Sr-Nd-Pb-Ca Isotopes of Holocene Basalts from Jingpohu, NE China: Implications for the Origin of Their Enriched Mantle Signatures

: The geochemistry on Holocene lavas from the Jingpohu volcanic ﬁeld in NE China are compared with other Cenozoic lavas from across the back-arc rift of NE China, in order to constrain their enriched mantle sources. Holocene lavas within Jingpohu volcanic ﬁeld comprise two separate “Crater Forest” (CF) and “Frog Pool” (FP) volcanic areas. FP lavas have lower MgO, CaO, and heavy rare earth elements and higher Al 2 O 3 , Na 2 O, K 2 O, and large-ion lithophile elements than CF lavas. Yet, both CF and FP lavas share similar isotopic signatures, with depleted Sr and Nd isotopes ( 87 Sr/ 86 Sr = 0.703915–0.704556, 143 Nd/ 144 Nd = 0.512656–0.512849) and unradiogenic Pb isotopes ( 208 Pb/ 204 Pb = 37.79–38.06, 207 Pb/ 204 Pb = 15.45–15.54, 206 Pb/ 204 Pb = 17.49–18.15), similar to oceanic island basalts. An important new constraint for the Jingpohu lavas lies in their Ca isotopes of δ 44/40 Ca from 0.63 to 0.77‰, which are lower than that of the bulk silicate earth (0.94 ± 0.05‰). By comparing the isotopic signatures of sodic lavas with that of the potassic lavas across NE China, we propose a three-component mixing model as the source for the sodic lavas. In consistence with geophysical results, we propose that subducting Paciﬁc plate induces asthenospheric mantle upwelling of an upper depleted mantle (DM), including subducted ancient sediments (EM I), which partially melted upon ascent. These primary melts further interacted with the lithospheric mantle (EM II), before differentiating within crustal magma chambers and erupting.


Introduction
Enriched mantle components typically have low 143 Nd/ 144 Nd, variable 87 Sr/ 86 Sr, and high 208 Pb/ 204 Pb and 207 Pb/ 204 Pb ratios (at a given value of 206 Pb/ 204 Pb) [1] and can be further subdivided into an EM I with relatively low 87 Sr/ 86 Sr ratios and an EM II with relatively high 87 Sr/ 86 Sr ratios [1]. Within the Cenozoic back-arc rift of NE China (Figure 1), both potassic and sodic lava suites have enriched chemical as well as isotopic signatures that are arguably derived from enriched mantle sources [2][3][4]. In recent decades, an increasing number of studies indicate that the potassic suites can be explained by a mixture of an enriched EM I and a depleted mantle (DM) component [4,5]. However, the chemical and isotopic signatures of sodic suites appear to require an additional enriched EM II source [6].
Several geodynamic models have tried to locate the origins of the enriched mantle components beneath NE China, including the metasomatized lithosphere [9][10][11], recycled sediments brought up by a mantle plume [3], and enriched fluids in the mantle transition zone released by the dehydration of the present Pacific slab and/or earlier subducted plates [4,6,12,13]. While it is widely accepted that the westward subduction of the Pacific plate plays a signifi-

Geological Background
Despite being located up to 2400 km behind the west Pacific subduction zone, many intraplate volcanoes, during the Cenozoic, have formed across NE China, e.g., Jingpohu, Changbaishan, and Wudalianchi volcanoes, inside what is best described as a back-arc rift setting. Spatially, the province's potassic rocks are mainly restricted within the northernmost part of China, such as Wudalianchi-Erkeshan-Keluo (WEK) and Nuomihe volcanic fields (Figure 1), while sodic lavas, including the Changbaishan, Longgang, and Jingpohu volcanic fields, scatter across a much larger area to the south and east of these ( Figure 1). As revealed through tomographic imaging, the subducted slab of the Pacific plate levels out along the base of the upper mantle [14,15] and could, thereby, likely play an additional role in supplying magma source components to the intraplate volcanism. However, whether the Pacific slab can still contribute to Cenozoic volcanism in NE China remains controversial [2,10,15].
The Jingpohu volcanic field is located close to Dunhua-Mizhi fault and erupted from Eocene to Holocene, with the most recent eruptions 5500-5200 year before present (BP) [16,17]. Magnetotelluric (MT) low-resistivity [18,19] and seismic low-velocity [20,21] anomalies at depths of 8-16 km and 30-100 km record likely partial melting beneath the Jingpohu volcanoes but at a lower degree than below the Changbaishan volcano [21]. During a 4-month mobile seismic survey in 2002, 39 seismic events mostly within the Jingpohu [22] indicated that the Jingpohu volcanoes may still erupt.
Thirteen young Holocene volcanic craters are found within the XiaoBeihu Forest and range in size from being dozens of meters to more than 400 m in diameter and from less than 10 m to more than 100 m in depth. While 11 of these craters cluster within a southern "Crater Forest" (CF) volcanic area, only two craters comprise a separate "Frog Pool" (FP) volcanic area in the north (Figure 2). A larger lava flow from the CF craters snaked down a river valley, blocked an older channel of the Mudan River and, thereby, formed the Jingpohu Lake. Located approximately 15 km apart, a correspondingly smaller lava flow emanates from the crater pair within the FP volcanic area. Even if carbon-14 ages show that the two volcanic areas erupted synchronously at 5200-5500 year BP [16,17], the petrology and geochemistry of the lavas from the two volcanic areas differ [23].
Minerals 2021, 11, x FOR PEER REVIEW 3 of 20 volcanic area. The study, furthermore, includes high-precision chemical and isotopic data, which will help constrain the sources of enriched mantle components beneath NE China.

