Periodicity in the Deccan volcanism modulated by plume perturbations at the mid-mantle transition zone

Abstract

In peninsular India, the Deccan Traps record massive, continental-scale volcanism in a sequence of magmatic events that mark the mass extinction at the Cretaceous-Paleogene boundary. Although the Deccan volcanism is linked with the Réunion hotspot, the origin of its periodic magmatic pulses is still debated. We develop a numerical model, replicating the geodynamic scenario of the African superplume underneath a moving Indian plate, to explore the mechanism of magmatic pulse generation during the Deccan volcanism. Our model finds a connection between the Réunion hotspot and the African large low shear-wave velocity province (LLSVP) to show pulse generation from a thermochemical plume in the lower mantle. The plume is perturbed at 660 km, and its head eventually detaches from the tail under the influence of Indian plate movement to produce four major pulses (periodicity: 5 - 8 Ma), each giving rise to multiple secondary magmatic pulses at a time interval of ~ 0.15-0.4 Ma.

Full text

This manuscript is a non-peer reviewed preprint submitted to EarthArXiv. A
1
copy of this manuscript has been submitted in Earth and Planetary Science
2
Letters (Elsevier). Currently in-revision.
3
4
Periodicity in the Deccan volcanism modulated by plume perturbations at
5
the mid-mantle transition zone
6
7
Dip Ghosh1, Joyjeet Sen2, and Nibir Mandal3
8
Department of Geological Sciences, Jadavpur University, Kolkata 700032, India
9
10
Abstract
11
In peninsular India, the Deccan Traps record massive, continental-scale volcanism in a
12
sequence of magmatic events that mark the mass extinction at the Cretaceous-Paleogene
13
boundary. Although the Deccan volcanism is linked with the Réunion hotspot, the origin of its
14
periodic magmatic pulses is still debated. We develop a numerical model, replicating the
15
geodynamic scenario of the African superplume underneath a moving Indian plate, to explore
16
the mechanism of magmatic pulse generation during the Deccan volcanism. Our model finds a
17
connection between the Réunion hotspot and the African large low shear-wave velocity
18
province (LLSVP) to show pulse generation from a thermochemical plume in the lower mantle.
19
The plume is perturbed at 660 km, and its head eventually detaches from the tail under the
20
influence of Indian plate movement to produce four major pulses (periodicity:  Ma),
21
each giving rise to multiple secondary magmatic pulses at a time interval of ~  Ma.
22
23
Keywords: Deccan Traps; Cretaceous-Paleogene extinction; Numerical simulation; African
24
LLSVP; Réunion hotspot; Mid-mantle transition.
25
26
For correspondence:
27
1dipghosh14@gmail.com ; 2senjoyjeet@gmail.com; 3nibir.mandal@jadavpuruniversity.in
28
29
30
31
32

1. Introduction
33
Deccan Traps (DTs), the most spatially extensive continental flood basalt (CFB)
34
province in peninsular India, witness a remarkable event of volcanism in the Phanerozoic
35
history of the Earth (Chenet et al., 2009), which in recent times has received particular attention
36
in connection with the mass extinction of biological species (Keller et al., 2012; Wilson, 2014).
37
A school of thought relates this sudden biotic crisis to the enormous volume (> 106 km3) of
38
basaltic magma eruptions in the Deccan provinces (Schoene et al., 2015; Wignall, 2001) during
39
late Mesozoic to early Cenozoic (Fig. 1a). This massive volcanism involved degassing on a
40
global scale, resulting in two significant environmental changes: the first being global
41
warming, carbon cycle disruption, and ocean acidification (Self et al., 2014) associated with
42
volatile emissions, with the second a poisoning of the entire ecosystem (Schmidt et al., 2016)
43
associated with SO2 injection into the upper atmosphere. Another school of thought has
44
proposed a Chicxulub bolide impact theory for the Cretaceous mass extinction (Alvarez et al.,
45
1980; Schulte et al., 2010), but the issue is still debated. The DTs have also stimulated
46
discussions on the long-standing critical question about the origins of large igneous provinces
47
(LIPs) (Campbell and Griffiths, 1990; Dannberg and Sobolev, 2015; Farnetani and Richards,
48
1994). What is the potential source of enormous magma supply to LIPs, and how are they
49
connected to lower mantle dynamics (Glišović and Forte, 2017; White and McKenzie, 1995)?
50
This Deccan volcanic province is excellent for studying LIPs as it is relatively young and
51
geographically extensive thus, allowing geoscientists to reliably reconstruct the eruption events
52
in space and time.
53
Based on volcanological and geochemical properties, the Deccan Volcanic Province
54
(DVP) is divided into three principal stratigraphic successions: Kalsubai, Lonavala, and Wai
55
subgroups (Fig. 1b). The volcanic event that defines the Cretaceous-Paleogene boundary
56
(KPB) at  Ma (Sprain et al., 2018) occurred  ka after the
57

emplacement of Kalsubai falls within Khandala, Bushe, or Poladpur Formations (Richards et
58
al., 2015). Using 40K/40Ar plagioclase geochronology of erupted basalts and U-Pb
59
geochronology of zircon from intervening ash beds, several workers have constrained the
60
timings of multiple eruptions pulses (Keller et al., 2012; Richards et al., 2015; Schoene et al.,
61
2019, 2015). All these studies agree that the main eruption phases started shortly before the
62
C30n-C29r geomagnetic reversal and ended following the C29r-C29n reversal. Above the
63
KPB, the Wai subgroup consists of geochemically and volcanologically distinct formations,
64
which suggest more voluminous eruptions (Renne et al., 2015; Richards et al., 2015; Sprain et
65
al., 2019).
66
This study aims to explore the mechanism of unsteady eruption dynamics in the
67
evolution of DVP through multiple pulses, punctuated by quiescent periods. Previous studies
68
based on geochemical data (Chenet et al., 2007) suggested three phases of DT eruptions, with
69
most of the volume, erupted before the KPB, where the second phase is considered responsible
70
for late Cretaceous environmental changes (Chenet et al., 2009; Petersen et al., 2016) (Fig.
71
1b,c). Alternative views emphasize the Chicxulub impact to propose that the DVP magma
72
eruptions were mostly a post-KPB event (Renne et al., 2015; Richards et al., 2015). More recent
73
investigations from high-precision U-Pb geochronology (Schoene et al., 2019) report three to
74
four discrete pulses during the main eruption event at KPB, each lasting < 100 ka. The first
75
eruption event that formed the lowermost seven formations lasted from ~ 66.3 to 66.15 Ma
76
ago, followed by the second, third, and fourth pulses at ~ 66.1 to 66.0 Ma, ~ 65.9 to 65.8 Ma,
77
and ~ 65.6 to 65.5 Ma to form the Poladpur Formation, the Ambenali Formation, and the
78
uppermost Mahabaleshwar Formation, respectively (Schoene et al., 2019).
79
A spectrum of geophysical and geochemical studies finds a linkage of the DVP events
80
with the Réunion hotspot (Bredow et al., 2017; Fontaine et al., 2015; Ganerød et al., 2011).
81
Geochemical proxies suggest a link of the source of Deccan basalts to ocean island basalts
82

