1 INTRODUCTION

On January 15, 2022, the Hunga Tonga-Hunga Ha'apai (HTHH) volcano (20.536°S, 175.382°W, Fig. 1a–b) erupted violently in the Kingdom of Tonga, southwest Pacific, reaching the volcano explosivity index (VEI) 6 (Poli and Shapiro, 2022). It is the largest volcanic eruption in the last 30 years. The circum-Pacific region is known as "the Ring of Fire", with abundant volcanic eruptions. Large eruptions are usually catastrophic. For example, both the Mt. Pinatubo and the Novarupta volcanos (Fig. 1c–d) have resulted in major climate changes (Fierstein and Hildreth, 1992; Newhall et al., 1996; Self et al., 1996; Robock, 2000). What controlled the formation of large volcanos (Allison et al., 2021; Popa et al., 2021)? Where are the danger zones of catastrophic eruptions (Sheldrake and Caricchi, 2017; Weber et al., 2020; Caricchi et al., 2021)? These questions are of critical importance to the human society, yet remain essentially unsolved. Here we show that slab windows and the subduction of carbonate are two favorable conditions for large explosive eruptions.

2 MATERIAL AND METHOD 2.1 Detailed seismic data compilation

We used the generic mapping tools (GMT) to draw the topography and bathymetric map of the study regions. The seismic data in the study regions are from the ISC-EHB catalog, which is a groomed version of the ISC Bulletin and contains thousands of seismic events from 1964 to 2018 (Engdahl et al., 1998). We counted and plotted the locations of the seismic event with M_b > 5 for the Tonga region (Fig. 2) and M_w > 4 for the Novarupta region (Fig. 3). Seismicity in the vertical section is the distribution of events along with a sideline 20 km wide for both Hunga Tonga-Hunga Ha'apai volcano and Novarupta volcano region. We found the slab tearing near both regions, which is the important condition for the catastrophic eruption.

2.2 Details of Hunga Tonga-Hunga Ha'apai volcanic eruption

The details of Hunga Tonga-Hunga Ha'apai volcanic eruption by collecting the latest reports (Cronin et al., 2017, 2022) and volcanic databases of the volcano (Venzke, 2013).

2.3 Volcano, seamount, and water depth data

A sufficient quantity of representative volcano, topographic and bathymetric data now exists to chronicle the catastrophic eruption, seamount and water depth data over time. Volcano, seamount, and water depth data from previous studies were collected and used to verify our study from different databases, including the Global Volcanism Program (Venzke, 2013), the Seamount Catalog (Koppers et al., 2010), and Bathymetric Data Viewer (Amante and Eakins, 2009). The water depths of Pacific Ocean, calcite compensation depth, and Tonga Trench used in the main text are from Bathymetric Data Viewer (Amante and Eakins, 2009) and previous articles (van Andel, 1975).

3 RESULT AND DISCUSSION 3.1 Volatiles in the HTHH volcano

The HTHH volcano has erupted ca. 5 km³ of volcanic ash and tephra into the air, with abundant CO₂ but only 0.4 Mt of SO₂ (Venzke, 2013; Cronin et al., 2017; Adam et al., 2022; Cronin, 2022). To produce magmas with such a low SO₂, the mantle source should have sulfur contents of less than 20 μg/g. The average sulfur abundance in the upper mantle is about 250 μg/g (Sun and McDonough, 1989). This means that the mantle source of the HTHH volcano is much depleted in sulfur.

Sulfur is a highly incompatible element under the oxygen fugacity of convergent margin magmas. Therefore, previous partial melting can dramatically decrease the sulfur contents in the volcano. In a welldeveloped arc, like the Tonga Arc, the sulfur content in the mantle wedge is lower than that of the upper mantle. Note that, CO₂, water, and other volatiles are also highly incompatible during partial melting. Therefore, a highly depleted mantle source is usually also depleted in volatiles. However, the explosive eruption of the HTHH volcano requires high volatile contents, e.g., CO₂.