Geological Background
Despite being located up to 2400 km behind the west Pacific subduction zone, many intraplate volcanoes, during the Cenozoic, have formed across NE China, e.g., Jingpohu, Changbaishan, and Wudalianchi volcanoes, inside what is best described as a back-arc rift setting. Spatially, the province's potassic rocks are mainly restricted within the northernmost part of China, such as Wudalianchi-Erkeshan-Keluo (WEK) and Nuomihe volcanic fields (Figure 1), while sodic lavas, including the Changbaishan, Longgang, and Jingpohu volcanic fields, scatter across a much larger area to the south and east of these ( Figure 1). As revealed through tomographic imaging, the subducted slab of the Pacific plate levels out along the base of the upper mantle [14,15] and could, thereby, likely play an additional role in supplying magma source components to the intraplate volcanism. However, whether the Pacific slab can still contribute to Cenozoic volcanism in NE China remains controversial [2,10,15].
The Jingpohu volcanic field is located close to Dunhua-Mizhi fault and erupted from Eocene to Holocene, with the most recent eruptions 5500-5200 year before present (BP) [16,17]. Magnetotelluric (MT) low-resistivity [18,19] and seismic low-velocity [20,21] anomalies at depths of 8-16 km and 30-100 km record likely partial melting beneath the Jingpohu volcanoes but at a lower degree than below the Changbaishan volcano [21]. During a 4-month mobile seismic survey in 2002, 39 seismic events mostly within the Jingpohu [22] indicated that the Jingpohu volcanoes may still erupt.
Thirteen young Holocene volcanic craters are found within the XiaoBeihu Forest and range in size from being dozens of meters to more than 400 m in diameter and from less than 10 m to more than 100 m in depth. While 11 of these craters cluster within a southern "Crater Forest" (CF) volcanic area, only two craters comprise a separate "Frog Pool" (FP) volcanic area in the north (Figure 2). A larger lava flow from the CF craters snaked down a river valley, blocked an older channel of the Mudan River and, thereby, formed the Jingpohu Lake. Located approximately 15 km apart, a correspondingly smaller lava flow emanates from the crater pair within the FP volcanic area. Even if carbon-14 ages show that the two volcanic areas erupted synchronously at 5200-5500 year BP [16,17], the petrology and geochemistry of the lavas from the two volcanic areas differ [23].

Analytical Method
Eleven lava samples were selected for geochemical analysis, including 8 from the FP and 3 from the CF volcanic area. The altered surfaces of collected samples were cut off, and only fresh parts crushed to an approximately 5 mm sized fragments by a steel jaw crusher. Fragments were hand-picked and ground into fine powder in an agate ball mill. Powders were then dried at 105 • C for 12 h in oven before being processed for whole-rock major and trace elements, as well as Sr-Nd-Pb-Ca isotopic analyses at the Sample Solution Lab in Wuhan, China.

Major and Trace Elements
Whole-rock major element analyses were conducted on a Primus II X-ray fluorescence machine (XRF; Rigaku, Japan). Firstly, approximately 1 g of dried sample powder was precisely weighted and fused at 1000 • C for 2 h inside a muffle furnace. After being cooled, the fused sample was weighted again for calculating any loss on ignition (LOI). Then, 0.6 g of the cooled sample was mixed with 6.0 g cosolvent (Li 2 B 4 O 7 :LiBO 2 :LiF = 9:2:1) and 0.3 g oxidant (NH 4 NO 3 ) in a Pt crucible and melted at 1150 • C for 14 min. Finally, the remelted sample was cooled into a flat disc for XRF analysis. For quality control, two Chinese national geological reference materials GBW07105 (basalt) and GBW07111 (granodiorite) were measured and found in good agreement with suggested values. Duplicate measurements of sample WJ-11 also demonstrate good reproducibility of the analytical method.
Whole-rock trace element analyses were performed on an Agilent 7700e inductively coupled plasma mass spectrometer (ICP-MS). Approximately 50 mg samples were weighed and dissolved in Teflon bombs with 1 mL HNO 3 and 1 mL HF. These Teflon bombs were secured in stainless steel pressure jackets and heated for more than 24 h at 190 • C in an oven. After cooling, the Teflon bombs were carefully opened and evaporated to incipient dryness at 140 • C, and then, 1 mL HNO 3 was added and again evaporated to dryness in order to remove fluorides from the initial HF dissolution. Afterwards, 1 mL HNO 3 , 1 mL MQ water, and 1 mL internal standard solution (1 ppm In) were added to the sample, sealed, and kept in an oven at 190 • C for more than 12 h. Finally, the solutions were transferred to a polyethylene bottle and diluted to 100 g by adding 2% HNO 3 for the ICP-MS analysis. Based on measurements of the standard AGV-2, BHVO-2, BCR-2, and RGM-2 and duplicate measurement of sample WJ-11, standard deviation is estimated to be <6% from suggested values for most elements, while the reproducibility of most elements is <7%. Total procedural blanks yielded element concentrations that are significantly lower than those of the Jingpohu samples.