(OIB), actively erupting on the island of La Réunion (Peters and Day, 2017). Glisovic et al.,
83
2017 based on their geophysical model, predicted a deep mantle origin of DVPs and proposed
84
a mantle plume hypothesis to show its connection with the Réunion hotspot. Interestingly, the
85
temporal coincidence of the Deccan volcanic events with the plume-induced accelerated
86
motion of the Indian plate further strengthens the mantle-plume hypothesis proposed for the
87
origin of Deccan CFB (Cande and Stegman, 2011; Glišović and Forte, 2017). Moreover, like
88
Iceland and Tristan da Cunha, the Réunion hotspot is thought to have originated from a laterally
89
vast thermochemical pile above the core-mantle boundary (CMB) beneath present-day Africa,
90
referred to as the African large low shear-wave velocity province (LLSVP) (Tsekhmistrenko
91
et al., 2021). This pile might have transported primordial material from CMB to the surface via
92
Réunion and other plumes, as evident from geochemical studies on Sr-Nd-Os systematics
93
(Peters and Day, 2017). Although geophysical and geochemical evidence suggests a connection
94
between the Réunion hotspot and African LLSVP, the mechanism of episodic Deccan
95
volcanism is still unknown.
96
In this article we examine the thermochemical scenario that favours the Réunion
97
hotspot to operate in pulsating fashion with characteristic periodicity, producing a huge
98
cumulative volume of Deccan basalt at the KPB. We then show how a single major pulse can
99
give rise to a number of secondary pulses of smaller timescales, as reflected from volcanic
100
episodes in the DVP on time scales less than a million years (Ma). Our thermochemical model
101
allows us to constrain a spectrum of the periodicity timescales (a few Ma to less than a Ma),
102
depending on the thermomechanical properties of the source materials. We present a budget
103
for the volume flux from the mantle to the surface.
104
105
2. Methods
106

2.1. Model set-up
107
The developer version of finite element code ASPECT 2.4.0 (Dannberg and Heister,
108
2016; Heister et al., 2017) is used to develop our thermochemical model, treating the mantle
109
as a system of stratified fluid layers with their density and viscosity varying as a function of
110
pressure, temperature, composition, and phase transformations. The model domain covers the
111
entire vertical depth (~ 2890 km) of the mantle with a horizontal width of 11560 km, which is
112
discretized into  5.5 km cells. Since this work primarily aims to study the dynamics and
113
pulsating nature of the plumes, we assume a pre-existing high-density basal layer of a specified
114
thickness of 150 km (Citron et al., 2020) at the CMB given by a single compositional field to
115
represent the pile (Fig. S2). The changes in the composition field are tracked using passive
116
tracers that advect with the global flow. We imposed an initial sinusoidal temperature profile
117
(Citron et al., 2020; Li et al., 2018) in the background mantle to initiate convection and two
118
thermal boundary layers (TBL) using error functions at the top and the bottom of the domain
119
to represent the CMB TBL and the lithosphere, respectively (Fig. S2c). In addition, heat is
120
introduced into the system by internal heating within the pile (Fig. S2b). We considered heating
121
rate up to 20 times that in the background mantle.
122
To calculate the physical parameters of different model components, we use a depth-
123
dependent composite material model built in the ASPECT material library. All material
124
properties are assigned from an incompressible base model; this model assumes constant
125
parameter coefficients to represent the ambient mantle values, except for the density and
126
viscosity. We use the depth-dependent material model to describe the different viscosities
127
assigned for the lithosphere and upper and lower mantle. Additionally, we consider the thermal
128
and compositional pre-factors to vary the viscosity as a function of temperature and
129
composition. Density varies mainly due to both thermal expansion and compositional
130
variations in our model. The depth-dependent density function also accounts for phase
131

transitions in the ambient mantle and the basal layer. We varied the excess density of the basal
132
layer (pile) from 150 to 450 kg/m3 and its viscosity from 0.1 to 100 times that of the ambient
133
mantle (Fig. S2 a-d).
134
The top and bottom model boundaries are subjected to isothermal conditions with T =
135
298K and T = 3300K, respectively (Fig. S2c). A uniform velocity condition is imposed at the
136
top boundary of the initial model, keeping all other boundaries under a free-slip condition. We
137
reset the boundary conditions of our model to accommodate the temporal variation of plate
138
velocity, to replicate the plate motion history using the plate motion model from previous
139
studies (Seton et al., 2012). A comprehensive list of the model parameters is provided in Table
140
1. To determine the physical properties of sequential plume surges, we consider a line segment
141
across the model box length at a depth of 400 km, which lies above the plume-pulses initiation
142
depth. We then find excess or deficit of physical properties from the peak amplitude from the
143
curve with respect to the background value representing the ambient mantle (Fig. S4).
144
To develop partial melting models, we used a 2D Cartesian box with a vertical depth
145
of 350 km from Earth’s surface and a horizontal length of 700 km (Fig. S5a). The dimensions
146
are reduced to achieve a high-resolution analysis of the melting phenomena. Unlike the whole
147
mantle model, here, we consider the compressibility of both the solid and the melt phases
148
within a two-phase model. The top thermal boundary layer represents the thermal structure of
149
the Indian shield during the Late Mesozoic with a LAB depth of ~160 km. A thermal
150
perturbation of 250-500 K is added at the bottom boundary to represent the excess temperature
151
(non-adiabatic temperature) derived by the plume head from the whole mantle model (Fig. S5b
152
ii). The boundary velocity condition is the same as in the previous model, except for the bottom
153
boundary, where mass can flow in and out, thus supplying plume material to generate
154
successive melt pulses. Initially, the system is considered to be free from porosity. We used
155
mesh deformation at the upper boundary to track the surface topography generated in the
156

successive melting events. The details of the model parameters is given in Table 2.
157
2.2. Problem formulation
158
Our 2D thermochemical convection simulations are developed in a theoretical framework
159
of Boussinesq approximation, using mass, momentum, and energy conservation equations:
160

(1)
󰇟󰇗󰇠
(2)


(3)
where , , , 󰇗 denote the following physical variables: velocity, dynamic pressure,
161
viscosity, and strain rate, respectively.  is the gravitational acceleration,  is the reference
162
density of the ambient mantle,  is the specific heat at constant pressure, and , , and  are
163
the absolute temperature, thermal conductivity, and the rate of internal heating, respectively.
164
To replicate Earth-like convective vigor, we choose a set of parameters to appropriately
165
fix the reference Rayleigh number for the mantle:
166


(4)
,  and  represent the reference values of the coefficients of thermal expansion, the
167
thermal diffusivity, and the viscosity of the ambient mantle, respectively. The basal layer has
168
a density difference with the ambient mantle, which is introduced in our modelling as
169
Buoyancy number:
170
 

(5)
B expresses the intrinsic density anomaly normalized to that caused by thermal expansion.
171

Discontinuous Galerkin method is used in ASPECT to implement tracking of compositional
172
fields. The advection of composition is given by
173

󰇛󰇜
(6)
where  is the compositional vector.
174
The density and viscosity in the material model vary according to the following equations.
175
󰇛󰇜󰇛󰇜󰇛󰇜
(7)
󰇛󰇜󰇛󰇜
(8)
where  and  are the viscosity and density calculated from the base model;  and  denote
176
their corresponding reference values.  is the coefficient of thermal expansion,  is the
177
density difference between the source layer and the ambient mantle,  stands for the first
178
component of the compositional vector . The temperature pre-factor in eq. 7 is expressed as
179
󰇛󰇜󰇧󰇛󰇜
󰇨
(9)
where  is the thermal viscosity exponent, and
180
󰇱
 

