Vp and Vs models show high velocity anomalies between 60 and 100 km depth that highlight a first order continuity of the Adria lithosphere (Figure 3), despite the obvious different resolution of regional and teleseismic models. At shallow depth (60–65 km), the two models show low velocity anomalies in the Adriatic offshore, suggesting lateral heterogeneities in the uppermost lithospheric mantle. The difference at the same depth beneath the belt is probably related to the different resolution of the two models, with a partial vertical smearing of deeper high Vs anomalies in the teleseismic model. The integration of the two models agrees with a continuity in the continental subduction of the high velocity Adria lithosphere along the entire mountain range (Figure 3). In the peri-Tyrrhenian mantle, a broad volume of low velocity is present from 60 to 100 km depth, consistent with a hot mantle wedge in front of the Adria lithosphere.

The process of continental subduction is well documented in the northern Apennines (Figure 4, section AA’), with the high velocity Adria lithosphere that deepens westward. The flexure of the lithosphere is well consistent with the decrease of velocity observed beneath the belt, indicating the subducting crustal material. The velocity model gives deep structural constraints on the role of sub-lithospheric load to flex the subducting lithosphere, proposed from analysis and modeling of the foredeep (Royden et al., 1987). The edge of delamination at 20 km depth is outlined by a spot of seismicity that indicates the point where the stripping of shallow material of the Adria lithosphere occurs. The west-dipping stripe of earthquakes is distributed within the flexed lower crust down to 70–80 km depth. The location of the positive anomaly of the Adria mantle lithosphere is fully consistent with the crustal features, indicating that the west-dipping continental subduction of Adria proceeds down to 160 km at least. A wide volume of high Vp/Vs resolved by teleseismic tomography is present beneath the belt and agrees with an intense hydration of the wedge that supports the topography of the belt and the active extension.

The gross structure in southern Apennines is not that different. The Adria lithosphere is deflected and the connection with the deep high Vp anomaly in the mantle is consistent with that of the northern sector (Figure 4, section BB’). A broad low Vp anomaly is present beneath a wide portion of the belt, indicating that crustal rocks deepened up to the mantle during the evolution. Anyway, seismicity in the lower crust is poor or absent, as well as the high Vp/Vs in the wedge.

Although tectonic rates in central Mediterranean are minor than in the eastern sector, active tectonic processes generate abundant seismicity that concentrates along the Apennines. The across-belt extension of about 2–4 mm/yr spreads along the entire mountain chain (D’Agostino et al., 2014; Serpelloni et al., 2022), generating large normal faulting earthquakes, with a higher energy release in the southern portion. Compression is well defined by earthquakes with inverse fault mechanisms developing along the main arcs of northern Apennines (Chiarabba et al., 2015).

Moderate magnitude events of the past decades align on contiguous fault segments of the extensional belt (Figure 2). Seismicity is rather shallow beneath the mountain range, while it deepens to depths of 20–30 km to the east of the chain (La Torre et al., 2023). This outer belt is interrupted at about N43°, isolating a 100 km long section without seismicity, but it resumes further south in the Apulian foreland and continues up to the limit with the Calabrian arc (Figure 2).

In the northern Apennines, seismicity concentrates within the upper 12–14 km of the crust beneath the mountain range and spread within a SW dipping layer down to 70–80 km depth (Figure 4A). At the connection between the shallow, extensional and the deeper seismicity, a cluster identifies the edge of the delamination of the Adria lithosphere. In the southern Apennines, seismicity occurs prevalently within clusters around more silent blocks (Figure 2) and down to 15–20 km depth, but deeper earthquakes like those in the northern arc are poor or absent (Figure 4B), except for a small sector at the edge of the southern Apennines (Figure 2).

Numerical modeling shows that a couple of paired compression and extension is generated by delamination beneath the external front and the inner belt, respectively (Gogus and Pysklywec, 2008). Seismo-thermo-mechanical numerical modeling shows that negative slab buoyancy broadly reproduces the coupled extensional and compressional tectonic regimes of the Northern Apennines, in terms of both stress orientations and velocities (D’Acquisto et al., 2020).