Chemical Separation
Digestion and column chemistry were employed to separate Sr-Nd-Pb-Ca elements for isotopic analysis. A total of three digestions were conducted, one for separating Sr and Nd, one for Pb, and another for Ca. For Sr and Nd separations, approximately 100 mg of dried sample powder was digested using the same method as for the trace element digestion described above. Sample solutions were loaded with 2.5 M HCl into ion-exchange columns packed with AG50W resin. Strontium fractions were eluted with 2.5 M HCl, and REE cuts were collected using 4 M HCl. Then, Sr fractions were purified in a second column loaded with Sr-specific resin, and REE cuts were loaded in a third column with Ln resin in order to separate the Nd. Another digestion of 100 mg sample was performed for the separation of Pb. These digested solutions were loaded with 1M HBr into columns with AG resin. After removing undesirable matrix elements with 1 M HBr, Pb fractions were eluted using 6 M HCl. Calcium separation required another digestion of approximately 50 mg sample using the same method. An aliquot sample solution containing 40 µg Ca was loaded into the PFA column with 250 µL DGA resin. The matrix was removed by 4 M HNO 3 , and Ca fractions were collected with MQ water. The collected Ca fractions were evaporated to dryness and dissolved in 2% HNO 3 in order to achieve 10 ppm Ca solutions for MC-ICP-MS analysis.

Mass Spectrometry
The Sr-Nd-Pb isotope analyses were performed on a Neptune Plus MC-ICP-MS (Thermo Fisher), which was equipped with seven fixed electron multiplier ICs, and nine  208 Pb on L3 to H3. Each measurement consisted of 10 blocks of 10 cycles. Standard NIST SRM 987, JNdi-1, and NBS SRM 981 were used to evaluating the accuracy and reproducibility for Sr, Nd, and Pb isotopes, respectively.
Calcium isotope analyses were also conducted on the plasma equipped with a quartz dual cyclonic-spray chamber and a Savillex 50 uL/min PFA MicroFlow Teflon nebulizer. Usually, a 10 ppm 44 Ca + solution can obtain a signal above 5 V. Before each injection, the sample-introduction system was cleaned with 5% HNO 3 for 2-3 min in order to keep the 44 Ca + signal below 1 mV and, thereby, avoid cross-contamination between samples. A standard sample bracketing method was employed for correcting instrumental drift. Due to a limited solution of commonly used standard NIST SRM 915a, an in-house Alfa Ca standard solution (Lot: 9192737) was used for bracketing in this laboratory. For comparison, intermediate measurements of NIST SRM 915a were also performed, and the results were reported relative to that NIST SRM 915a by adding a conversion factor of 0.58. Consequently, Ca isotope compositions are reported as: δ 44/42 Ca sample = [ 44 Ca/ 42 Ca sample / 44 Ca/ 42 Ca NIST SRM 915a − 1] × 1000. Each sample was measured 3 times. In addition, a calcium carbonate standard NIST SRM915b, a basalt rock standard BHVO-2, and seawater samples were also analyzed to evaluate both accuracy and reproducibility. The long-term average calculated δ 44/42 Ca SRM915a is 0.001 ± 0.058‰ (2 SD, n = 155).

Petrography
All lavas are porphyritic, but with megacrysts only found within FP lavas and never in CF lavas. Many phlogopite and anorthoclase megacrysts were found inside lavas that were located along at the rim of the FP craters (Figure 3a,b), while a previous study has also found clinopyroxene megacrysts in these FP lavas [23]. Phenocrysts (<15%) in the FP lavas are otherwise mainly euhedral olivines and set in a glassy groundmass with plagioclase and olivine microlites ( Figure 3c). The lavas from the CF volcanic area comprise 5-15% phenocrysts of olivine, clinopyroxene, and plagioclase set in a glassy ground mass with pyroxene, olivine, and plagioclase microlites ( Figure 3d).

Whole-Rock Major Elements
Whole-rock major elements of the Jingpohu samples are provided in Table 1 and are, for comparison, consistently plotted together with literature data of Holocene samples from Jingpohu volcanic field. The total alkali versus silica diagram (TAS), after a calculation to 100% on an anhydrous basis (Figure 4a), shows that compositions for the Holocene FP and CF volcanic areas are significantly different. These FP samples mostly cluster along the boundary between the diagram's phono-tephrite and tephri-phonolite fields, whereas CF samples mostly range across its tephrite/basanite and trachy-basalt field. All Holocene samples display K2O/Na2O ratios between 0.5 and 0.8 (Figure 4b), exhibiting correlations with each other and belonging to the shoshonitic series. CF samples can be further subdivided into a low-and a high-alkali group. The three CF samples of this study all belong to the high-alkali basanitic (~15% normative olivine) group, which also has slightly lower SiO2 contents than that of the low-alkali trachy-basaltic group. Jingpohu samples yield SiO2 contents between 43.76 and 50.08%, and even if FP and CF samples have overlapping SiO2 ranges, the two datasets exhibit significant differences in all other major elements except for Fe2O3 (T) and TiO2. As shown by Harker diagrams (Figure 5), FP samples have higher Al2O3, Na2O, and K2O contents and lower MgO and CaO compared to the CF samples.