(10)
󰇛󰇛󰇜󰇜.  and  represents the minimum and the maximum values
181
of the thermal pre-factors, respectively. The compositional pre-factor in eq 7 is taken in the
182
form:
183
󰇛󰇜
(11)
 is the compositional viscosity pre-factor corresponding to composition c0. From a depth-
184

dependent model, we find model viscosity:
185
󰇛󰇜󰇛󰇜󰇛󰇜

(12)
where 󰇛󰇜 is the depth-dependent viscosity calculated from a depth-dependent model.
186
Depth dependent phase transition is defined in ASPECT, the expression of which follows,
187


(13)
 denotes the phase-transition zone width.  is the pressure difference across the width of
188
phase transition zones, given by
189
󰇛󰇜
(14)
where  is the Clapeyron slope.
190
ASPECT calculates the dynamic topography from the stress at the surface in the following
191
way. First, it evaluates the stress component that acts in the direction of gravity at the centres
192
of the cells along the top model surface. The dynamic topography is then calculated using,
193
 
󰇛󰇜
(15)
where  is the dynamic topography,  is the stress calculated in the previous step,  is the
194
density of the corresponding cell center, and  is the component of gravity.
195
The melting model is implemented in ASPECT by separating out the fluid phase from its solid
196
counterpart, which is related by compaction pressure as,
197
󰇛󰇜
(16)
where  is porosity,  is the solid pressure and  is the fluid pressure. After computing the
198
stokes equation, the fluid velocity is calculated from Darcy’s equation,
199



(17)
where  is the fluid velocity and  is the solid velocity,  is the Darcy coefficient, and 
200
is fluid density. The porosity is advected using the following relation,
201

 
󰇛󰇜
(18)
 is the rate of melting. Permeability is then calculated from
202
󰇛󰇜
(19)
 is the reference permeability.
203
204
3. Results
205
3.1. Pulsating rise of thermochemical plumes
206
We consider pile density, viscosity, concentration of heat-producing elements (HPE), and
207
major phase transitions in the mantle to obtain a reasonable plume model for the Deccan LIP
208
evolution in the geodynamic framework of the Réunion hotspot. In this modelling, the
209
buoyancy number (), which measures density contrast of the pile with the ambient lower
210
mantle, accounts for varying relative proportions of eclogite and peridotite within the basal
211
layer. As the viscosity and heat-producing element concentration of the pile are not well
212
constrained, we varied them within a plausible range of their values found in the literature
213
(Citron et al., 2020; Dannberg and Sobolev, 2015; Heyn et al., 2020; Li et al., 2018). A velocity
214
boundary condition is imposed at the upper model boundary to replicate the lithospheric plate
215
kinematics that prevailed during Réunion hotspot activities. The details of the model domain
216
and parameters used for the simulations along with the initial model boundary conditions are
217
provided in the Methods section and Supplementary Figs. S1, S2.
218

The plate velocity induces downwelling flow in the mantle, which forces the thermal
219
boundary layer (TBL) at the CMB to pile up laterally and increase its thickness
220
(󰇜 (Fig. S3a). The TBL is pushed towards the pile to increase  further (Fig.
221
S3b), amplifying the Rayleigh number (Ra) in the TBL to locally exceed the critical . Under
222
this threshold condition, the buoyancy head becomes high enough to force the material to flow
223
vertically against gravity, forming a thermochemical plume (Figs. S3c, d). Due to its strong
224
buoyancy flux, the plume grows mainly in the vertical direction within the lower mantle.
225
However, on encounter with the upper mantle, it faces two processes that significantly hinder
226
its continuous growth: 1) influence of the plate velocity and 2) phase transition between 300
227
and 400 km (Dannberg and Sobolev, 2015). At this stage, the plate-driven flow extends to a
228
depth of 660 km and exert drags to the plume head (Fig. S3e), detaching it from the tail
229
counterpart. The buoyancy ultimately takes over the drag, allowing the head to move vertically
230
upward in the form of a solitary pulse (Figs. 2a i-iv). The ascending head undergoes phase
231
transformations: coesite to stishovite and pyroxene to garnet, increasing the plume density.
232
Ultimately, inherent high excess temperatures enable plumes to overcome this density-
233
enhancing barrier to reach the lithosphere-asthenosphere boundary (LAB), where they spread
234
laterally in the horizontal direction. This stagnation process facilitates thermal mixing and
235
mechanical entrainment within the mantle.
236
The model run shows that the plume upwells in a pulsating fashion to produce multiple
237
heads in the course of the ascent event (Figs. 2a i-iv). The primary head gives rise to the first
238
pulse following its detachment from the main body after crossing the 660 km boundary (Fig.
239
2a i). The initiation of the plume destabilizes the pile margin (Figs. 2a i, ii), as indicated by
240
reducing pile volumes and high rates of its lateral migration (~ 10 km/Ma) (Figs. 2b ii, iii), that
241
produces relatively high eclogite proportions (~ 10%) and heat-producing element
242
concentration in the plume (Fig. S4a). A large buoyancy head due to the high excess
243

temperature (> 500 K) and density contrast (> -50 kg/m3) (Figs. S4b, c) facilitates the surface
244
to attain high dynamic topography with an elevation of ~1600 m (Fig. 2a i inset) and inflates
245
the pulse volume () (Fig. 2b i). The pile margin remains unstable (Fig. 2a ii),
246
forcing a large volume of material to upwell through the plume tail and produce a second pulse
247
(Fig. 2b ii) as time elapses after the first pulse allowing new materials to accumulate in a
248
threshold volume at 660 km. Unlike the first pulse, the second pulse evolves with a moderate
249
amount of eclogite and heat-producing elements (HPE) to form significantly lower pulse
250
volume () and dynamic topography (~800 m) (Fig. 2a ii inset) owing to its lower
251
excess temperature (~400 K) and density contrast (> -40 kg/m3) (Figs. S4b, c). With time the
252
pile moves further away from the plume axis but the rate of the movement reduces to ~5-6
253
km/Ma. It sustains the periodic material supply to the 660 km boundary to produce tertiary
254
pulses (Figs. 2a iii, iv; 2b iii). The pile eventually attains a stable state, and unstable to stable
255
transition results in a drastic reduction in material volume supply to the plume (Fig. 2b ii),
256
marked by much lower pulse volume () and a low positive dynamic
257
topography (~100-200 m) (Figs. 2a iii, iv insets) with low excess temperature (~ 250 K) and
258
density contrast (~-20 kg/m3).
259
Although all the sequential pulses ultimately reach the LAB and take part in melting
260
and subsequent volcanism, the primary (first) pulse, owing to its sufficient excess temperature
261
(> 500 K), and volume () takes the lead role in forming LIPs. The
262
thermochemical pile, which is the primary material feeder for the pulses, stratifies a specific
263
set of physio-chemical parameters to generate a reasonable melt volume and dynamic
264
topography required for the formation of Deccan LIP.
265
3.2. Buoyancy effects on plume rise dynamics
266

