Discovery of Deccan Inclination Anomaly and its possible geodynamic implications over the Indian Plate

The rapid northward drift of the Indian plate during Deccan volcanism assumes a gradual shallowing of paleomagnetic inclinations in subsequent lava ow formations. A comparison of palaeomagnetic data produced during the last six decades reveals an inclination anomaly during Chron C29r (66.398–65.688 Ma) along with brief clockwise-counter-clockwise rotations during and after the main phase Deccan eruption. This interval temporally coincides with i) an accelerated Indian ocean spreading rates, ii) brief incursion of an inland ‘seaway’ and iii) a major drop in the sea level at the southern tip of the Indian Peninsula. Furthermore, the restoration of tilt later during C29n agrees with the withdrawal of the inland seaway and the development of a regional southward dip of the Deccan lava ow formations. Here, we produce an evolutionary model to postulate the interaction of the Réunion plume with the Indian lithospheric plate with coincident geological evidences demanding further exploration.


Introduction
Recent studies on mantle plume-lithosphere interactions indicated that spreading plume heads below the lithosphere can develop signi cant asthenospheric ows to exert the 'plume-push' force and act as potential drivers for accelerated plate motions and/or initiation of subductions (e.g., Pusok and Stegman 2020;Cande and Stegman 2011;van Hinsbergen et al. 2011). The Deccan large igneous province (LIP) is the product of lithospheric interactions of the Réunion hotspot over the northward drifting Indian plate during Late Cretaceous and early Paleogene times (Courtillot et al., 1986;Basu et al., 1993).
Geochronological records indicate a tholeiitic basalt peak during 66.4-65.4 Ma, i.e., precisely within the geomagnetic Chron C29r. This peak is widely referred to as the Main Deccan eruption phase (Sprain et al., 2019 and references therein), denoted here by the DE M . However, the style and repercussions of the impact of the Réunion mantle plume over the Indian lithospheric plate are inadequately explored (e.g.

Raval and Veeraswamy 2019).
Globally, Deccan traps represent one of the classical palaeomagnetic records with extensive databases that were produced during the last six decades (e.g., Clegg et al., 1955, Vandamme et al., 1991Wensink 1973;Chenet et al., 2008;. We considered over 1600 statistically signi cant mean directions obtained from all over the Deccan trap ows and dikes (Fig. 1).
The high ferrimagnetic concentrations in the Deccan basalt mineralogy enabled classical approaches of successful demagnetization to obtain the characteristic remanence (ChRM) directions referred to as primary magnetization. Here, we compiled the data from 56 widely referred publications representing the entire Deccan province, although they were largely dominated by Central Province and Chron 29r. After compilation, we classi ed the data into geomagnetic Chrons C30n, C29r and C29n (methods and treatment of data described below and the compiled data are available in Supplemental le). As mentioned, the Chron 29r (66.398 Ma -65.688 Ma) represents the highest number of data points in agreement with the prevailing knowledge that over 80% of the lavas erupted during this D EM Chron (e.g. Renne et al., 2015;Schoene et al., 2015;Sprain et al., 2019). The compilation observed an unambiguous inclination anomaly of more than 10º and clockwise + anticlockwise rotations of 2 to 5 degrees during C29r (see Tables 1 and 2 and the data treatment in sections below). Inclination Anomaly ----12 -----

Methodology
The inferences and conclusions drawn in this manuscript are based on the palaeomagnetic database developed from published literature to date (from 1955 to 2020). Over 65 publications presenting palaeomagnetic approaches were published during this time, out of which over 50 publications were widely and repeatedly referred independently or as cross references. The vast majority of these papers unambiguously reported directions in agreement with the N-R-N sequence of the C30n-29r-29n geomagnetic polarity time scale. These data were produced globally by different teams, and the analysis was performed in many different reputed laboratories with varied sets of instrumental con gurations and sensitivities. We elaborate here on the criteria and methods adopted to compile and treat the data. We also describe the sources of error and the rationale of ltering the data for mean calculations.
The published papers generally presented demagnetization data (using thermal or alternating elds) and the estimation of characteristic remanent magnetization (ChRM) as primary remanence. The routine statistical methods of spherical distribution used in paleomagnetism allowed the means to be estimated at the specimen level and then at the sample level and site levels. It enabled standard parameters (e.g., Alpha-95, precision parameter and maximum angular deviations) to describe their scatter, facilitating global comparison. It further gives an idea about the quality of data in addition to describing the normal/reverse polarities and calculating the apparent and true poles. Routine statistical methods are based on the classical approach of Fisher statistics (Fisher, 1967) for spherical distribution of the vector data. This criterion has been used universally for rejection of the data and depiction of its quality.
In the majority of the papers, the reversal is unambiguously assigned to C29r, the reversal followed by normal to C29r-29n, and the normal followed by reversal to 30n-29r polarity chrons. The data are very well supported by eld stratigraphic knowledge or chemostratigraphy. For the present analysis, we complied only with the declination/inclination (D/I) directions from the published literature, facilitating the site mean data points (given in the supplementary le). This is because the NRM intensities are found to have large deviations due to style of presentation and laboratory standards and instrumental sensitivities from individual attempts. Palaeomagnetic analysis involves various protocols of demagnetization adopted by different workers and instrumental sensitivities. Therefore, the standardization and comparison of NRM intensities across different attempts is not feasible. However, since our inferences are founded entirely on the D/I data, the NRM intensities are not considered further.