Whole-Rock Major Elements
Whole-rock major elements of the Jingpohu samples are provided in Table 1 and are, for comparison, consistently plotted together with literature data of Holocene samples from Jingpohu volcanic field. The total alkali versus silica diagram (TAS), after a calculation to 100% on an anhydrous basis (Figure 4a), shows that compositions for the Holocene FP and CF volcanic areas are significantly different. These FP samples mostly cluster along the boundary between the diagram's phono-tephrite and tephri-phonolite fields, whereas CF samples mostly range across its tephrite/basanite and trachy-basalt field. All Holocene samples display K 2 O/Na 2 O ratios between 0.5 and 0.8 (Figure 4b), exhibiting correlations with each other and belonging to the shoshonitic series. CF samples can be further sub-divided into a low-and a high-alkali group. The three CF samples of this study all belong to the high-alkali basanitic (~15% normative olivine) group, which also has slightly lower SiO 2 contents than that of the low-alkali trachy-basaltic group. Jingpohu samples yield SiO 2 contents between 43.76 and 50.08%, and even if FP and CF samples have overlapping SiO 2 ranges, the two datasets exhibit significant differences in all other major elements except for Fe 2 O 3 (T) and TiO 2 . As shown by Harker diagrams ( Figure 5), FP samples have higher Al 2 O 3 , Na 2 O, and K 2 O contents and lower MgO and CaO compared to the CF samples.   [24,25]. For comparison, literature data of Holocene Jingpohu samples [2,[26][27][28] are also plotted.

Whole-Rock Trace Elements
Whole-rock trace elements of the Holocene Jingpohu samples are presented in Table  1 and in Figure 6. As illustrated by the primitive mantle-normalized spider diagram (Figure 6a), the Jingpohu samples exhibit more enriched patterns than an oceanic island basalt (OIB), but share some of its characteristics, including an absence of negative Nb-Ta anomalies. A chondrite-normalized rare earth element (REE) diagram ( Figure 6b) shows that Jingpohu samples are characterized by enriched light rare earth elements (LREE) relative to heavy rare earth elements (HREE). Compared with each other, the FP samples exhibit slightly higher concentrations of large-ion lithophile elements (LILE) and lower HREEs than those of CF samples ( Figure 6).

Whole-Rock Trace Elements
Whole-rock trace elements of the Holocene Jingpohu samples are presented in Table 1 and in Figure 6. As illustrated by the primitive mantle-normalized spider diagram (Figure 6a), the Jingpohu samples exhibit more enriched patterns than an oceanic island basalt (OIB), but share some of its characteristics, including an absence of negative Nb-Ta anomalies. A chondrite-normalized rare earth element (REE) diagram (Figure 6b) shows that Jingpohu samples are characterized by enriched light rare earth elements (LREE) relative to heavy rare earth elements (HREE). Compared with each other, the FP samples exhibit slightly higher concentrations of large-ion lithophile elements (LILE) and lower HREEs than those of CF samples ( Figure 6).
FP samples are also slightly more LREE enriched than CF samples, as shown by somewhat higher La n and especially by higher (La/Yb) n ratios (subscript n denotes to their normalization with a primitive mantle), which for both suites, correlate negatively with SiO 2 contents (Figure 7a,b). Samples do not exhibit evident Eu anomalies, with Eu/Eu* = 0.93-1.07 (Eu/Eu* = Eu n × 2/(Sm n + Gd n )). The Jingpohu samples have Ti/Ti* ratios that range from 0.57 to 0.9 (Ti/Ti* = Ti n × 2/(Eu n + Tb n )) but without any correlation with SiO 2 .   Ocean island basalt (OIB) [29] and lavas from Changbaishan [11] and Longgang [2] volcanic fields are presented for comparison. Values of primitive mantle and chondrite are from [29].
FP samples are also slightly more LREE enriched than CF samples, as shown by somewhat higher Lan and especially by higher (La/Yb)n ratios (subscript n denotes to their normalization with a primitive mantle), which for both suites, correlate negatively with SiO2 contents (Figure 7a,b). Samples do not exhibit evident Eu anomalies, with Eu/Eu* = 0.93-1.07 (Eu/Eu* = Eun × 2/(Smn + Gdn)). The Jingpohu samples have Ti/Ti* ratios that range from 0.57 to 0.9 (Ti/Ti* = Tin × 2/(Eun + Tbn)) but without any correlation with SiO2. Figure 6. Primitive mantle-normalized incompatible element (a) and chondrite-normalized rare earth element diagrams (b). Ocean island basalt (OIB) [29] and lavas from Changbaishan [11] and Longgang [2] volcanic fields are presented for comparison. Values of primitive mantle and chondrite are from [29].
OR PEER REVIEW 11     Table 2 presents the whole-rock Sr-Nd-Pb-Ca isotopes of the Holocene lavas from Jingpohu volcanic field together with the USGS and other reference materials. For comparison, these are presented together with other published isotopic data on the Jingpohu as well as other sodic and potassic volcanic fields in NE China. As shown in Figure 8a, the Holocene lavas from Jingpohu volcanic area exhibit depleted Sr and Nd isotopes relative to a bulk silicate earth (BSE) [30], with 87 [30], and result in values that range from 0.63 to 0.77‰ and are lower than that of the BSE (0.94 ± 0.05‰; Figure 8d) [31]. Additionally, the CF and FP samples have overlapping Sr-Nd-Pb-Ca isotopic characteristics.

