Large amounts of CO2 are transferred from Earth’s interior to the surface by volcanism. On a geological time scale, the rate of CO2 emission has controlled the evolution of Earth’s atmosphere and climate, as well as the dynamic processes that take place in the mantle and core. The total rate of natural CO2 emission from Earth has been estimated on the basis of CO2 flux from arc, mid-ocean-ridge, and hotspot volcanism. However, previous estimates have overlooked the CO2 emitted from a recently discovered type of volcanism—petit-spot volcanism—that occurs on the deep-sea floor. Here, we measure the CO2 and H2O contents of glassy basalts produced by petit-spot volcanism and estimate the initial contents to be >5 wt% and 1.0–1.1 wt%, respectively. Based on these values and magma flux of petit-spot volcanism, we show that the rate of CO2 emission from petit-spot volcanoes (2.7–5.4 × 1011 g CO2 yr–1) is a few percent of the CO2 emissions from arc and mid-ocean-ridge volcanism, and up to ∼14% of that from hotspot volcanism. Thus, the contribution to the carbon cycle on Earth of the large amounts of CO2 that have been emitted from the deep-sea floor by petit-spot volcanism has not previously been recognized.

The determination of the total rate of CO2 emission by volcanism is the basis for the modeling of the global carbon cycle of Earth, and is absolutely necessary for the understanding of the evolution of the Earth surface, i.e., atmosphere, sea, and life, as well as Earth’s interior (Berner et al., 1983; Kasting, 1993; Kerrick, 2001). The rate of CO2 emission by volcanism has been estimated on the basis of present arc, mid-ocean-ridge, and hotspot volcanism; the total flux is on the order of 1013 g CO2 yr–1 (Williams et al., 1992; Sano and Williams, 1996; Marty and Tolstikhin, 1998; Dasgupta and Hirschmann, 2006). This estimate is used to consider the present and past global carbon cycle of Earth, though the past flux of CO2 emissions still remains an open question (Kerrick, 2001; Dasgupta and Hirschmann, 2010). However, previous estimates do not include the emission of CO2 by the recently discovered petit-spot volcanism.

Petit-spot volcanoes were first found at locations on the 135-m.y.-old Pacific plate near the Japan Trench and ∼600 km to the southeast of the trench (Hirano et al., 2006). More recently, they have been reported on the Nazca plate off the coast of Chile (Hirano et al., 2013). They are believed to represent volcanism along fractures in the lithosphere formed by plate flexure prior to subduction. If petit-spot volcanoes are created by this mechanism, it is possible that they occur on subducting plates around the world. If many more petit-spot volcanoes exist, CO2 emissions produced by them may contribute significantly to the total emissions of CO2 transferred from Earth’s interior to the surface. Here, we measure the CO2 and H2O contents of glassy basalts collected from three petit-spot volcanoes, and investigate the rate of emission of CO2 from petit-spot volcanoes.

Samples of basaltic glass containing olivine crystals were collected from the deep-sea floor at ∼6000 m depth near the axis of the Japan Trench (site A) and at a site ∼600 km southeast of the Japan Trench (site B) (Hirano et al., 2006) (Fig. 1; see the GSA Data Repository1 and Fig. DR1 therein for details of the survey area).

The contents of CO2 and H2O dissolved in the glass samples were determined using infrared spectroscopy (see the Data Repository for details of the analytical methods). To determine the CO2 and H2O contents, absorbances at the 1515 and 1430 cm–1 bands for CO2 (carbonate) (Dixon and Pan, 1995) and at the 3550 cm–1 band for H2O (hydroxyl group) (Dixon et al., 1995) were measured after baseline correction. The concentration of H2O molecules (H2Om) was also determined on the basis of the 1630 cm–1 absorbance (Dixon et al., 1995). The volume fraction of bubbles (vesicularity) in glassy basalts was determined from backscattered electron images obtained by a scanning electron microprobe.

Before the calculation of CO2 and H2O contents, we determined the effect of hydration. When the temperature is higher than the glass transition temperature, the reaction of H2Om with the silicate melt proceeds rapidly (H2Om + O ↔ 2OH, where H2Om, OH, and O represent H2O molecules, the hydroxyl group, and oxygen in the silicate structure, respectively), and a water speciation equilibrium can be achieved (Nowak and Behrens, 1995). On the other hand, the dissociation of H2Om into hydroxyl groups is extremely slow below the glass transition temperature; hence, hydration at low temperatures results in a relatively high H2Om content (Yokoyama et al., 2008).