We performed a series of simulations to investigate how the buoyancy number () of
267
the pile influenced the process of pulse generation at the mid-mantle transition zone for a given
268
viscosity ratio ( ~1) and HPE concentration. For low  values (), the mantle flow
269
efficiently drags the pile horizontally to widen the exposed CMB fraction, causing both 
270
and pile height () to increase at high rates (Figs. 3a, c; 4a). Consequently, the pile becomes
271
unstable (Fig. 3a) to accelerate material flux into the plume and gives rise to initial pulses with
272
large volumes () and dynamic topography (> 1500 m) (Figs. 3a, c inset; Fig.
273
4b). Increasing  weakens the interaction of mantle flow with the pile due to a high intrinsic
274
density of the basal layer, leading to TBL thickening at slow rates. (Figs. 3b, d; Fig. 4a). As a
275
result, the plume having the same initial excess temperature produces pulses of much smaller
276
volumes () (Fig. 4b) and dynamic topography (<1100 m) (Figs. 3b, d insets).
277
Moreover, the volume difference in primary, secondary, and tertiary pulses are much more
278
pronounced at a lower B value (Fig. 4a).
279
3.3. Viscosity effects on pulse-driven processes
280
Geophysical studies suggest that the viscosity of thermochemical piles can be up to
281
1000 times higher than the ambient mantle (Heyn et al., 2020). We find that an increase in the
282
viscosity ratio () from 1 to 100 considerably dampens the vertical growth of piles, allowing
283
them to remain stable for given values of B and HPE concentration (Fig. 3e-h), as reflected
284
from the lower rates of pile volume changes (Fig. 4c). This increase in , on the other hand,
285
strengthens the interaction of mantle flow with the pile, as evidenced from large exposed CMB
286
areas (Fig. 4a). Such a strong interaction increases the horizontal shortening of the pile at the
287
cost of vertical growth, eventually reducing pulse volumes by up to 12 % (Fig. 4b) and
288
widening the time periodicity of pulse generation.  also significantly influences the dynamic
289
topography. The model estimates for  yield a large dynamic topography ( m)
290

when  is low (< 0.8), which is considered unrealistic for thermochemical plumes. Increasing
291
 to 100 depresses the topography to < 2000 m for lower values of  which can be correlated
292
with Deccan volcanic events.
293
3.4. Effect of internal heat production on plume dynamics
294
Geochemical observations on OIBs support the presence of enriched mantle reservoirs
295
as mantle heterogeneity and/or variable mantle reservoirs (Peters and Day, 2017). Some of
296
these sources are less degassed and hence, are more enriched in HPEs. One possibility is that
297
such reservoirs could be present within LLSVPs since they are primarily composed of
298
primordial material, subducted Hadean crust, or recycled oceanic crust remnants from a
299
decomposed subducted plate (Deschamps et al., 2011). Previous estimates, based on heat
300
budget calculations, show that the heat-producing element concentrations (󰇜 can be as high
301
as 20 to 25 times that of the background mantle (Citron et al., 2020). To study the role of this
302
factor on the pile dynamics, we increased  of the pile by up to 20 times that of the ambient
303
lower mantle. Such enrichment augments the pile buoyancy with time to set a gravitationally
304
unstable state of the pile even under a high condition. Plumes that originate from pile edges
305
in our models entrain HPE-enriched pile materials to increase its excess temperature. However,
306
 has relatively weak effects, as compared to other parameters, such as viscosity ratio ()
307
(Figs. 3 c-d, g-h).  primarily affects the dynamic topography and, more importantly, the
308
material supply to thermochemical plumes (Figs. 4b, c). Increase in  amplifies the dynamic
309
topography and also enhances material supply to the plume, especially at a lower value of the
310
buoyancy number (). The other remarkable effect of  on plume geometry is that the
311
plume develops a thick tail, which facilitates pile material transport to the mid-mantle region
312
in larger volumes (Fig. 3g), compared to that produced in a lower  condition. This is also
313
reflected in higher rate of reduction in pile volumes with time (Fig. 4c), which implies a more
314
effective pile material entrainment into the plume tail. In addition, high  causes the plume
315

to gain a higher excess temperature that results in dynamic topography with a realistic elevation
316
of ~1600 m for the primary pulse for  = 100 (Fig. 3g inset).
317
3.5. Melt transport from a thermochemical plume
318
When the plume head approaches the LAB, the temperature inside the plume exceeds
319
the local solidus to initiate the melting process in the plume materials. This phenomenon
320
inevitably increases the porosity of the system, which thereby enhances permeability in the top
321
region of the plume head (Fig. S6). The Indian shield (a stable craton) had a thickness of 150-
322
200 km before it started to interact with the plume (Naganjaneyulu and Santosh, 2012),
323
implying a deep upper thermal boundary layer. Depending upon the initial temperature,
324
composition, and volume of the plume head, the melting process is onset at a depth varying
325
from . During the initial phase of ascent, the magnitudes of melt and plume
326
velocities lie compatibly in a range of  (Fig. S5b i), but as the plume ascends
327
to a shallower depth, the melts owing to their lower density (2700 kg/m3), gain a much higher
328
velocity (> 1.2 m/year) to segregate from the plume materials at the LAB (Fig. S5b v). Our
329
model results suggest that the melt-ascent velocity is directly proportional to the porosity in the
330
system, which increases steadily with the plume evolution. Unlike the plume head, the
331
segregated melts always ascend nearly in a vertical direction, implying that the plate velocity
332
hardly affects the upward melt flow dynamics. At a depth of ~60-80 km, the segregated melts
333
start to spread laterally, forming a melt pool below the lithosphere (a permeability barrier) (Fig.
334
5a). The melt front interacts with the lithosphere to produce horizontal shear that sets in small-
335
scale downwelling and causes thinning of the TBL. Upwelling of the melt front within the
336
lithosphere ultimately gives rise to volcanism. We evaluated the melt volume, velocity, time
337
scale of the melt rise, and dynamic topography as a function of the initial plume volume,
338
temperature, and density, which are presented in Fig. 5 and Fig. S5b.
339

Since the primary plume pulse has the highest volume (~ km3), it contains a
340
high concentration of HPEs. This condition, aided with a high excess temperature (~500 K)
341
(Fig. S5), enables the pulse to overcome the upper-mantle buoyancy barriers. Model results
342
show that the higher excess temperatures and HPE concentrations result in a greater melting
343
depth (~250 km) of the initial melt pulse (Fig. 5a i), and also enhance the excess buoyancy,
344
that accelerates the upward flow of melts to reach a depth of 50 km within 150-180 kyr (Fig.
345
5b). The porosity evolution, coupled with a high excess temperature, facilitates melt generation
346
during the plume ascent to produce an enormous volume (km3) of melts at the
347
LAB (Fig. 5b). This melt pool then efficiently incorporates lithospheric materials by thermal
348
erosion to increase the melt volume further (km3), ultimately giving rise to
349
massive volcanism. Following this melt pulse generation, the plume head is then significantly
350
depleted in HPE concentration. Secondly, the heat dissipation to the ambient mantle lowers the
351
excess temperature (~300 K) in the plume. The thermal change by these mechanisms relocates
352
the melting depth at a shallower level (150 to 180 km) during the subsequent pulses, where a
353
moderate excess temperature, a relatively low HPE concentration, and smaller plume volume
354
set the upward melt flows at slow rates (), taking up to 300 kyr to reach the LAB.
355
These second-generation pulses reduce their melt volumes to  km3. In addition,
356
the thermal erosion of the lithosphere at the LAB by the melt pools becomes less effective and
357
fails to substantially increase the melt volumes (Fig. 5a ii). Thus, they produced erupted
358
volumes significantly lower than those produced in the first pulse. The smaller pulses are
359
manifested in relatively low topographic elevations (Figs. 5b, c). The tertiary melt pulses
360
further reduce their volumes and their excess temperatures (~ 250 K) and lose their capacity
361
for large-scale thermal erosion of the lithosphere and attaining a stagnation state at a depth of
362
~50 km (Fig. 5a iii).
363
364