Crustal Assimilation and Fractional Crystallization
During magma ascent, fractional crystallization may modify the chemical compositions of the parental magma and can be evaluated through correlations with a geochemical differentiation index. The presence of anorthoclase, phlogopite, and clinopyroxene megacrysts in the FP lavas suggests that these magmas experienced prolonged durations of medium-to high-pressure crystallization before eruption [23], which would likely also allow for fractionation at depth. Thus, the negative Ba anomalies of FP lavas (Figure 6a) are likely the result of anorthoclase fractionation, while it is otherwise difficult to specify any other fractionating phase in this way. Neither of the two suites exhibit any significant Eu anomaly, negating the fractionation of plagioclase (Figure 7c). Both suites exhibit weak correlations between SiO2 and other elements, such as MgO, Al2O3, Na2O, K2O, and La in  Table S1). The field for EM I-type (magenta; Pitcairn-Gambier Chain) and EM II-type oceanic island basalts (green; Samoa) are from the GEOROC database (http://georoc.mpch-mainz.gwdg.de/georoc/ accessed on 3 March 2021). The δ 44/40 Ca value for bulk silicate earth (0.94 ± 0.05‰) is from [31].

Crustal Assimilation and Fractional Crystallization
During magma ascent, fractional crystallization may modify the chemical compositions of the parental magma and can be evaluated through correlations with a geochemical differentiation index. The presence of anorthoclase, phlogopite, and clinopyroxene megacrysts in the FP lavas suggests that these magmas experienced prolonged durations of medium-to high-pressure crystallization before eruption [23], which would likely also allow for fractionation at depth. Thus, the negative Ba anomalies of FP lavas (Figure 6a) are likely the result of anorthoclase fractionation, while it is otherwise difficult to specify any other fractionating phase in this way. Neither of the two suites exhibit any significant Eu anomaly, negating the fractionation of plagioclase (Figure 7c). Both suites exhibit weak correlations between SiO 2 and other elements, such as MgO, Al 2 O 3 , Na 2 O, K 2 O, and La in Figures 5 and 7, indicating some degree of fractional crystallization that could potentially also relate to different proportions of mega/phenocrysts. In contrast, there are no obvious correlations between SiO 2 and Fe 2 O 3 (T) (Figure 5d) or Ti anomaly (Figure 7d), which preclude any systematic involvement with oxides.
Due to the distinct compositional differences between the crust and mantle-derived magmas, crustal assimilation could also explain some of the chemical and isotopic variations exhibited by Holocene Jingpohu samples. However, their low SiO 2 contents ( Figure 5), depleted Sr-Nd isotopes (Figure 8), and lack of any negative Nb-Ta anomalies (Figure 6a) are all inconsistent with any significant contributions of crustal assimilation.

Differences between the CF and FP Lavas
Carbon-14 dating results showed that the CF and FP lavas in the Jingpohu volcanic field erupted synchronously between 5200 and 5500 years BP [16,17]. However, even though both suites have similar Sr-Nd-Pd-Ca isotope signatures, each has distinctly different geochemical characteristics, including lower MgO and CaO contents and higher Al 2 O 3 , Na 2 O, and K 2 O contents of FP lavas compared to those of the CF lavas. Since clinopyroxene megacrysts-only found in FP lavas-also exhibited an opposite higher MgO and CaO contents and lower Al 2 O 3 , Na 2 O, and K 2 O contents ( Figure 5), yet had comparable SiO 2 contents, we suggest that the geochemical differences between the CF and FP lavas can mainly be attributed to FP lavas having experienced high-pressure clinopyroxene fractionation, while more primitive CF trachybasalts-basanites never crystallized clinopyroxene at depth. Although clinopyroxene is also commonly more HREE-enriched relative to LREEs [32], clinopyroxene fractionation modeling cannot alone lower HREEs as much as observed in the FP lavas. An alternative interpretation for the fractionated HREEs is different partial melting degrees of a garnet-bearing source, which is also consistent with negative correlations between SiO 2 and La n and (La/Yb) n (Figure 7a,b). This scenario can also explain the relatively large SiO 2 variation (43.76-50.08%), while most other oxides remain almost constant ( Figure 5). Additional fractionation of anorthoclase is needed to also explain the negative Ba anomaly exhibited by FP lavas.
Bai et al.'s review [13] of Jingpohu lavas attributed the lower HREE of FP lavas to an eclogite-bearing peridotite source, while the CF lavas were derived from a garnet peridotite. However, since magmas from different sources are likely to inherit different isotopic characteristics, we prefer to think that the coeval CF and FP suites acquired their geochemical differences after having been derived from a common source assemblage. Therefore, based on geochemical differences, we suggest that the CF and FP lavas originated from a same garnet-bearing source experiencing different melting degrees, while the FP lavas further fractionated clinopyroxene and anorthoclase megacrysts at depth and before eruption.