Tectonically, the HTHH volcano is located on the Tonga-Kermadec trench (Fig. 1b), which is well known for large eruptions (Cronin et al., 2017; Adam et al., 2022; Cronin, 2022). As an intra-oceanic subduction zone far away from continents, the Pacific Plate subducting underneath Tonga is lack of sediments, with a total sediment thickness of < 100 m (Plank and Langmuir, 1998). The water depths of the Pacific Plate to the east of the Tonga Trench are > 5 000 m (Amante and Eakins, 2009), which are deeper than the carbonate compensation depths (CCD, 4 500 m) of the Pacific Ocean (van Andel, 1975). Therefore, not much carbonate is expected in the sediments, either. Nevertheless, the carbon isotopes of arc volcanic rocks in this region show clear carbonate recycling characteristics, which is attributed to carbonate from the altered oceanic crust (Plank and Manning, 2019). We find that the subducting Louisville Seamount Trail is a more important carbonate supplier.

3.2 The subduction of the Louisville Seamount Trail

The remnant of the Louisville Seamount Trail is ca.75 km wide and 4 500 km long, and located exclusively to the south of the Osbourn Trough. It is a typical plume tail formed by the eruption of the Louisville mantle plume. The Osbourn Seamount, which is now located close to the trench, is the oldest seamount that has not been subducted yet in this trail. It erupted right next to the fossil spreading center of the Osbourn Trough (Fig. 4), at ~76.7 Ma (Koppers et al., 2004).

The Osbourn Trough is a fossil ridge initiated after the eruption of the Ontong Java Plateau at ~119– 125 Ma (Tejada et al., 2002; Taylor, 2006), and was abandoned at ~86 Ma (Worthington et al., 2006), when the ridge jumped further south (Seton et al., 2012; Sun et al., 2022). This indicates that seamounts of the Louisville Seamount Trail older than 77 Ma erupted either on or to the north of the Osbourn Trough. Considering that the Osbourn Trough was spreading before 86 Ma, all the Louisville seamounts older than 86 Ma were once carried further north from the Osbourn Trough.

Plate reconstruction shows that the Louisville Seamount Trail started to subduct from the northeast corner of the Tonga Arc, and migrated southward (Ruellan et al., 2003). The HTHH volcano is located right above the subduction pathway of the Louisville Seamount Trail ca. 2 million years ago (Fig. 4). The Louisville Seamount Trail is pointing towards 336°, i.e., the angle between it and the Tonga subduction zone is about 36° (Ruellan et al., 2003).

The north end of the Tonga subduction zone is the fastest convergent plate boundary in the world with a subduction rate of 240 km/Ma (Benz et al., 2011) at the northeast end and declines southwestward. This is likely due to the collision of the Louisville Seamount Trail against the Tonga Arc and the opening of the Tonga back-arc basin. Consistently, the Tonga Arc is clearly curved westward at ~23.5°S, which likely marks the location of the latest collision/subduction. The northwestward drifting rate of the Pacific Plate is ca. 70 km/Ma, whereas the opening of the Lau Basin in the middle of the Tonga Arc near the volcano is ca. 60 km/Ma (Benz et al., 2011), i.e., the subduction rate near the HTHH volcano is ca.130 km/Ma.

Distance between the Tonga Trench and the HTHH volcano, is about 190 km. Seismic results show that the depth of subducting slab beneath the HTHH volcano is about 80 km. Hence, at a subduction rate of 130 km/Ma, it takes ca. 2 million years for the subducting seamounts to reach the location underneath the HTHH volcano. As illustrated in Fig. 5, there must be subducted Louisville seamounts near the mantle source of the HTHH volcano.

Compared with altered oceanic crust with water depths deeper than the CCD, the Louisville Seamount Trail is more abundant in carbonate. It consists of many individual seamounts of 30–75 km in diameter (Koppers et al., 2011) and 2 865–4 480 m in elevation from seafloor. The water depths on the tops of these seamounts range between 250 and 1 920 m, which are much shallower than the CCD, such a carbonate is stable on these seamounts, which is supported by IODP Expedition 330 drill cores (Koppers et al., 2011).