4. Discussion
365
4.1. Deccan volcanism - African superplume connection
366
It is now a well-accepted hypothesis that the existence of African LLSVP dates back to
367
at least the Pangea event (Zhang et al., 2010). During the Gondwana-proto-Laurussia
368
convergence, several cold subducting slabs assembled in the lower mantle beneath the African
369
continental lithosphere to form this distinct layer above the CMB, whose current location and
370
shape have been framed in the post-Pangea subduction history. Recent mantle convection
371
models coupled to continuously evolving plate boundaries (Hassan et al., 2016; Müller et al.,
372
2016) track the African LLSVP positions through time, considering the subduction driven
373
mantle flow due to Neo-Tethys Ocean closure, as illustrated in Fig. 6a, b. The model results
374
suggest that the western margin of African LLSVP remained almost stable during the entire
375
Cretaceous period, but the eastern flank has continuously relocated its position. The time-
376
dependent effect of subduction on the north (closure of Tethys) produced a strong southward
377
lower-mantle poloidal flow (Fig. 6a), leading to mantle upwelling in the south. The upwelling
378
dynamics, in turn, induced a convective mantle “roll” that forced the eastern flank of the
379
African LLSVP boundary to migrate southward and the Indian plate to move northward at a
380
higher velocity (Glišović and Forte, 2017). These interpretations are further validated by
381
geophysical observations that predict deformation and southward movement of the African
382
LLSVP under east Africa (Ford and Long, 2015).
383
Our modelling domain considers a north-south cross-section of the mantle to replicate
384
the Indian plate movement in late Mesozoic and Cenozoic (past 130 Ma) and reconstruct the
385
eastern flank position of the African LLSVP relative to the Indian subcontinent (Fig. S1). The
386
model simulations suggest that a large mantle roll, formed as a consequence of the subduction
387
in the north (cf. Glisovic et al., 2017), forced the pile to move in the southward direction at a
388
rate of 17-19 km/Ma at the beginning of the Late Cretaceous (Fig. 6c ii). This postulate is
389

consistent with the inferences from other studies that claimed the southward movement of
390
African LLSVP due to the presence of deep-mantle southward poloidal flow as a consequence
391
of the Tethyan subduction over the past 130 Ma (Hassan et al., 2020). The poloidal flow
392
resulted in a thermal instability within the exposed CMB on the north of the LLSVP, which
393
subsequently migrated towards the African LLSVP and amplified the pile at its eastern flank
394
to attain a thickness of ~800-1000 km (Fig. S3). The laterally migrating TBL instabilities
395
climbed up the pile edge to reach the crest and finally formed a mature plume. The positional
396
reconstruction of the African LLSVP and the Indian plate for this time period allows us to
397
conclude that the eastern flank of African LLSVP coincided with the Indian plate location in a
398
time frame of 70-65 Ma (Fig. 6b). This plume then generated successive pulses upon reaching
399
the mid-mantle transition zone through the late Mesozoic and Cenozoic, where the first pulse
400
corresponds to the Deccan events at 66 Ma. The plume initiation decelerated the southward
401
pile migration to ~ 6-7 km/Ma (Fig. 6c ii) because the plume forced pile materials to effectively
402
upwell in the vertical direction. Subsequently, the pile migrated further south-westward,
403
whereas the Indian plate had north-eastward movement.
404
The plume continued to form periodically the secondary and tertiary pulses at mid-
405
mantle depth at an interval of 5-8 Ma, giving rise to successive eruptions from the Réunion
406
hotspot. The plume process eventually reduced pulse volumes and involved a sharp change in
407
the chemical characteristics of the Réunion lava flows during the post-Deccan volcanism
408
period (Peters and Day, 2017). With time, the eastern margin of the pile shifted its position
409
further southwest to reach its current location (Fig. 6b). The present model suggests that the
410
process of sequential plume-head detachment at the mid-mantle transition zone modulated the
411
periodic pulse generation and determine the time scale, volume, and topography associated
412
with each of these pulses. Considering a CMB temperature of  and an initial pile
413
thickness of , the model results for B in a range  yield a periodicity of 5-8 Ma,
414

similar to that of Réunion activity throughout the Cenozoic. To tally the dynamic topography,
415
the pile also needs to be ~ 100 times viscous () and  times HPE enriched than the
416
ambient lower mantle. This condition produces a primary pulse volume of 
417
 and dynamic topography of  related to the Deccan event, followed by the
418
next generation of pluses with volumes of , ,
419
(Fig. 4a) and topography of , , and .
420
4.2. The Deccan volcanic periodicity
421
To study the time periodicity of Deccan volcanism, we focus on the melting process in
422
the primary plume head obtained from our thermochemical model (Fig. 5; Fig. S5). The model
423
results suggest that the plume head locally underwent melting within the asthenosphere to
424
create three eruptive events within a time scale of 1 Ma, where the first event occurred within
425
0.15 Ma from the plume head stagnation with a cumulative volume of  (Fig.
426
5b), correlated with the lowermost seven formations produced during the period ~ 66.5-66.3
427
Ma. The second event took place after a quiescent period of ~ 0.3 Ma with a volume of 
428
, which corresponds to the ~ 66.0 Ma Poladpur Formation. Finally, the third pulse that
429
initiated after 0.4 Ma produced a volume of which can be equated with the
430
Ambanali and later formations deposited during ~ 65.6-65.3 Ma. Based on these model
431
calculations, we estimate a volume flux of ~8-9 km3/year for the first event, subsequently
432
reduced to  and , respectively, for the second and third
433
events. This estimate implies that the rate of Deccan volcanic eruption in a pulse (time scale ≤
434
100 Ka) exceeded the global value (3 to 4 km3/year) by a factor of 1.5 to 3. Moreover, there
435
must be hiatuses in the order of tens of thousands of kiloyears within the pulses to balance the
436
total volume estimates. Geochemical proxies also suggest a sharp increase of mantle
437
contributions to later volcanic formations, such as Poladpur and Ambenali, indicating a
438

reduction of magma-crust interface area (Renne et al., 2015). The higher rates of thermal
439
erosion at the LAB during the first two events effectively thinned the lithosphere and weakened
440
the vigorousity crust-mantle interaction during the subsequent melt pulse events, as revealed
441
from our models (Fig. 5a).
442
Although our model estimates broadly agree with the time gaps between different
443
episodes of the Deccan volcanism, they somewhat underestimate the erupted volumes
444
predicted from petrological and geochemical studies (Schoene et al., 2019). Groups of flows
445
within the Poladpur and Mahabaleshwar Formations, each potentially comprising > 50,000
446
km3, lack any secular variation of paleomagnetic poles, suggesting the eruption at high rates,
447
 on decadal to centuries scales. Our volume and flux estimates for eruptions
448
prior to the KPB tally well with the available data; however, they do not account for either the
449
high melt volumes or the rate of eruption for the post-KPB eruptions. We thus hypothesize that
450
there was a transition in the nature of volcanism across the KPB, the explanation of which
451
demands the possible effects of other internal or external factors. One possible explanation
452
could be that the Chicxulub bolide impact accelerated the eruption rates, as suggested by the
453
previous workers (Renne et al., 2015).
454
4.3. Comparison with major global LIP events
455
We will now discuss the Deccan volcanism that occurred sequentially in three major
456
pulses in the context of similar episodic volcanic events from other LIPs and hotspots, such as
457
Hawaii, Réunion, Yellowstone, and others (Morrow and Mittelstaedt, 2021). They show the
458
periodicity of the volcanic events on varied timescales (Fig. 7). For example, the Hawaii-
459
Emperor hotspot track records a sequence of magmatic pulses at around 64 Ma, 50 Ma, 42 Ma,
460
and 28 Ma, implying a pulsating time scale of about 10 Ma (Van Ark and Lin, 2004). On the
461
other hand, from bathymetry analysis Wessel (2016) has established a much shorter pulsating
462