Mantle Petrogenesis for the Sodic Lavas
The Holocene Jingpohu lavas exhibit low SiO 2 contents and K 2 O/Na 2 O ratios that make these sodic as opposed to Nuomihe and Wudalianchi-Erkeshan-Keluo's potassic suites (Figure 4b). While samples exhibit steeper incompatible element patterns than OIBs ( Figure 6), with (La/Yb) n ratios between 16.6 and 61.0 (Figure 7b), they have similar depleted Sr-Nd isotopes (Figure 8a) and overall patterns on the primitive mantle-normalized diagram (Figure 6a). These features not only characterize Jingpohu lavas but are shared by Cenozoic sodic lavas from the nearby Changbaishan and Longgang volcanic fields (Figures 6 and 8), suggesting that these three volcanic fields may have similar petrogeneses and mantle sources.   (Figure 1), exhibiting the most depleted Sr-Nd isotopes and positive Nb-Ta and negative Pb anomalies, are identified as representing the depleted asthenospheric mantle endmember [4,9]. For Cenozoic potassic lavas, on the other hand, mixing lines between an enriched component (EM I) and the lithosphere on K/U and Hf/*Hf versus 87 Sr/ 86 Sr and 206 Pb/ 204 Pb diagrams suggest a depleted lithospheric endmember [4]. However, by adding sodic suites to the plots, stronger mixing lines between the EM I and a depleted asthenosphere (represented by the Halaha lavas) are achieved (Figure 9). This can be explained by the enriched components having significantly higher Sr and Pb concentrations than those of the depleted component and, thereby, allowing relatively small additions of the enriched components to significantly modify Sr and Pb isotopes, while not influencing K/U and Hf/*Hf ratios as much. A depleted asthenospheric source is further supported by tomographic images that reveal an upwelling upper mantle, triggered by a westward subduction and levelling of the Pacific slab on top of the 660 km discontinuity [14,33], which could, thereby, induce decompressional melting. More importantly, xenoliths from the lithospheric mantle beneath the Changbaishan and Longgang volcanic fields exhibit obvious EM II ( Figure 10) [34,35], rather than depleted mantle signatures. Notably, sodic suites (colored symbols) exhibit mixing lines that differ from potassic suites (black symbols) and may in fact be the only ones associated with a lithospheric EM II component (Figure 9). Since xenoliths from the lithospheric mantle beneath Changbaishan and Longgang volcanic fields exhibit similar EM II signatures ( Figure 10) [34,35], this not only adds a third component to the mix, but also argues for the depleted component to be asthenospheric. Therefore, we propose that the depleted components in all of NE China's Cenozoic volcanoes were sourced from an upwelling asthenospheric, rather than the lithospheric mantle. (Figure 9). Since xenoliths from the lithospheric mantle beneath Changbaishan and Longgang volcanic fields exhibit similar EM II signatures ( Figure 10) [34,35], this not only adds a third component to the mix, but also argues for the depleted component to be asthenospheric. Therefore, we propose that the depleted components in all of NE China's Cenozoic volcanoes were sourced from an upwelling asthenospheric, rather than the lithospheric mantle.  While the Halaha lavas are thought to most closely represent the depleted asthenosphere [4,8], their overlap with Jingpohu lavas in the Sr-Nd isotopic diagram (Figure 8a) exposes another insight. Thus, since the Jingpohu lavas record a δ 44/40 Ca range from 0.63 to 0.77‰, which is significantly lower than that of a BSE (0.94 ± 0.05‰) [31], this argues for their derivation from a source that was more enriched than a typical depleted asthenospheric upper mantle. Even though the Halaha lavas exhibit negative Pb anomalies, While the Halaha lavas are thought to most closely represent the depleted asthenosphere [4,8], their overlap with Jingpohu lavas in the Sr-Nd isotopic diagram (Figure 8a) exposes another insight. Thus, since the Jingpohu lavas record a δ 44/40 Ca range from 0.63 to 0.77‰, which is significantly lower than that of a BSE (0.94 ± 0.05‰) [31], this argues for their derivation from a source that was more enriched than a typical depleted asthenospheric upper mantle. Even though the Halaha lavas exhibit negative Pb anomalies, whereas the Jingpohu lavas display positive Pb anomalies, their other geochemical similarities to OIB [4,8] and overlap in the Sr-Nd isotopic diagram ( Figure 10) suggest that even the depleted source of the Halaha lavas has been metasomatized by an enriched component. Therefore, we propose the depleted DM asthenosphere must be even more depleted than the Halaha and Jingpohu lavas.