Velocity structure and seismicity indicate that the two processes are active in the Apennines belt in agreement with the evolution of decoupled collisions (Faccenda et al., 2009). First, the paired system of extension and compression expected by delamination is observed in the Apennines, at least down to 43°N of latitude. The evolution of such processes over the past 17 Myr generated a thinned lithosphere in the northern Tyrrhenian back arc, intense stretching, elevated heat flow, and melting of the crust (Rossetti et al., 1999; Della Vedova et al., 2001; Peccerillo, 2005; Lustrino et al., 2011). All these features are coherent with predictions from numerical modeling of delamination. On the same side, the intermediate depth seismicity (35–70 km depth, Figure 4) is indirect evidence for the eclogitization of the Adria lower crust, and the branching of positive pulses in RF is coherent with a thick layer of delaminated lithosphere beneath the belt (Figure 5), whose active edge is prompted by seismicity at the kink of the lithosphere (Figure 4). Furthermore, the doubling of the Moho beneath the mountain range and the shallow Moho in the inner sector (Figure 5, see Bianchi et al., 2010) also agree with predictions from numerical models. Velocity images document that the eclogitization of lower crust material is active beneath the northern Apennines coupled with a strong hydration of the mantle wedge (broad high Vp/Vs anomaly in Figure 5) and a retreat of the delamination hinge. The progressive eclogitization of the underthrusting Adria led first to a weakening of the lower crust with decoupling and exhumation of the still buoyant crust (Labrousse et al., 2010) and then a densification of granulitic material that take place along seismic and ductile deformation and metamorphism (Austhreim, 2013). The process starts at about 35–40 km depth and is almost complete at 70–80 km depth, as traced by the seismicity in the lower crust and velocity changes at depth (Figure 4A). This depth interval is consistent with PT conditions for eclogitization of the lower crust (Labrousse et al., 2010). Conversely, deep seismicity is poor and eclogitization seems slowed in the southern Apennines, reasonably associated with a stalled retreat (Figure 4). High velocities down to 160 km indicate that no plate rupture has occurred.

Evidence for a fossil process of eclogitization comes from the western Alps. Tomographic models show a distinct pattern of the subducting continental lithosphere (Menichelli et al., 2023). We observe a change from low to high Vp and low Vp/Vs at about 50–60 km depth within the European lower crust, and a broad high Vp/Vs anomaly in the Adria hydrated wedge on top of the subducting lithosphere (Figure 6). Intermediate depth earthquakes are poor or absent within the lower crust, and we hypothesize that the process is well mature and probably no more active. Anyway, the fossil example of the western Alps suggests that eclogitization, taking place during the late evolution of subductions, contributes to the generation of compositional anomalies within the continental mantle, documented here by the high Vp/Vs at the mantle top of Adria.

The Apennines has been a repeated target of geophysical investigations for many decades (Amato et al., 1993; Spakman et al., 1993; Piromallo and Morelli, 2003; Di Stefano et al., 2009; Benoit et al., 2011; Giacomuzzi et al., 2011; Giacomuzzi et al., 2012; Giacomuzzi et al., 2022). A solid consensus exists for some gross and main features of the mantle structure, but there are still some differences that fuel the debate on contrasting models of the belt evolution. Many models reported the existence of high velocity anomalies in the uppermost mantle compatible with subducted lithosphere (Lucente et al., 1999; Piromallo and Morelli, 2003; Benoit et al., 2011; Giacomuzzi et al., 2011; Giacomuzzi et al., 2012). Models with west-dipping slab largely prevail, although the almost vertical geometry of the high velocity slab. In addition, the lack of lateral and vertical continuity of the high velocity anomalies is interpreted as possible evidence for a slab break off (Wortel and Spakman, 2000; Lucente and Margheriti, 2008) or slab window (Amato et al., 1993; Lucente et al., 1999), although the meaning of the latter remains rather ambiguous in terms of origin and evolution. More recently, Giacomuzzi et al. (2022) proposed that pre-existing compositional anomalies of the Adria continental lithosphere might explain the reduced Vp and low Vp/Vs alternatively to the slab window or slab interruptions.