Possible Sources of Error
The data were retrieved and rechecked several times to check the typo errors. Below, we discuss the sources of errors based on which the ltering strategy was adopted.

1) Manual error
The rst and foremost source of error is generally developed during the collection of oriented samples in the eld. The oriented samples are collected manually (oriented hand samples) or by gasoline-driven portable rock coring machines (manually handled). The samples were marked carefully using either the north compass or the sun compass method. This has a greater chance of introducing manual errors at various stages from marking in the eld to creating cylindrical specimens in the laboratory. Manual errors can also be introduced during laboratory handling of specimens. For most spinner magnetometers, the samples are to be handled over six directions of measurements at every stage. There is no clue to de ne the manual error, although it may be represented in the nal data as scatter that can be de ned by the standard palaeomagnetic data presentation procedures but with unknown contribution.
2) Laboratory standards: The palaeomagnetic data in Deccan traps are produced from various laboratories that are commonly equipped with spinner magnetometers. The high NRM intensities often permit complete demagnetization, even with routine spinner magnetometers with low sensitivities (e.g., Minispin from Molspin UK, Sensitivity: 0.05 mA/m). The other common spinner magnetometers of better sensitivities used are the DSM-Schonstedt (~ 10e-4 A/m) and the JR-4 to 6 series of AGICO Czech (~ 2.4 x µA/m). Both of these instruments thus provide better con dence over a large number of palaeomagnetic data, although the quality of data carefully produced from other instruments, such as Astatic magnetometers, is ascertained considering the excellent repeatability and the higher intensities of the Deccan basalt samples. Furthermore, the fully automated AGICO instruments prevent manual errors of sample positioning, and the standardized data interface software, statistics and plotting interface allows rapid, error-free processing. The cryogenic magnetometer (e.g., 2G) gives the nest sensitivity in paleomagnetic analysis; however, the strong remanence in Deccan basalt does not demand such analysis unless paleointensity and secular variation such as studies are aimed.
The detailed palaeomagnetic analysis involves demagnetization of a large array of specimens to produce statistically signi cant data by the removal of noisy results. The two most common methods of demagnetization used are thermal and alternating eld demagnetizations. While thermal demagnetization can introduce laboratory-induced errors during heating and cooling, af demagnetization is most successful in Deccan traps due to its soft ferrimagnetic mineralogy for both primary and secondary components. Individual workers have used different protocols of demagnetization strategy, and the demagnetizers themselves can introduce spurious elds during analysis, producing deviations rather than direct errors. Furthermore, the skills and experience of individual workers during interpretation varies, which may lead to some manual bias error component. b) Geotectonics: The shield type geometry of the Deccan province in general refutes any major intrashield tectonics to affect the palaeomagnetic directions. However, the lineaments and other structural features within Deccan Province, if contemporary, can be inferred for tectonically induced errors. Few authors have reported such tectonic relations, but they are mainly related to vertical movement rather than internal rotations and deformation and do not express any major anomaly in paleomagnetic data.
These references justifying the tectonic component are avoided in our database approach. Chron C29r is the main focus of this study, and the majority of the palaeomagnetic data for this chron belong to the main/central Deccan province, which does not show such intra-shield tectonics at large to affect the internal rotations and tilt. If such incoherence is present, it should be re ected by deviation of D/I directions internally, for which we have applied the ltering criteria discussed below.