The Louisville mantle plume is located at ca. 50°S. In general, limestone reef is not popular at such a high latitude. However, one drill hole recovered ~15-m thick algal limestone reef, whereas the other three drill holes also recovered several condensed pelagic limestone intervals of up to 30 cm thick and multiple layers of foraminiferal sand interlayered with lava and carbonate-cemented volcanic breccia (Koppers et al., 2011). Seamounts drilled during IODP Expedition 330 all erupted before the warm period in the early Cenozoic, when the global temperature was up to 5–15 ℃ higher than that in the late 20^th century (Burke et al., 2018), such limestone reefs may form at high latitudes during this period. Note that, nummulitic limestone samples have been dredged from Burton Guyot in the Louisville Seamount Trail, which indicates the presence of the Eocene shallow-water reef in the high- to mid-latitude Pacific (Koppers et al., 2011).

The Louisville guyots were originally islands high above sea level (Koppers et al., 2011). All the drill holes of IODP Expedition 330 were on the flat top of different guyots. The drill-hole samples show eruption at shallow water depths or even above water, with erosional unconformities (Koppers et al., 2011). The diameter of the flat top of these guyots ranges 5–15 km (Buchs et al., 2018), suggesting that islands were originally thousands of meters above sea level, which took a long time for deplanation. These guyots were possibly formed even after the early Cenozoic warm period. Therefore, drill holes on flat top of these guyots are not representative in term of coral reefs. The hillsides may have much thicker limestone reefs.

Significantly, carbonates on the seamounts of the Louisville Seamount Trail are well protected by the interlayered lava. The subduction of these seamounts may carry a large amount of carbonate into the mantle wedge (Plank and Manning, 2019).

Carbonate may lower the solidus temperature of mantle peridotite by up to ca. 300 ℃ (Dasgupta, 2013). Therefore, the subduction of the seamounts is favorable for the formation of large volcanos. More importantly, it provides CO₂ that charged the magma chamber and powered the explosive eruptions.

The distribution of earthquakes shows that the subducting Pacific Plate was torn apart underneath near this region, forming slab windows (Chen and Brudzinski, 2001). In addition to the slab window roughly perpendicular to the trench, the slab underneath the HTHH volcano is also torn (Fig. 5), forming a slab window roughly parallel to the trench, which is likely responsible for a new volcano line behind the Tonga Arc, the Tofua Arc. The HTHH is one of them.

These slab windows expose the cold mantle wedge to the hot asthenosphere mantle, resulting in abnormally high temperatures. The subduction of shallow seamounts and slab windows together, endorse the formation of volatile-rich large volcanos.

3.3 The Mt. Pinatubo eruption

The Mt. Pinatubo is the largest volcano in the last 100 years. It erupted in 1991 on the Luzon Island, the Philippines (120.35°E, 15.13°N; Fig. 1c). The Mt. Pinatubo volcano is scaled VEI 6, with a total eruption volume of 8.4–10.4 km³, including 5–6 and 3.4–4.4 km³ of ignimbrite and fallout deposits, respectively (Self et al., 1996). The total CO₂ erupted in the Pinatubo is estimated 50 Mt (Gerlach, 2011). An amount of 15–20 Mt of SO₂ was ejected up to 40 km high, into the stratosphere (Self et al., 1996). It has periodic eruptions, which is the same as the HTHH volcano. The last major eruption occurred in A. D. 1500 with roughly the same size as those of 1991 (Newhall et al., 1996).

Similar to the HTHH volcano, the Mt. Pinatubo volcano is also associated with seamount subduction and a slab window nearby. It is located at the convergent margin in the west Philippines, where the fossil ridge of the South China Sea subducts underneath the Luzon Island (Zhan et al., 2015). Distribution of adakite and seismic imagines show that the subducting ridge is teared, forming a slab window (Yang et al., 1996). The Mt. Pinatubo volcano is located near the south edge of the slab window.