Receiver function modeling yields the definition of the Adria lithosphere, which is westward plunging beneath the Apennines (Bianchi et al., 2010). Several studies focused on the modeling of the Moho geometry (Agostinetti et al., 2011; Di Stefano et al., 2011; Spada et al., 2013), and on the Lithosphere-Asthenosphere boundary depth (Miller and Agostinetti, 2012). The correlation between intermediate-depth seismicity and positive pulses in RF profiles across the chain suggests that part of the crust underthrusts the belt with the Adria lithosphere, with a flow at the nose of the wedge facilitated by slab dehydration (Agostinetti et al., 2011). The difference between the thickness of the seismic layer (∼30–35 km, LaTorre et al., 2023) and the depth of the moho (∼70 km) beneath the mountain belt is consistent with the hypothesis that seismicity and deformation of the roof plate are decoupled from the westward plunging delaminated lithosphere. Anyway, the reconstruction to greater depths and the break in the west-dipping Adria lithosphere seem to conflict with a westward-subduction scenario continuous from the Oligocene (Chiarabba et al., 2014).

The SKS anisotropy pattern suggests a differential evolution of the trench-retreat process along the Northern Apennines (Salimbeni et al., 2013; Pondrelli et al., 2022). The pattern is typical of retreat in the southern part of the Northern Apennines, with the no orogen-normal measurements found in the inner sector of the northern part in contrast to a significant pure retreat that has been interpreted as an oblique motion since around 5 Ma (Salimbeni et al., 2008). The different behavior is also evidenced by the lithosphere reconstruction that supports an irregular delamination/retreat process along the belt, with a maximum in the southern portion of northern Apennines (Chiarabba et al., 2014).

Our results show that the subduction of Adria followed the delamination of the lower crust and the underthrusting of the lithosphere beneath the orogen (Figure 4). The first order similarity in velocity structure across northern and southern Apennines indicates that the process is broadly similar for the entire Apennines belt. Anyway, difference in structural details, the discontinuous presence of deep seismicity and fluids within the wedge, and, finally, anomalies in the Adria mantle (Figures 3, 4) suggest that continental delamination is not uniform and is influenced by heterogeneities of the Adria lithosphere.

The lateral space/time heteropy of the belt might be partially controlled by difference in buoyancy of the Adria continental mantle. Such differences can be even more pronounced in continental margins, like the Adria microplate at the margin of the Neo-Tethys Ocean (Stampfli and Borel, 2002; Handy et al., 2010), where intense underplating of material occurs during the ridge formation.

The lateral varying buoyancy of the lithosphere played a key role in conditioning the subduction of the Adria lithosphere. This lateral variation in buoyancy resulted in changes in lithospheric stiffness, which in turn influenced the overall configuration of the Apennines belt. These observations match the predictions by numerical models (Lo Bue et al., 2021). Delamination is the primary process responsible for shaping the northern Apennines over the last 10 million years. However, in the southern Apennines, this process interacts with the retreat of the Ionian lithosphere beneath the southern Tyrrhenian subduction zone, and it is also influenced by the thermal and compositional characteristics of the underlying mantle. Tomographic images reveal the presence of a pronounced low-velocity (Vp) anomaly beneath southern Italy. This anomaly may arise from either an upwelling of the asthenosphere or the presence of a thermally and compositionally distinct mantle region. A compelling hypothesis to explain these observations is the varying distribution of fluids associated with the eclogitization process, which may account for the differences in the retreat rates of the two subduction arcs. The deep geometry that we have defined documents new evidence for modeling the prediction that buoyancy variations in the subducting lithosphere exert a first-order control on subduction rate, with more buoyant slab segments correlated with slower subduction rates and steeper slab dip (Royden and Husson, 2009).