Origin of the EM I Component
Both the potassic and sodic lavas from NE China are plotted within or close to the EM I-type OIB area (represented by Pitcairn-Gambier Chain basalts; Figure 8), implying a larger contribution from this EM I component. Prior to the seismic detection of the Pacific slab beneath NE China [36], studies attributed the EM I components to either the continental lithosphere or recycled sediments in a mantle plume. For example, Zhang et al. [9] suggested that Cenozoic potassic lavas in NE China were produced by partial melting of phlogopite-bearing garnet peridotite in the lithosphere. This inference provided a good explanation for the potassic lavas, which often host phlogopite-bearing peridotite xenoliths, while hydrous phases are absent in mantle xenoliths entrained in the sodic lavas from Changbaishan and Longgang lavas. As mentioned, however, their lithospheric peridotite xenoliths have an EM II Sr-Nd isotope signature ( Figure 10) [34,35], rather than EM I, and it is, therefore, unlikely for the EM I component to have originated from the lithospheric mantle.
Based on similar Sr-Nd-Pb isotopic characteristics between the Jingpohu and Hawaii basalts, Zhang et al. [27] proposed a mantle plume source for the Jingpohu lavas. In this scenario, the plume brought up recycled lower crustal materials from the lower mantle, which released enriched fluids that metasomatized the lithospheric mantle. Consequently, they suggested that the metasomatized lithospheric mantle acquired a mixed character between a depleted mantle and EM I. However, seismic tomography shows that the low-velocity anomaly beneath NE China is shallow,~700 km deep [37], and its underlying subducted Pacific plate blocks the ascent of any deep-seated mantle plume transporting recycled crust. Furthermore, since 3 He predominantly originates from the deep earth [38], deep-seated mantle plumes are distinguished from other shallow sources by their extremely high helium isotope ratios ( 3 He/ 4 He), e.g., 13.7-15.9 Ra from Kilauea volcano [39], while volcanic gases from Changbaishan and Wudalianchi retain much lower 3 He/ 4 He < 6 Ra [10,40]. In essence, more recent geophysical and geochemical observations do not support a model where recycled crustal sediments are brought up by a deep mantle plume.
After the Pacific plate was detected beneath NE China (e.g., [36]), several studies have suggested that the EM I component of the sodic lavas originate from fluids released by the subducting plate, e.g., [23]. However, 230 Th excesses of the youngest eruptions in NE China strongly argue against such fluid contribution from stagnant Pacific plate [2]. In addition, if the west Pacific plate is the main contributor of an EM I source, volcanoes closer to Japan Trench should be more enriched in Sr-Nd isotopes, which is inconsistent with more distal Wudalianchi volcanoes (approximately 2000 km from Japan Trench), exhibiting a great contribution from the EM I component than more proximal Changbaishan and Jingpohu volcanoes (approximately 1200 km from Japan Trench) (Figure 8a). Besides, it is doubtful that the west Pacific plate can still dehydrate after subducting more than 1200 km beneath NE China [2]. Even if the Pacific plate could still be dehydrating, it would not induce such a distinct EM I signature, where Cenozoic lavas exhibit 207 Pb/ 206 Pb ratios that are higher than a meteorite isochron, whereas the 207 Pb/ 206 Pb ratio of the Pacific slab is lower than that of the meteorite isochron [41]. Furthermore, potassic lavas in oceanic arcs are characterized by Ce/Pb < 46.5 (Gill et al., 2004), while potassic lavas in NE China have much higher Ce/Pb > 65. In conclusion, although we cannot completely exclude the dehydration of the west Pacific slab, we do not think it was a major contributor towards the EM I source.
Instead, Cenozoic lavas from NE China retain high Ba/Th and 207 Pb/ 206 Pb ratios, which are more consistent with a record of more ancient dehydration events (>1Ga) in the mantle transition zone [42]. By modeling the Pb isotope evolution through time, Wang et al. [4] suggest that the extremely unradiogenic Pb isotopes can be attributed to recycled sediments isolated in the mantle transition zone for~2.2 Ga. This is also consistent with tomographic images, suggesting that the transition zone is extremely hydrous [43,44]. Therefore, after excluding the possibility of an EM I originating from the lithospheric mantle, a mantle plume, or the Pacific slab, we strongly suggest that the EM I components were derived from recycled sediments that have remained isolated within the mantle transition zone for >1 Ga and, therefore, not recycled with the Pacific plate.

Origin of EM II Endmembers
By comparing the isotopic characteristics of Cenozoic potassic and sodic lavas from NE China, we suggested that additional EM II components were involved in the generation of the sodic lavas. The EM II fingerprint in the Jingpohu and Changbaishan lavas was also revealed by [6,13] based on their recognition of a trend towards an EM II endmember in their isotopic plots, and they proposed that it originated from the Pacific slab. However, EM II fingerprints are too sporadically recognized throughout NE Asia. For example, in South Korea, EM II components were detected in the lavas from Jeju Island, but not in nearby volcanic areas such as Jeongok and Baengnyeong [7]. Therefore, the EM II components are more likely formed through localized processes, rather than large-scale processes such as the west Pacific subduction. Localized volcanic structures within the lithosphere beneath Changbaishan volcano have been identified geophysically [45], but more importantly, xenoliths entrained in Changbaishan and Longgang Cenozoic lavas exhibit, as mentioned, isotopic EM II signatures ( Figure 10) [34,35], suggesting that it is lithospheric.
Mantle xenoliths from Changbaishan and Longgang volcanic fields primarily comprise spinel lherzolites, with estimated equilibration temperatures ranging from 750 to 1139 • C, supporting a lithospheric mantle origin [34,35]. Lu-Hf and Sm-Nd model ages of residual clinopyroxene within these mantle xenoliths further suggest that the lithospheric mantle was metasomatized between the Early Proterozoic and Phanerozoic ages [34]. Additionally, decoupling of Sr and Nd isotopes within some xenoliths is consistent with diachronous melt extraction [34], which might have introduced the EM II components to the sodic lavas in NE China. Therefore, we propose that the EM II components most likely originated from the lithospheric mantle that was previously metasomatized during ancient subduction events, prior to the subduction of the West Pacific plate.

