Palaeomagnetic results from an archaeological site near Rome (Italy): new insights for tectonic rotation during the last 0.5 Myr

Approximately 20 km north-east of Rome, along the modern trace of the Tiburtina road, recent archaeological diggings have brought to light a system of aqueduct galleries constructed by Roman engineers. This site falls inside the Acque Albule Basin, a travertine plateau Upper Pleistocene in age, that has been interpreted as a rhombshaped pull-apart basin created by strike-slip faulting within a N-S shear zone. This study provides evidence that two narrow water channels of this aqueduct system were significantly deformed by tectonic movement that occurred subsequent to their construction (II-III century A.D.). The geometry of the deformation pattern is compatible with that expected for a shear zone bounded by N-S oriented, right-lateral faults. The palaeomagnetic study of the volcanic formation («Pozzolane Rosse» Formation, 457 ± 4 kyr) containing the Roman aqueduct system evidences significant clockwise rotation around sub-vertical axis, consistent with the above-mentioned tectonic style. Mailing address: Dr. Fabio Florindo, Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143, Roma, Italy; e-mail: florindo@ingv.it


Introduction
Along the modern trace of the Tiburtina road, approximately 20 km north-east of the city of Rome (fig.1), recent archaeological diggings have brought to light a system of aqueduct galleries constructed by Roman engineers during the II-III century A.D. Geological and structural data collected and presented by Marra et al. (2004) along two narrow water channels of this aqueduct system (fig.2a-d) evi-denced deformations due to tectonic movement that occurred subsequent to their construction.
The archaeological site falls at the western edge of the Acque Albule Basin (AAB in fig.1), a travertine plateau Upper Pleistocene in age, with a medium thickness of approximately 60 m.AAB has been interpreted as a rhombshaped pull-apart basin (7 km long, 4 km wide) created by strike-slip faulting within a N-S shear zone, whose evolution is attributed to Middle-Upper Pleistocene times (Faccenna et al., 1994).The water channels were excavated in the pyroclastic terrain («Pozzolane Rosse» Formation, 457 ± 4 kyr, Karner et al., 2001) emplaced by the eruptive activity of the nearby Colli Albani volcanic district southwest of the basin (fig. 1) (De Rita et al., 1988, 1995;Karner et al., 2001).
The principal N-S water channel is affected by both brittle (extensive) and ductile (compressive) deformations, whereas the shorter difference in elevation of about 1.8 m, N-side down (portable GPS measurement) between its edges (see fig. 6 in Marra et al., 2004).The average gradient of 1.5% along the total length is comparable with that of the topography.In the northern part of the channel a conspicuous accumulation of sedimentary soil gives testimony to the occurrence of one or more instances of subsidence following the time of construction.However, subsidence levels in the range of one metre cannot alone explain the distortion manifested through horizontal displacement, resulting in an estimated total offset at the southern end of the principal aqueduct, of approximately 8 m and a consequent clockwise rotation of approximately 10°.A clockwise rotation in the same range also affects the minor aqueduct channel.
A detailed survey (Marra et al., 2004) of the principal channel indicates a segmented course of the entire structure, with orientations ranging between N10°E and N10°W, and with one section oriented at N35°W.Numerous extensional joints and faults were surveyed all along the archaeological excavations of the principal channel.Elaboration by way of a circular diagram reveals a preferential orientation on N30°-40°E and a dip ranging between 70°a nd 90°towards the SE (in most instances) rather than towards the NW (see fig. 7 in Marra et al., 2004).

Palaeomagnetic investigation
In order to detect possible horizontal rotations linked to strike-slip tectonics, we drilled 35 cores in 3 sites along a limited outcrop of lithified, fine grained portion of the Pozzolane Rosse pyroclastic-flow deposit (fig.4).This is a massive, poorly cemented, scoriaceous ash deposit erupted 457±4 kyr (Karner et al., 2001) by the Alban Hills volcanic district, whose petrographic features are well distinguishable with respect to other pyroclastic-flow units of the same volcanic district (De Rita et al., 1988;Karner et al., 2001).
We used a petrol-powered portable drill, and the cores were oriented in situ using a magnetic compass.Sampling was adjusted to avoid intervals with pebbles and scoriae that characterize this formation.The magnetic measurements were made in the shielded room of the palaeomagnetic laboratory of the Istituto Nazionale di Geofisica e Vulcanologia, INGV (Rome).All the samples were measured on a 4.5-cm small access pass-through 2G cryogenic magnetometer.Alternating Field (AF) demagnetization was systematically used.Peak fields were set at 10 milliTesla (mT) increments to 40 mT, then at 20 mT increments up to 120 mT.The maximum peak field was set at 150 mT.
For 77% of the samples, stable palaeomagnetic behaviour was evident in the vector component diagrams (fig.5) with characteristic remanence (ChRM) directions that generally tended toward the origin.Efficiency of AF  a best-fit line that was constrained through the origin of the vector component diagram (Kirschvink, 1980).For the remaining samples the demagnetisation data are noisy and it was cleaning (fig.5) and thermomagnetic analyses (fig.6) concur to indicate that magnetite is the dominant ferromagnetic mineral.For each samples, the ChRM direction was determined using not possible to determine a clear ChRM direction.This behaviour is attributable to the presence of lava clasts and scoriaes within the matrix.Site-mean palaeomagnetic directions were calculated using Fisher's (1953) statistics where only stable ChRM were isolated.
The mean palaeomagnetic directions obtained from the three sites are plotted on an equal area projection in fig. 7.All samples from the three sites show normal palaeomagnetic directions.The mean direction obtained from the statistical analysis of sites 1 and 2, close to the distorted portion of the aqueduct, indicates a clockwise rotation of about 50°with respect to the local north (N =15, Dm= 52.1°, Im= 43.8°, κ =19.3 and α95 = 8.6) with an average rotation rate of ∼ 0.1°/kyr.In addition, the data from sites 1 and 2 show a significant flattening of the palaeomagnetic vectors, with inclination data that are noticeably lower than those expected for a time-averaged GAD field.On the contrary, the mean direction obtained from site 3 is oriented N-S (N=10, Dm=347.5°,Im=61.8°,κ =21.5 and α 95=10.1)and the flattening is absent or negligible (fig.7).At this site, we cannot exclude the possibility that the N-S direction may arise from a very recent phase of remagnetization.
Finally, we rule out that a significant tilting of the investigated volcanic layer may have caused the observed palaeomagnetic rotation, since no evidence for it is inferred from the textural features of the rock, or from the surrounding structural-geological setting.