time scale (< 2 Ma) for the post-22 Ma volcanism, as observed in the Deccan volcanism. The
463
Yellowstone LIP started its volcanism at around 18 Ma (Schutt et al., 2008), followed by two
464
distinct magmatic peak events at around 11 Ma and 5 Ma (Stachnik et al., 2008; Waite et al.,
465
2006). In a recent study of the Yellowstone super-volcano, the tomographic P-wave model has
466
detected hot pulses in the upper mantle (Huang et al., 2015). These discrete bodies, most
467
probably pockets of partial melts, represent episodic pulses produced by a large plume source
468
in the mantle, as predicted from our numerical model simulations. The Réunion hotspot
469
displays a major emplacement in Deccan traps at 66-68 Ma, with later magmatic peaks at 57
470
Ma, 48 Ma, 35 Ma, 8 Ma, and 2 Ma (Mjelde et al., 2010).
471
4.4. Model limitations
472
The model presented here treats the lithosphere as an upper thermal boundary layer,
473
which does not account for visco-plastic rheology with a failure criterion, which is a limitation.
474
in our simulations. Secondly, the creep processes that are often activated in the upper mantle
475
could influence the shape and the ascent rate of the plume head, which are not explored in this
476
study. In addition, our primary model excludes any compressibility effect of the solid phases.
477
The plume melting models consider a reaction time scale of 103 years due to computational
478
constraints. This might overshoot the overall timescale of melting and melt migration.
479
480
5. Summary and Conclusions
481
The 2D thermochemical simulations demonstrate that the following parameters: pile-
482
ambient mantle viscosity ratio (), buoyancy number (), and heat-producing element
483
concentration () have controlled the Réunion hotspot dynamics and its connection to the
484
seismically observed African LLSVP. The position of India is found to match with the African
485
LLSVP location at the end of the Cretaceous Period, where the LLSVP acted as the source of
486

the Réunion hotspot materials to produce the Deccan LIP and subsequent eruption events.
487
We show that an instability in the TBL above the CMB played a critical role in the
488
Reunion hotspot formation. The instability initiated on the eastern flank of the African LLSVP
489
during the Neo-Tethys subduction (130-150 Ma) but migrated to the pile crest to form a plume.
490
The plume ascent was perturbed at the mid-mantle transition zone to produce four major pulses
491
on a time interval of 5-8 Ma. The model calculations suggest that, at the onset time (Late
492
Cretaceous) of Réunion Hotspot volcanism, the African LLSVP had a Buoyancy number (󰇜
493
in the range of , pile-ambient mantle viscosity ratio () in the order of 100 and heat-
494
producing element concentration (󰇜times that of the ambient lower mantle. The
495
primary pulse of the Réunion plume had thereby sufficient volumes () and
496
excess temperature () to produce the Deccan LIPs. The partial-melting model
497
envisages that the primary pulse subsequently gave rise to 3 melt pulses with volumes in the
498
order of ~  km3 at a time interval of  Ma, as recorded in the
499
Deccan traps.
500
Finally, we conclude that most of the LIPs evolve in pulses on characteristic time scales,
501
under the influence of combined action of the pile processes operating at the CMB and the
502
feeding mechanism into the plume stem, modulated by a mid-mantle perturbation. The entire
503
sequence of pulses is divided into two categories: major pulses with a periodicity of 5-8 Ma,
504
determined by the plume-head detachment at the mid-mantle transition zone, and minor melt
505
pulses with a 0.15-0.4 Ma time periodicity, determined by the melting phenomenon within each
506
major pulse. The temporal variations in magma eruption characteristics are consistent with
507
depth-dependent compositional heterogeneity of the plume source.
508
509

Acknowledgments
510
This work has been supported by the DST-SERB through the J. C. Bose fellowship
511
(SR/S2/JCB-36/2012) to NM. We thank the Computational Infrastructure for Geodynamics
512
(geodynamics.org) which is funded by the National Science Foundation under awards EAR-
513
0949446 and EAR-1550901 for supporting the development of ASPECT. D.G. and J.S. are
514
thankful to Juliane Dannberg for helpful discussions. We especially thank Dr. Simon
515
Gatehouse, BHP who critically read an earlier version of this manuscript and provided many
516
constructive suggestions.
517
518
Data availability
519
The model parameters required to produce the results are given in the tables and the
520
supplementary information. The simulation code is freely available online under the terms of
521
the GNU General Public License at https://github.com/geodynamics/aspect. Parameter files
522
that reproduce the findings of this study are available from the corresponding author upon
523
reasonable request.
524
525
526

References
527
Alvarez, L., Alvarez, W., Asaro, F., Michel, H. V., 1980. Extraterrestrial Cause for the Cretaceous-
528
Tertiary Extinction. Science (80-. ). 208, 1095–1108.
529
Bredow, E., Steinberger, B., Gassmöller, R., Dannberg, J., 2017. How plume-ridge interaction shapes
530
the crustal thickness pattern of the Réunion hotspot track. Geochemistry, Geophys. Geosystems
531
18, 2930–2948. https://doi.org/10.1002/2017GC006875
532
Campbell, I.H., Griffiths, R.W., 1990. Implications of mantle plume structure for the evolution of
533
flood basalts. Earth Planet. Sci. Lett. 99, 79–93. https://doi.org/10.1016/0012-821X(90)90072-6
534
Cande, S.C., Stegman, D.R., 2011. Indian and African plate motions driven by the push force of the
535
Réunion plume head. Nature 475, 47–52. https://doi.org/10.1038/nature10174
536
Chenet, A.L., Courtillot, V., Fluteau, F., Gérard, M., Quidelleur, X., Khadri, S.F.R., Subbarao, K. V.,
537
Thordarson, T., 2009. Determination of rapid Deccan eruptions across the Cretaceous-Tertiary
538
boundary using paleomagnetic secular variation: 2. Constraints from analysis of eight new
539
sections and synthesis for a 3500-m-thick composite section. J. Geophys. Res. Solid Earth 114.
540
https://doi.org/10.1029/2008JB005644
541
Chenet, A.L., Quidelleur, X., Fluteau, F., Courtillot, V., Bajpai, S., 2007. 40K-40Ar dating of the
542
Main Deccan large igneous province: Further evidence of KTB age and short duration. Earth
543
Planet. Sci. Lett. 263, 1–15. https://doi.org/10.1016/j.epsl.2007.07.011
544
Citron, R.I., Lourenço, D.L., Wilson, A.J., Grima, A.G., Wipperfurth, S.A., Rudolph, M.L., Cottaar,
545
S., Montési, L.G.J., 2020. Effects of Heat‐Producing Elements on the Stability of Deep Mantle
546
Thermochemical Piles. Geochemistry, Geophys. Geosystems 21, 1–17.
547
https://doi.org/10.1029/2019GC008895
548
Crameri, F., Shephard, G.E., Heron, P.J., 2020. The misuse of colour in science communication. Nat.
549
Commun. 11, 1–10. https://doi.org/10.1038/s41467-020-19160-7
550
Dannberg, J., Heister, T., 2016. Compressible magma/mantle dynamics: 3-D, adaptive simulations in
551
ASPECT. Geophys. J. Int. 207, 1343–1366. https://doi.org/10.1093/gji/ggw329
552
Dannberg, J., Sobolev, S. V., 2015. Low-buoyancy thermochemical plumes resolve controversy of
553
classical mantle plume concept. Nat. Commun. 6, 1–9. https://doi.org/10.1038/ncomms7960
554
Deschamps, F., Kaminski, E., Tackley, P.J., 2011. A deep mantle origin for the primitive signature of
555
ocean island basalt. Nat. Geosci. 4, 879–882. https://doi.org/10.1038/ngeo1295
556
Farnetani, C.G., Richards, M.A., 1994. Numerical investigations of the mantle plume initiation model
557
for flood basalt events. J. Geophys. Res. 99. https://doi.org/10.1029/94jb00649
558
Fontaine, F.R., Barruol, G., Tkalčić, H., Wölbern, I., Rümpker, G., Bodin, T., Haugmard, M., 2015.
559
Crustal and uppermost mantle structure variation beneath La Réunion hotspot track. Geophys. J.
560
Int. 203, 107–126. https://doi.org/10.1093/gji/ggv279
561
Ford, H.A., Long, M.D., 2015. A regional test of global models for flow, rheology, and seismic
562
anisotropy at the base of the mantle. Phys. Earth Planet. Inter. 245, 71–75.
563
https://doi.org/10.1016/j.pepi.2015.05.004
564
Ganerød, M., Torsvik, T.H., van Hinsbergen, D.J.J., Gaina, C., Corfu, F., Werner, S., Owen-Smith,
565
T.M., Ashwal, L.D., Webb, S.J., Hendriks, B.W.H., 2011. Palaeoposition of the seychelles
566
microcontinent in relation to the deccan traps and the plume generation zone in late cretaceous-
567
early palaeogene time. Geol. Soc. Spec. Publ. 357, 229–252. https://doi.org/10.1144/SP357.12
568
Glišović, P., Forte, A.M., 2017. On the deep-mantle origin of the Deccan Traps. Science (80-. ). 355,
569

