data for the Baffin Bay picrites were compiled by Wang et al. (2013). The starting materials used in crystal fractionation experiments are also shown. These figures show that magma water concentrations significantly affect the LLDs at MgO b 8.0 wt%.

and OIB-types suggest a non-arc environment for the Dashizhai basalts. This is consistent with the bimodal feature of the Dashizhai volcanic rocks. The Al2O3–LLD method was used to estimate the pre-eruptive H2O content of these basalts. As shown in Fig. 29, the maximum Al2O3 before plagioclase saturation is about 19 wt% for the Benbatu and Dashizhai basalts, and 16 wt% for the Amushan basalts. The corresponding H2O contents are 4.41 wt% (Benbatu and Dashizhai) and 0.39 wt% (Amushan), respectively. A quantitative index of Fe enrichment called the Tholeiitic Index (THI = Fe4/Fe8, where Fe4 is the average FeOtotal concentration of samples with 4 ± 1 wt% MgO, and Fe8 is the average

FeOtotal at 8 ± 1 wt% MgO) can also be used to constrain the average H2O content of magmatic systems (Zimmer et al., 2010). Forty-one magmatic suites from different tectonic settings define a function of H2O (wt%) = exp((1.26–THI)/0.32), R2 = 0.85. Because the Benbatu and Amushan basalts define a broad positive correlation between FeOt and MgO at MgO ≤ 8.0 wt%, the THI was calculated by FeOt according to the upper and lower trends in Fig. 29. The result shows that the average H2O contents for the Benbatu and Amushan basalts were about 5.4– 6.8 wt%. Because of the lack of samples with MgO ranging from 3.0 to 5.0 wt%, the Tholeiitic Index method cannot be applied to the Amushan volcanic rocks. The Al2O3–LLD and the Tholeiitic Index methods have



31

12

20

Karoo 10

Na2O+K2O (wt%)

FeOt (wt%)

15

10

5

6 4 2

(a)

(d)

0

0 0

2

4

6

8

10

0

12

24

2

4

6

8

10

60

SiO2 (wt%)

16

12

55 50 45

(b)

(e)

8

40 0

2

4

6

8

10

12

0

6

2

4

6

8

10

12

1.2

CaO/Al2O3 (wt%)

TiO2 (wt%)

12

65

20

Al2O3 (wt%)

8

4

2

0.8

0.4

(c)

(f) 0.0

0 0

2

4

6

8

10

12

MgO (wt%)

0

2

4

6

8

10

12

MgO (wt%)

Fig. 24. Comparing (a) FeOt, (b) Al2O3, (c) TiO2, (d) Na2O + K2O, (e) SiO2, and (f) CaO/Al2O3 versus MgO for hydrous and anhydrous Karoo basalt's LLDs (see details in Fig. 23). The same comparison was conducted for the Central Atlantic, Emeishan, Deccan, Siberian, and Columbia River CFBs and the Basin and Range basalts (see Appendix Figs. 7–12).

similar systematic error, about 1–2 wt%. Considering the similarity and large errors for these two methods, the H2O contents constrained by the two independent methods are comparable. The estimated H2O content of the Amushan basalts by primary melt composition is about 0.4 wt%. This is similar to the average H2O concentration in different primary Hawaiian magmas (0.4–0.6 wt%, which mostly varying from 0.4 to 0.5 wt%; Hauri, 2002). In contrast, the estimates for the Benbatu and Dashizhai imply that water flux melting played an important role in generation of these basalts. The low magma water concentration matches well with asthenospheric mantle-derived trace element distribution patterns (N-MOBR-like), whereas the high magma water concentration matches well with the arc-like signatures. This suggests that these arc-like signatures were produced by water flux melting. 5. Mantle transition zone water filtering: A link between deep-Earth fluid cycling and large-scale intra-continental basaltic magmatism CFBs are LIPs erupted onto continental crust and characterized by anomalously high rates of mantle melting, representing the largest

volcanic events in Earth's history. They are commonly associated with the early stages of continental breakup, but whether they arise due to processes related to the continental lithosphere (e.g., thinning, delamination, and insulation) or instead are derived from melting of a deep-mantle plume remains an issue of discussion (Anderson, 1994, 2005; Beccaluva et al., 2009; Campbell, 2005, 2007; Campbell and Griffiths, 1990; Coltice et al., 2009; Heinonen et al., 2014; Jourdan et al., 2007; Sobolev et al., 2011). The majority of CFBs and some riftrelated basalts display enrichment in fluid-mobile elements and depletion of HFSEs (Figs. 15–21), similar to basalts developed in the subduction environment (Puffer, 2001 and this study). Such a feature is dissimilar to ocean island basalts (OIBs) that are associated with hot spots—the surface expressions of mantle plumes (Morgan, 1971). The occurrence of arc-like signatures correlated with eruption relative to the distance to paleo-subduction zones is evident in some CFB provinces (Wang et al., 2015). The most important feature of CFBs is that every CFB province is located at the margin of a Precambrian craton, commonly over the suture with an accreted terrane (Anderson, 1994; Cox, 1978). The evidence for a craton-accreted terrane setting is particularly well established for the Central Atlantic LIP, where



22

t an in m do gla

g

12

la

nt)

)P

ina

14

or

om

P

px

in

O l+ C

m

Al2O3 (wt%)

+(

18

px

g(d

C

16

Al 2 O 3 m a x = 20wt%

l+

18

22

Karoo

O

Al 2 O 3 m a x = 20wt% Ol +C px +P la

20

Al2O3 (wt%)

32

14 Ol+Cpx

10

10 8

Samples with Eu/Eu*