The concentration profiles of CO2 and H2Om around bubbles are shown in Figure 2. The concentration of H2Om increases toward the bubbles; this increase was caused by hydration. Because the hydration layers cover a distance of ∼50 μm (samples KR07-06 7K#387R09-2 and KR04-08 D08), analyses of the glass part more than ∼50 μm away from the bubbles were performed to determine the original H2O content in glassy basalts. The distance of 50 μm and the age of 1–10 Ma for the rock samples (Hirano et al., 2006) provide a diffusion coefficient on the order of 10−11 to 10−12 μm2 s–1 for the H2O molecules in basaltic glass, on the basis of the relation DL2/t, where D is the diffusion coefficient, L is the diffusion distance, and t is time. This value is comparable to the diffusion coefficient of H2O molecules in rhyolitic glasses at temperatures of 15–21 °C (Yokoyama et al., 2008).

In contrast, the CO2 content shows no concentration variation. The absence of variation in the CO2 concentration implies that the diffusion of CO2 was fast enough to maintain the equilibrium concentration according to the solubility curve during magma ascent. There may be a very thin concentration profile around bubbles (<15 μm), which is scarcely measured by the Fourier transform infrared microspectrometer due to the limitation of spatial resolution. If the concentration profile within a distance <15 μm was formed by the diffusion of CO2 into the bubble, the diffusion time is estimated to be less than ∼22.5 s on the basis of the relationship tL2/D, because the diffusion coefficient of the CO2 in basalt melts is ∼10 μm2 s–1 (Zhang et al., 2007) at a temperature of 1200 °C (Hirano et al., 2004). This time is too short to be explained by the phenomenon of magma ascent. For example, if the magma ascent time required to erupt from the depth of 10 km is 20 s, an unrealistic ascent velocity of 500 m s–1 would be necessary. We therefore infer that the CO2 degassed at equilibrium during magma ascent, and magma was accelerated near the surface and rapidly cooled in seawater after magma fragmentation.

The concentrations of OH and H2Om are shown in Figure 3, together with water speciation at equilibrium at temperatures of 500 °C and 1000 °C (Lesne et al., 2011). The OH content was calculated from the difference between the total H2O and H2Om contents. The measured water speciations (except for a sample [site A sample KR07-06 7K#385R09] whose data are not discussed in the following section because there is a possibility that the H2O content was influenced by the hydration as shown below) are similar to those at equilibrium at a temperature of ∼1000 °C. Because the magma temperature was estimated to be ∼1200 °C, the difference between the magma temperature and the equilibrium water speciation temperature is ∼200 °C. When we consider the hydration of 0.1 wt% H2Om, the calculated equilibrium temperature decreases by ∼150 °C. Hence, if the quenched temperature (∼1000 °C) had been influenced by the hydration of more than 0.1 wt% H2Om, magma temperature would be >1200 °C. Therefore, the maximum uncertainty of the total H2O content induced by hydration is estimated to be 0.1 wt%. In contrast, one sample (site A sample KR07-06 7K#385R09) shows water speciation at equilibrium at a temperature of ∼500 °C. This may reflect the effect of hydration and/or the slow cooling of samples. The hydration causes an increase in the H2Om content, resulting in a decrease in the equilibrium temperature recorded in water speciation. The slow cooling of magma also decreases the equilibrium temperature because it allows the reaction between water and the silicate melt to proceed (Nowak and Behrens, 1995).

We used only the data that were not influenced by hydration, and determined CO2 and H2O contents of glassy basalts to be 547–949 ppm and 0.9–1.0 wt%, respectively, which we interpreted to represent the saturation points for petit-spot magma at pressures of 156–185 MPa (Fig. 4A). These saturation pressures were much higher than the hydrostatic pressure on deep-sea floors at ∼6000 m (∼60 MPa), which indicates that the petit-spot magmas were vigorously released into seawater.

The measured CO2 and H2O contents represent residues after magma degassing, which is induced by decompression during volcanic eruption. The volatile solubility decreases as decompression proceeds during magma ascent, resulting in the growth of gaseous bubbles. Thus, we estimated the initial CO2 and H2O contents by tracing back the degassing path from the measured contents. The degassing path was simulated using a solubility model of CO2 and H2O in magma that calculates thermodynamic equilibrium between gaseous and liquid volatile components at given temperature, pressure, and chemical composition (Papale et al., 2006). Equilibrium degassing during magma ascent was assumed because no evidence of diffusive disequilibrium of CO2 and H2O was observed around the bubbles (Fig. 2).