Discussion and conclusions
Clockwise rotation around sub-vertical axis, as predicted in the theoretical model suggested in Marra et al. (2004), may explain the observed deformation on the archaeological structure as well as the geometry and kinematics of the surveyed tectonic elements.The large angle of rotation may be interpreted as a consequence of the small dimensions of the rotated block and of the sum of repeated instances of movement in the time span 457 ± 4 kyr-Present (fig.8).
Several theoretical models for block-rotation induced by strike-slip faults have been pro-Fig.8. Schematic tectonic model for the study area (after Marra et al., 2004).A clockwise rotation around subvertical axis, could explain the deformation as well as the geometry and kinematics of the tectonic elements.
posed in the literature (e.g., Sonder et al., 1994).Here we suggest that rotation of small (several tens of meters) rigid blocks can occur among overstepping, N-S oriented, right-lateral fault segments.These small blocks must have also a shallow detachment level, in correspondence of major lithostratigraphic discontinuities.Therefore, it is expected that block-rotation in the area of Rome can be observed only in very limited sectors, in the proximity of N-S right-lateral faults.
The activity of right-lateral N-S faults, within the framework of a larger NE-SW extensional tectonic regime (Montone et al., 1995) in the area of Rome, has been interpreted (Marra, 1999(Marra, , 2001) ) as a local kinematics induced by a disengagement zone at the eastern boundary of the Northern Apennines.Based on the absence of significant local earthquakes in the historical sources (CPTI, Working Group) and in the instrumental seismicity, we believe that the strike-slip faults in this area are characterized by creeping and are responsible for slow aseismic deformation.
In conclusion, brittle and ductile elements affecting the two water channels and the surrounding rocks are coherent with a local stressfield characterized by a tensor of maximum stress (σ 1) oriented NE-SW on the horizontal plane, and a tensor of minimum stress (σ 3) oriented at 90°to the former (fig.8).Thus, the geometry of the deformational pattern is consistent with strike-slip tectonics previously described in this area (Faccenna et al., 1994) and interpreted as responsible for the «pull-apart» origin of the Acque Albule Basin.Palaeomagnetic data reveal that a prevalently horizontal displacement, causing locally large CW rotation, occurred in agreement with that expected for a tectonic deformation induced by a set of parallel N-S right-lateral faults.
Whereas a generalised CW rotation is expected in the vicinity of right-lateral faults, a more detailed model can be proposed to explain the presence of rotation only at two out of three investigated sites.We suggest that conjugate fracture systems in between the two N-S faults that border the area subject to rotation (as theoretically foresighted in Jones and Tanner, 1995), can originate a polygonal, roughly circular area (fig.8), that can accommodate the rigid rotation.Differently, the zones outside of this area do not experience any significant rotation.The location of the palaeomagnetically investigated sites and the results are in good agreement with the proposed model.However, such a model requires further verification, with a larger number of tested locations, to be considered a possible deformation style associated to strike-slip faults in the area of Rome.

Fig. 6 .
Fig. 6.Temperature dependence of the low-field magnetic susceptibility (κ) for a selected sample from site 2.

Fig. 5 .
Fig. 5. Example of AF (Alternating Field) demagnetization behaviour on vector component diagram, in geographic coordinates.White symbols denote projection onto the vertical plane.Grey symbols denote projection onto the horizontal plane.Numbers adjacent to data points indicate peak AF values in milliTesla (mT).

Fig. 7 .
Fig. 7. Equal area projection (lower hemisphere) of ChRM directions for sites 1+2 and 3. Mean declination (Dm) and inclination (Im) with confidence limit (α 95), precision parameter (κ ) and number of samples (N) are listed.The red star indicates the expected direction for a Geocentric Axial Dipole field (GAD).