Data Reduction (rejection) and Filtering
The Deccan traps represent one of the richest databases for a short geological interval of less than 5 Ma, marked by the distinct polarity zone of N-R-N of the Late Cretaceous/Paleogene. The ample data produced globally from different laboratories are within close agreement, and a simple ltering and reduction of data based on routine spherical distribution statistics is feasible.
A previous compilation made by Vandamme et al. (1991) resulted in de ning the Deccan Super pole based on contemporarily available data. With the updated database up to 2020, we recalculated the Deccan Super pole, which is in close agreement with Vandamme et al 1991 (see Table 1). The deviation of values seen in this table is simply due to enrichment by the new data during the latter 30 years since the publication of Vandamme et al 1991. Therefore, considering these Super pole directions as central tendencies, we de ned the limits/windows for ltering out the data, apart from rejecting the data with large scatter de ned by the precision parameter (k) and alpha-95 of Fisher statistics. Table 2 Considering the means for whole data in 5th column of Table 1 as the central tendency of the updated database, we further applied lters to remove the noise in data due to the possible errors described above. The data for C30n, C29r and C29n are ltered individually in a declination window of +/-36 (10% of 360) and inclination window of +/-18 (10% of 180).  During 80 to 60 Ma -12 The Inclination Anomaly and Rotation The observed mean inclination for C29r is signi cantly higher than the anticipated paleolatitude derivative of 35º, and an average value of 10.78 can be assigned from various approaches expressed in Table 3. This '+10º' inclination anomaly observed during C29r is simply a mathematical expression of a signi cant northerly dip of the Indian plate. It is much oversighted in the context of equatorward drift of the Indian plate, which anticipates either inclination shallowing or, at most, inclination values intermediate to C30n and C29n. Moreover, no record of such large magnitude changes during the Late Cretaceous geodynamo does exist (Coe et al., 2000;Pechersky et al., 2010;Velasco-Villareal et al., 2011) and therefore also refutes the geodynamo effect. In contrast, coincident geological evidence from the Indian subcontinent endorses the anomaly by possible effects of plate tilting (Fig. 2 and the evidence produced below).
Very high inclinations (D/I=140/60º) during C29r are reported from the Deccan trap rocks of the Cauvery region in southern India (Mishra et al. 1989), and although the data are inadequate, they suggest a southern extent. The pre-Deccan Late Cretaceous strata from the Cauvery Basin on the southern Peninsula record a shallower inclination (338/-38, N=80) (Venkateswarulu 2020), substantiating the existence of the Deccan anomaly. The inclinations for C29n and C30n agree well with the anticipated paleolatitudes (Table 3), which indicates that the tilt was absent in C30n and restored during C29r as the Indian plate drifted away from the Réunion plume head (e.g., see Fig. 2). The regional southward dip for the Deccan lava ows (Fig. 3) is widely documented (Jay and Widdowson 2008;Shoene et al. 2015) and verify our proposed model based on palaeomagnetic inclination (Figure 3).
Considering the tilt and rotation estimates from the palaeomagnetic database (Tables 1 to 4), we further con rm our model by supporting geological evidence accounted below.

Magmatic records of plume head arrival and plate tilt
There is a close temporal and spatial linkage between voluminous LIPs, their tholeiitic and alkaline magmatism and mantle plumes (e.g., Bryan and Ernst 2007). The LIPs are generally characterized by short-lived (<1-5 Ma) igneous pulses responsible for large volume (>75%) magma outpours. Alkaline rocks associated with LIPs are typically formed due to low degrees of partial melting of mantle owing to minor thermal effects from an impinging or receding mantle plume (e.g., Gibson et al. 2006). The impact of the Réunion plume over the Indian plate is in close agreement with this convention through the observed episodes of tholeiitic and alkaline magmatism. The initial impact of the Réunion plume head started ~0.2 million years before the DE M and produced nepheline syenites and alkali gabbros during these early Deccan eruptions (Fig. 3), corresponding to terminal C30n/early C29r. Recent high-precision geochronological data (Renne et al., 2015;Sprain et al., 2019) indicated outpouring of bulk Deccan tholeiites between 65.4 and 66.4 Ma within C29r. This rapid extrusion requires higher amounts of partial melting under a considerably elevated geothermal gradient during a fully developed plume head, which precisely coincides with the duration of C29r (Cande and Kent, 1995). Small-volume volatile-rich magmatism of lamprophyres and carbonatites between 65.8-65.2 Ma (Fig. 3), which was mostly emplaced towards the terminal part of the DE M, typically intruded the Deccan lavas. This terminal phase is an artifact of small-fraction melting caused by thermal weakening during the waning stage of the Réunion plume. The occurrence of DE M precisely within C29r therefore elucidates the short span (<1 million years) of geodynamic interaction of the plume head with the Indian plate. This rapid impact of the plume head below the western margin of the Indian plate therefore appears to have resulted in tilt and rotation, as recorded by the paleomagnetic data.
Accelerated convergence at the end of C29r The spreading rates in the Indian Ocean reached a maximum between ~66 and 63 Ma during C29n (Cande and Stegman, 2011; compiled and redrawn in Fig. 4). The initial anomalously high rates of drift of the Indian plate from less than 100 mm/y to ~160 mm/y during 68 to ~66 Ma are explained by the arrival of the plume head (Eagles and Hoang, 2014). Therefore, the later increase in spreading (up to 180 mm/y after ~66 Ma) during the waning and withdrawal stages of the plume/plume head needs to be explained.
We postulate this later increase in convergence rates as a result of i) the termination of plume-induced rotational and tilt components that resolved into northward drifting kinematics, in addition to ii) the establishment of double subduction.