The degassing path shows that at high pressures, the CO2 content decreased rapidly as decompression proceeded and that the decrease in H2O content accelerated at low pressures (Figs. 4B–4D). If initial H2O contents of >0.9–1.0 wt% are assumed, the measured H2O contents are reproduced by the degassing path. On the other hand, the measured CO2 contents can be explained by a wide range of initial CO2 contents. For example, the measured CO2 and H2O contents of sample KR04-08 D07 (site B) can be roughly simulated by assuming initial CO2 and H2O contents of either 1–5 wt% and 1.0 wt%, respectively, or 10 wt% and 1.1 wt%, respectively (Fig. 4C). Thus, the initial CO2 content cannot be determined from the degassing path alone. Hence, we used the volume fraction of bubbles (vesicularity) in the samples to estimate the initial CO2 content. Because the degassed CO2 and H2O formed gaseous bubbles in the magma, the vesicularity can be calculated from the difference between the initial and measured contents of CO2 and H2O. The vesicularity of 43 vol% of sample KR04-08 D07 can be explained by initial CO2 and H2O contents of 10 wt% and 1.1 wt%, respectively (Fig. 4C). All samples included bubbles and, when the initial CO2 content is assumed to be greater than ∼5 wt%, the vesicularities of all samples (22–43 vol%) are well reproduced (Figs. 4B–4D). If the degassing process involves gas escape (i.e., open-system degassing), more CO2 is needed to explain the vesicularity. In contrast, the other volatiles such as sulfur may contribute to the formation of gas bubbles, but their contents are thought to be much lower than 5 wt% in basaltic magma (Mavrogenes and O’Neill, 1999). Hence, we consider that the CO2 content of the erupted magma was more than 5 wt%.

The worldwide amount of CO2 emissions from petit-spot volcanoes is dependent on the flux of their magmas, which is not well understood. However, the amount of emissions can be approximated from the number and size of volcanoes found on subducting plates. More than 100 petit-spot volcanoes that are younger than 8.5 Ma have been found within the ∼112,000 km2 area of the Pacific plate near the Japan Trench (Hirano et al., 2008). The volume of magma released from each volcano has been estimated to be >0.1–1 km3 because the average volumes of volcanic edifices are 0.181 and 0.086 km3 for sites A and B (Hirano et al., 2008), the volume of magma intrusion of each volcano is ∼1 km3 on the basis of a seismic reflection survey (Fujiwara et al., 2007), and fragmented and dispersed magmas, which were not counted, seem to contribute to the emission of CO2. Additional volcanoes are also expected on the Pacific plate because only 36% of the area has been covered by the shipboard acoustic surveys (Fig. DR1). These data imply magma fluxes of 4.12 × 10−5 km3 yr–1 near the Japan Trench (Table 1). In contrast to the magma flux near the Japan Trench, the flux at the Tonga Trench is estimated to be 1.51 × 10−6 km3 yr–1 (Table 1). The difference of magma flux at the Japan and Tonga Trenches may be caused by the different age of the plate and other controlling factors, but it is still an open question. As a first estimate, we use the average of the magma flux at the Japan and Tonga Trenches to estimate the CO2 flux by petit-spot volcanism.

If petit-spot volcanism occurs on subducting plates around the world, the amount of naturally emitted CO2 may be greater than previous estimates. Worldwide, trenches associated with subduction extend over a total length of ∼40,000 km (Reymer and Schubert, 1984), and the survey lengths at the Japan and Tonga Trenches are 440 km and 170 km, respectively (Table 1). Thus, we estimate that the total magma flux for this type of volcanism could be 2.05 × 10−3 km3 yr–1 (Table 1). Assuming 5–10 wt% CO2 in magma, we estimate the associated CO2 emission to be up to 2.72–5.43 × 1011 g yr–1 (Table 1). This emission rate represents 0.1%–1.2% of the emissions from arc and mid-ocean-ridge volcanism (6.6–14 × 1013 and 4.4–22 × 1013 g yr–1, respectively) (Dasgupta and Hirschmann, 2010) and 0.3%–14% of that from hotspot volcanism (0.4–11 × 1013 g yr–1) (Dasgupta and Hirschmann, 2010). Our estimate may be low because it includes only the CO2 that erupts with magma, whereas during subaerial volcanism, magmatic volatiles are released from magma without eruption (Shinohara, 2008; Burton et al., 2013). For example, volcanic CO2 flux from persistently degassing volcanoes such as Mount Etna (Italy) and Kilauea (Hawaii) volcanoes is on the order of 1012 g yr–1 (Burton et al., 2013). Therefore, we suggest that large amounts of CO2 that erupt from the deep-sea floor during petit-spot volcanism may have made a previously unrecognized contribution to the carbon cycle on Earth.

This study received financial support from Grants-in-Aid for Scientific Research in Japan (no. 21684025) and a Toray Science and Technology grant (no. 11-5208). The solubilities of CO2 and H2O in magma were calculated using SOLWCAD code ( We thank two anonymous referees and Andrew Barth for their constructive comments.

1GSA Data Repository item 2013325, details of survey area and analytical methods, is available online at, or on request from or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.