The fossil ridge of the South China Sea consists of large seamounts, with thick coral reefs. Among them, the Huangyan Island is the largest one. The subduction of this fossil ridge has been carrying down a large amount of carbonate since ca.15.5 Ma ago (Briais et al., 1993). The oldest eruption identified in the Mt. Pinatubo is 1.1 Ma (Newhall et al., 1996), likely after the opening of the slab window.

The main difference between the Mt. Pinatubo and the HTHH volcanos is that the former is close to the continent with a large amount of sediments subducting alongside a younger ridge. Therefore, far more sulfur was released from Mt. Pinatubo volcano, which had a major influence on the climate (Bândă et al., 2015).

3.4 The Novarupta volcano

The Novarupta volcano (Fig. 1d) is the largest volcano in the 20^th century (VEI 6). It erupted in 1912 with a total output of ~28 km³, including at least 17 km³ of fall deposits and about 11 km³ of ash-flow tuff (Fierstein and Hildreth, 1992). It is volatile-rich and had a major influence on global climate (Fierstein and Hildreth, 1992; Oman et al., 2005).

The Novarupta volcano is located at the end of the Alaska Peninsula, whereas the Pacific Plate is subducting northwestward along the Aleutian-Alaska trench. Seismic image shows that a seamount is subducting towards the Novarupta volcano (Frederik et al., 2020). The average water depth of the Pacific Plate to the southeast of the Aleutian-Alaska trench is shallower than the CCD (Amante and Eakins, 2009) (Fig. 1d). The seafloor south of the trench has been drilled in four locations, Site 178–180 and 183, off the Eastern Aleutians/lower Alaska Peninsula (The Shipboard Scientific Party et al., 1973a, b, c, d). None of the sites are on the seamount. Site 183 has nanofossil chalk about 10 m thick and several meters of carbonate rock, as well as a small amount of carbonate cement (The Shipboard Scientific Party et al., 1973d). Carbonate (chalk) and less abundant carbonate cements appear in the Site 178 for 1 m (The Shipboard Scientific Party, 1973c). Calcareous nanofossils are enrichment in the Site 179 at depth of 0–42 m (The Shipboard Scientific Party, 1973b). Calcareous nanofossils are less abundant but widely distributed in the Site 180 (The Shipboard Scientific Party, 1973c). There are thick sediments on the ocean floor and seamounts. The subducting sediments of the Aleutian-Alaska trench are more than 350 m thick, comprised of carbonates of up to 40 m thick, interlayered with pelagic clay, ashy-siliceous clay, and turbidites (The Shipboard Scientific Party, 1988; Plank, 2014). Seismic data also show that there are slab windows nearby (Gou et al., 2019) (Fig. 3). Once again, both volatile-rich sediments and slab window are favorable for the formation of such a large volcano.

4 CONCLUSION

The water depths of the west Pacific Plate are usually deeper than the CCD, and thus carbonate content is low on ocean floors in general. Seamount subduction is the most favorable process that forms violent large volcanos. Subducted seamounts with water depths shallower than the CCD provide carbonates that could enhance the melting capacity of the mantle wedge and release carbon dioxides to drive explosive eruptions. Slab windows nearby are another key factor, which could elevate temperatures and increase the melting capacity of the mantle wedge.

In the case of Tonga, subducted carbonate has very low sulfur contents, much lower compared to sediments of continental origin. Meanwhile, cyclical large eruptions have purged sulfur out of the source, resulting in the cleanest large volcanic eruption ever reported. For large volcanos associated with the subduction of continental sediments, e.g., the Mt. Pinatubo and the Novarupta volcanos, sulfur contents are usually much higher and more catastrophic.

5 DATA AVAILABILITY STATEMENT

The sources of all data are available in the manuscript or the public database.

6 ACKNOWLEDGMENT

We thank R. J. ARCULUS from the Australian National University, whose encouragement and suggestions greatly improved the early versions of the manuscript.