Acceleration of the intra-oceanic subduction
Mantle plumes have been considered drivers of regional subduction initiation (Gerya et al. 2015;Pusok and Stegman 2020;van Hinsbergen et al. 2021, Rodriguez et al. 2021. Multiple subductions are evident for the India-Asia convergence; however, the nal subduction during ~66 Ma to 65 Ma is little explored in the context of Deccan volcanism and the Réunion plume push force. We infer that the quick geodynamic response of the Indian plate over the Réunion plume head during DE M marked by the tilt and counter rotations during C29r might have exerted signi cant changes in pre-existing plate kinematics at the India-Asia subduction interface. Possible resultant deformation due to plate tilt and rotation added to the previously accelerated rate of convergence may have signi cant repercussions on the initial stage of intraoceanic subduction (Fig. 2). The combination of quick clockwise and anticlockwise rotations along with tilt and drift appears to have superimposed over the pre-existing kinematics for the Indian plate demanding detailed kinematic modelling in this context.
Opening of the 'Sea-way' within the plate Based on paleontological nds, a short-lived 'seaway' associated with Deccan traps has been reported precisely at the end of C29r (e.g. Keller et al. 2009Keller et al. , 2012. This inland 'sea-way' formation (/marine in uence) along pre-existing rift valleys (i.e. Narmada and Godavari Rifts, shown in Fig. 3) is evident by the stressed marine fauna. The brief north/northeast tilting of the Indian plate therefore offers a possibility to explain the formation of the brief inland 'sea-way'.
Biostratigraphically well-documented localities ~800-1000 km inland of the Narmada and Godavari rifts contain brief and stressed planktic foraminiferal assemblages within terrestrial palustrine to freshwater facies (Keller et al 2009(Keller et al , 2012. The absence of benthic species among these localities (Keller et al 2012) indicates only a brief marine incursion that can be explained by a major tectonic event such as the lithospheric tilt reported here. The paleosols developed over this zone designate upland conditions and further support the restoration of tilt during C29n, as depicted in Figure  Finally, although more geological evidence with precise dating is required, the present perceptions ( Fig. 2-4 and Table 4) strongly support the geodynamic developments over the Indian plate precisely during C29r and the main Deccan eruption.  In the northern region of the central province, the majority of the normal polarity ows overlay the reverse polarity, and the rest of the occurrences with prominent reverse polarity of C29r compile the tripartite subdivision of the Deccan traps into the C30n-C29r-C29n sequence. Stratigraphically thicker (/longer) records of C29r are most widely documented in the Central province. Site Abbreviations-Amboli (Ab) , Shahapur (Sh), Singarchori(Si), Sarnu (Sr), Tapola (Ta), Trimbak (Tr), Umred (Um), Varandha Ghat (VG), Vikarabad (Vi), Wai (Wai).

Figure 2
Evolutionary staged model to depict the mechanism of geodynamic interaction of Indian plate with the Réunion plume/hotspot imparting the inclination anomaly. During C30n, the subcontinent approached the impinging mantle plume, as documented by the alkaline magmatism in the northern part of the Deccan province. The interaction of the Indian plate with a fully developed plume head further during C29r resulted in a north/northeast tilt to record the ambient reverse eld. As the plate moved farther from the waning plume, the tilt was restored, and the reverse inclination steepened, producing the inclination anomaly of C29r (detailed in text).  Relative probability of alkaline and small-volume, volatile rich magmatism spatially and temporally related to the Deccan LIP based on high-precision 40Ar/39Ar and U-Pb determinations (n=12) (adopted from Dongre, et al. 2021). The period of bulk Deccan basalt volcanism is also shown in green (from Sprain et al., 2019). Inset: Spreading rates between India-Antarctica and India-Africa ridges (After Cande and Stegman, 2011).