613–616. https://doi.org/10.1126/science.aah4390
570
Hassan, R., Müller, R.D., Gurnis, M., Williams, S.E., Flament, N., 2016. A rapid burst in hotspot
571
motion through the interaction of tectonics and deep mantle flow. Nature 533, 239–242.
572
https://doi.org/10.1038/nature17422
573
Hassan, R., Williams, S.E., Gurnis, M., Müller, D., 2020. East African topography and volcanism
574
explained by a single, migrating plume. Geosci. Front. 11, 1669–1680.
575
https://doi.org/10.1016/j.gsf.2020.01.003
576
Heister, T., Dannberg, J., Gassmöller, R., Bangerth, W., 2017. High accuracy mantle convection
577
simulation through modern numerical methods - II: Realistic models and problems. Geophys. J.
578
Int. 210, 833–851. https://doi.org/10.1093/gji/ggx195
579
Heyn, B.H., Conrad, C.P., Trønnes, R.G., 2020. How Thermochemical Piles Can (Periodically)
580
Generate Plumes at Their Edges. J. Geophys. Res. Solid Earth 125.
581
https://doi.org/10.1029/2019JB018726
582
Huang, H., Lin, F., Schmandt, B., Farrell, J., Smith, R.B., Tsai, V.C., 2015. The Yellowstone
583
magmatic system from the mantle plume to the upper crust. Science (80-. ). 348, 773–776.
584
Kale, V.S., Bodas, M., Chatterjee, P., Pande, K., 2020. Emplacement history and evolution of the
585
Deccan Volcanic Province, India. Episodes 43, 278–299.
586
https://doi.org/10.18814/EPIIUGS/2020/020016
587
Keller, G., Adatte, T., Bhowmick, P.K., Upadhyay, H., Dave, A., Reddy, A.N., Jaiprakash, B.C.,
588
2012. Nature and timing of extinctions in Cretaceous-Tertiary planktic foraminifera preserved in
589
Deccan intertrappean sediments of the Krishna-Godavari Basin, India. Earth Planet. Sci. Lett.
590
341–344, 211–221. https://doi.org/10.1016/j.epsl.2012.06.021
591
Li, M., Zhong, S., Olson, P., 2018. Linking lowermost mantle structure, core-mantle boundary heat
592
flux and mantle plume formation. Phys. Earth Planet. Inter. 277, 10–29.
593
https://doi.org/10.1016/j.pepi.2018.01.010
594
Mjelde, R., Wessel, P., Müller, R.D., 2010. Global pulsations of intraplate magmatism through the
595
Cenozoic. Lithosphere 2, 361–376. https://doi.org/10.1130/L107.1
596
Morrow, T.A., Mittelstaedt, E.L., 2021. Quantifying Periodic Variations in Hotspot Melt Production.
597
J. Geophys. Res. Solid Earth 126. https://doi.org/10.1029/2021JB021726
598
Müller, R.D., Seton, M., Zahirovic, S., Williams, S.E., Matthews, K.J., Wright, N.M., Shephard, G.E.,
599
Maloney, K.T., Barnett-Moore, N., Hosseinpour, M., Bower, D.J., Cannon, J., 2016. Ocean
600
Basin Evolution and Global-Scale Plate Reorganization Events since Pangea Breakup. Annu.
601
Rev. Earth Planet. Sci. 44, 107–138. https://doi.org/10.1146/annurev-earth-060115-012211
602
Naganjaneyulu, K., Santosh, M., 2012. The nature and thickness of lithosphere beneath the Archean
603
Dharwar Craton, southern India: A magnetotelluric model. J. Asian Earth Sci. 49, 349–361.
604
https://doi.org/10.1016/j.jseaes.2011.07.002
605
Peters, B.J., Day, J.M.D., 2017. A geochemical link between plume head and tail volcanism.
606
Geochemical Perspect. Lett. 5, 29–34. https://doi.org/10.7185/geochemlet.1742
607
Petersen, S. V., Dutton, A., Lohmann, K.C., 2016. End-Cretaceous extinction in Antarctica linked to
608
both Deccan volcanism and meteorite impact via climate change. Nat. Commun. 7, 1–9.
609
https://doi.org/10.1038/ncomms12079
610
Renne, P.R., Sprain, C.J., Richards, M.A., Self, S., Vanderkluysen, L., Pande, K., 2015. State shift in
611
Deccan volcanism at the Cretaceous-Paleogene boundary, possibly induced by impact. Science
612
(80-. ). 350, 76–78.
613
Richards, M.A., Alvarez, W., Self, S., Karlstrom, L., Renne, P.R., Manga, M., Sprain, C.J., Smit, J.,
614