Geodynamic Implications by the Sodic Lavas
Since seismic tomography revealed the west Pacific plate beneath NE China, its subduction process and geodynamic impact remain a hot research topic, e.g., [36,46]. Even though geophysical studies suggest that the westward subduction of the Pacific plate played a significant dynamic role in instigating sodic volcanism within a back-arc-like setting, geochemical studies have not found any important material connections between these [2], as also supported by a volcanic gas study in Changbaishan volcanic field [10].
The hydrous upper mantle, imaged as low-velocity anomaly by seismic tomography [46], is attributed to the upwelling of ancient sediments from the mantle transition zone, rather than from a dehydrating west Pacific slab ( Figure 11). These ancient sediments in the mantle transition zone is the main EM I reservoir for sodic volcanism in NE China and may exist mainly as a mantle K-hollandite phase [4,47]. K-hollandite is a major repository of LILE (e.g., K, Rb, Ba, Pb, and LREE) and exists as solid phase throughout the transition zone down to the lower mantle [48]. Subduction of the Pacific slab may have trig-gered a rise in the mantle above the slab [32] and, thereby, brought these ancient sediments to shallower depth. Ascending within an upwelling upper mantle, the decompressional melting of both the ancient sediments and asthenospheric mantle produced a mixed array of EM I and DM components, respectively. When the mixed EM I + DM melts reached the lithosphere, these may locally have interacted with the lithosphere [49] and acquired an additional third EM II contribution from that ( Figure 11).
The hydrous upper mantle, imaged as low-velocity anomaly by seismic tomography [46], is attributed to the upwelling of ancient sediments from the mantle transition zone, rather than from a dehydrating west Pacific slab ( Figure 11). These ancient sediments in the mantle transition zone is the main EM I reservoir for sodic volcanism in NE China and may exist mainly as a mantle K-hollandite phase [4,47]. K-hollandite is a major repository of LILE (e.g., K, Rb, Ba, Pb, and LREE) and exists as solid phase throughout the transition zone down to the lower mantle [48]. Subduction of the Pacific slab may have triggered a rise in the mantle above the slab [32] and, thereby, brought these ancient sediments to shallower depth. Ascending within an upwelling upper mantle, the decompressional melting of both the ancient sediments and asthenospheric mantle produced a mixed array of EM I and DM components, respectively. When the mixed EM I + DM melts reached the lithosphere, these may locally have interacted with the lithosphere [49] and acquired an additional third EM II contribution from that ( Figure 11). Figure 11. Cartoon showing the three-source model for the Cenozoic sodic lavas in NE China. within the mantle transition zone, ancient sediments (EM I) subducted into the mantle transition zone~2.2 Ga ago. During Cenozoic, westward subduction of the Paleo-Pacific and Pacific plates caused the upwelling of ancient sediments (EM I). When the ancient sediments (EM I) were brought up to shallower depth, decompression caused partial melting of both ancient sediments (EM I) and their host asthenosphere (DM). Segregated and ascending primary melts interacted with the lithosphere and acquired their EM II signatures from that.

Conclusions
New chemical and isotopic data of the Holocene lavas from Jingpohu volcanic field are presented. The main results are summarized as follows.
(1) Jingpohu's Holocene lavas are mainly restricted to the CF and FP volcanic areas.
Both FP and CF lavas are otherwise coeval and retain such similar Sr-Nd-Pb-Ca isotopic characteristics, suggesting that they originated from the same sources and experienced similar petrogeneses. The more evolved FP phonotephrites exhibit lower MgO, CaO, and HREEs and higher Al 2 O 3 , Na 2 O, K 2 O, and LILEs than those of CF basanites and trachybasalts, which may be attributed to early and high P fractionation of clinopyroxene and anorthoclase, as well as different partial melting degrees of a garnet-bearing source. (2) By comparing the geochemical characteristics of Jingpohu lavas with other Cenozoic sodic and potassic lavas in NE China, we suggested that sodic lavas originate from three sources, DM, EM I, and EM II, which were derived from the upper asthenosphere, ancient (~2.2 Ga) sediments stored in the mantle transition zone and a locally metasomatized lithospheric mantle, respectively. Isotopic differences between Jingpohu, Changbaishan, and Longgang lavas may be attributed to different proportions of these enriched components during their petrogenesis. (3) Cenozoic volcanism in NE China was likely triggered by the westward subduction of the Pacific slab, inducing the upwelling and decompressional melting of a transitional mantle with ancient sediments. Even though we cannot completely exclude any material contributions from the Pacific slab, we do not think it is necessary to explain a varied petrogenesis of Cenozoic magmatism across NE China.