In the central Mediterranean area, seismicity distribution and strain rate from geodetic data clearly show that the deformation, confined at plate boundaries along the Alpine and Dinarides collision zones, spreads for hundreds of kilometers within the continental lithosphere of Adria (Serpelloni et al., 2022). Focusing on the Apennines (Figure 7), deformation pervaded the roof plate that consists of a mixture of continental terranes accreted during the belt formation (Bally et al., 1986). The process is not that of a coupled collision like the Alps, but the continental underthrusting of Adria concurs to edify the Apennines Mountain belt with crustal material that is peeled off and accreted in east-verging stacks. The evolution of this decoupled collision system is characterized by the disruption of the accretionary wedge and the formation of an extensional basin in the inner part of the orogen (Faccenda et al., 2009). This thinned lithosphere block (i.e., the Tyrrhenian block) has an independent motion from that of Europe and Adria (Serpelloni et al., 2022). A 2–4 mm/yr of across-belt extension characterizes the mountain range (Figure 7), paired by compression along the external front in the eastern Adriatic area. The low deformation rate is responsible for devastating earthquakes that struck the area in recent and in historical times (Rovida et al., 2022). A close look at recent large earthquakes of the Apennines (Figure 7) reveals that they prevalently clustered in the extensional and compressional deformation spines along the mountain range and the eastern Adria front, while some strike the Apulian foreland area, where strain rate is rather low. While extension is defined for almost the entire belt, compression is well defined only down to 43°N, becoming smaller or absent southward, or transferred Eastward (i.e., along the southern Croatian and Montenegro coasts (Figure 1). This observation opens a debate on whether crustal delamination is still active or if it is in an immature stage in southern Apennines.

Although the low strain rate, the completeness of the historical catalog for M>6 in the past millennium (Rovida et al. 2011) ensures a proper imaging of the long-term seismic energy release of the belt. The extension rate is almost similar between the northern and southern Apennines (Figure 7), but the energy released by historical earthquakes is higher in the south (Figure 8). Higher magnitude could be partially explainable by the greater thickness of the seismogenic layer (see Chiarabba and De Gori, 2016; LaTorre et al., 2023) and the dimension of fault segments. This feature could be consistent with the hypothesis that the energy release is controlled by the thermal state of the Apennines crust. Alternatively, we are inclined to attribute this difference to a different state of delamination, progressing from a more mature state in the north to a less advanced state in the south. The ductile deformation of the intermediate crust is notably favored by the greater maturity of the delamination process in the northern region.

Oppositely to the higher seismic release to the south in historical times, l arge magnitude earthquakes of the past decades (Figure 6) are more clustered in the northern arc, with the impressive decade-long series of events that struck central Italy. Seismicity is sparser to the south, with M>3 earthquakes occurring within clusters that isolate blocks with low seismicity, suggesting that fault segments in this portion of the belt are locked during the interseismic period. Two regions of large strain rate and poor seismicity are present at the northern and southern tips of the segment repeatedly struck in central Italy. These two zones of the normal faulting system deserve effort and ad hoc monitoring. An ambiguous region is the Apulian foreland, where the energy released by past earthquakes is high, but the strain rate is low. This continental margin represents a transition zone between the ea stward-verging Apennines and west-verging Albanides-Hellenides. The flexural bending of the Apulia foreland is linked with E-W strike slip kinematics and extension (Doglioni et al., 1994; Argnani et al., 2001; Maesano et al., 2020). We hypothesize that the transpressive tectonics in the region can be attributed to one of two factors: either an interaction between the subduction of the Adria plate beneath the two mountain belts or an incipient delamination in the Gargano area. In the latter scenario, the stress reorientation induced by the delamination process in the Gargano area has not reached the level observed in the northern arc, where the delamination process is mature.

Several portions of the compressional belt lack consistent seismicity over the past decades, suggesting that the thrust system is either locked or only partially coupled. Earthquakes typical of the belt develop on shallow S- and SW- dipping segments forming broad arcs like for the 2012 Emilia seismic sequence (Govoni et al., 2014; Chiarabba et al., 2014). The Mw 5.7 earthquake of November 2022 ruptured a small segment of the system with the rupture of a deep portion of the thrust plane (Pezzo et al., 2023). Earthquakes like these suggest that the compressional system is highly segmented with a limited vertical and lateral continuity of faults at depth.