Vanderkluysen, L., Gibson, S.A., 2015. Triggering of the largest Deccan eruptions by the
615
Chicxulub impact. Bull. Geol. Soc. Am. 127, 1507–1520. https://doi.org/10.1130/B31167.1
616
Ritsema, J., Deuss, A., Van Heijst, H.J., Woodhouse, J.H., 2011. S40RTS: A degree-40 shear-velocity
617
model for the mantle from new Rayleigh wave dispersion, teleseismic traveltime and normal-
618
mode splitting function measurements. Geophys. J. Int. 184, 1223–1236.
619
https://doi.org/10.1111/j.1365-246X.2010.04884.x
620
Schmidt, A., Skeffington, R.A., Thordarson, T., Self, S., Forster, P.M., Rap, A., Ridgwell, A., Fowler,
621
D., Wilson, M., Mann, G.W., Wignall, P.B., Carslaw, K.S., 2016. Selective environmental stress
622
from sulphur emitted by continental flood basalt eruptions. Nat. Geosci. 9, 77–82.
623
https://doi.org/10.1038/ngeo2588
624
Schoene, B., Eddy, M.P., Samperton, K.M., Keller, C.B., Keller, G., Adatte, T., Khadri, S.F.R., 2019.
625
U-Pb constraints on pulsed eruption of the Deccan Traps across the end-Cretaceous mass
626
extinction. Science (80-. ). 363, 862–866. https://doi.org/10.1126/science.aau2422
627
Schoene, B., Samperton, K.M., Eddy, M.P., Keller, G., Adatte, T., Bowring, S.A., Khadri, S.F.R.,
628
Gertsch, B., 2015. U-Pb geochronology of the Deccan Traps and relation to the end-Cretaceous
629
mass extinction. Science (80-. ). 347, 182–184. https://doi.org/10.1126/science.aaa0118
630
Schulte, P., Alegret, L., Arenillas, I., Arz, J.A., Barton, P.J., Bown, P.R., Bralower, T.J., Christeson,
631
G.L., Claeys, P., Cockell, C.S., Collins, G.S., Deutsch, A., Goldin, T.J., Goto, K., Grajales-
632
Nishimura, J.M., Grieve, R.A.F., Gulick, S.P.S., Johnson, K.R., Kiessling, W., Koeberl, C.,
633
Kring, D.A., MacLeod, K.G., Matsui, T., Melosh, J., Montanari, A., Morgan, J. V., Neal, C.R.,
634
Nichols, D.J., Norris, R.D., Pierazzo, E., Ravizza, G., Rebolledo-Vieyra, M., Reimold, W.U.,
635
Robin, E., Salge, T., Speijer, R.P., Sweet, A.R., Urrutia-Fucugauchi, J., Vajda, V., Whalen,
636
M.T., Willumsen, P.S., 2010. The chicxulub asteroid impact and mass extinction at the
637
cretaceous-paleogene boundary. Science (80-. ). 327, 1214–1218.
638
https://doi.org/10.1126/science.1177265
639
Schutt, D.L., Dueker, K., Yuan, H., 2008. Crust and upper mantle velocity structure of the
640
Yellowstone hot spot and surroundings. J. Geophys. Res. Solid Earth 113, 1–14.
641
https://doi.org/10.1029/2007JB005109
642
Self, S., Schmidt, A., Mather, T.A., 2014. Emplacement characteristics, time scales, and volcanic gas
643
release rates of continental flood basalt eruptions on Earth. Spec. Pap. Geol. Soc. Am. 505, 319–
644
337. https://doi.org/10.1130/2014.2505(16)
645
Seton, M., Müller, R.D., Zahirovic, S., Gaina, C., Torsvik, T., Shephard, G., Talsma, A., Gurnis, M.,
646
Turner, M., Maus, S., Chandler, M., 2012. Global continental and ocean basin reconstructions
647
since 200Ma. Earth-Science Rev. 113, 212–270. https://doi.org/10.1016/j.earscirev.2012.03.002
648
Sprain, C.J., Renne, P.R., Clemens, W.A., Wilson, G.P., 2018. Calibration of chron C29r: New high-
649
precision geochronologic and paleomagnetic constraints from the Hell Creek region, Montana.
650
Bull. Geol. Soc. Am. 130, 1615–1644. https://doi.org/10.1130/B31890.1
651
Sprain, C.J., Renne, P.R., Vanderkluysen, L., Pande, K., Self, S., Mittal, T., 2019. The eruptive tempo
652
of deccan volcanism in relation to the cretaceous-paleogene boundary. Science (80-. ). 363, 866–
653
870. https://doi.org/10.1126/science.aav1446
654
Stachnik, J.C., Dueker, K., Schutt, D.L., Yuan, H., 2008. Imaging Yellowstone plume-lithosphere
655
interactions from inversion of ballistic and diffusive Rayleigh wave dispersion and crustal
656
thickness data. Geochemistry, Geophys. Geosystems 9. https://doi.org/10.1029/2008GC001992
657
Tsekhmistrenko, M., Sigloch, K., Hosseini, K., Barruol, G., 2021. A tree of Indo-African mantle
658
plumes imaged by seismic tomography. Nat. Geosci. 14, 612–619.
659
https://doi.org/10.1038/s41561-021-00762-9
660

Van Ark, E., Lin, J., 2004. Time variation in igneous volume flux of the Hawaii-Emperor hot spot
661
seamount chain. J. Geophys. Res. Solid Earth 109, 1–18. https://doi.org/10.1029/2003JB002949
662
Waite, G.P., Smith, R.B., Allen, R.M., 2006. VP and VS structure of the Yellowstone hot spot from
663
teleseismic tomography: Evidence for an upper mantle plume. J. Geophys. Res. Solid Earth 111,
664
1–21. https://doi.org/10.1029/2005JB003867
665
White, R.S., McKenzie, D., 1995. Mantle plumes and flood basalts. J. Geophys. Res. 100.
666
https://doi.org/10.1029/95jb01585
667
Wignall, P.B., 2001. Large igneous provinces and mass extinctions. Earth Sci. Rev. 53, 1–33.
668
https://doi.org/10.1016/S0012-8252(00)00037-4
669
Wilson, G.P., 2014. Mammalian extinction, survival, and recovery dynamics across the Cretaceous-
670
Paleogene boundary in northeastern Montana, USA. Spec. Pap. Geol. Soc. Am. 503, 365–392.
671
https://doi.org/10.1130/2014.2503(15)
672
Zhang, N., Zhong, S., Leng, W., Li, Z.X., 2010. A model for the evolution of the Earth’s mantle
673
structure since the Early Paleozoic. J. Geophys. Res. Solid Earth 115, 1–22.
674
https://doi.org/10.1029/2009JB006896
675
676
677

678
Figures and Tables
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
Fig. 1. Geology of the Deccan volcanic province (DVP). (a) Map showing the four main sub-
provinces of DVP. The Deccan traps (DTs) rest on Precambrian basement rocks (shown in
various legend patterns). The terrain contains a number of structural zones, such as lineaments
and escarpment (marked as green dashed lines). Blue lines depict the major rivers flowing
across DVP. WGE = Western Ghat Escarpment, EGMB = Eastern Ghat Mobile Belt.
Reconstructed from (Kale et al., 2020) (b) Stratigraphic succession of the DVP (Left column)
and their corresponding cumulative eruption volumes (Right column) along with ages for the
three main subgroups of DVP in the Western Ghats (Renne et al., 2015). The panel shows the
following elements (from left to right): cumulative stratigraphic height, geological time scale
with the KPB indicated by the gray area, timescale of geomagnetic polarity with various
magnetic chrons, and cumulative volume of Deccan lava. It also includes the probabilistic
volumetric eruption rate and the Chicxulub impact time from Schoene et al., 2019 (c) A
thematic geological cross-section of the DTs to illustrate the three major phases and their
corresponding formations (Chenet et al., 2009). Color legends correspond to those used in (b).

698
699
700
701
702
703
704
705
706
707
Fig. 2. Pulsating ascent dynamics of thermochemical plume at mid-mantle transition zone. (a)
Development of successive four pulses (i-iv) from a thermochemical plume in models with
buoyancy number (B) = 0.8, viscosity ratio (μ) = 100 and heat producing element concentration
(cHPE) same as the background mantle. Colors (Crameri et al., 2020) represent the temperature
and dashed yellow lines delineate the pile margin. Insets show the dynamic topography (in km)
corresponding to each pulse. (b) Calculated plots of the pulse volume (i), the pile volume (ii),
and the locations of plume (black) and pile margin (yellow) (iii) during the four pulse events
(denoted in different colors). The volumes are calculated based on the initial volume provided
in Table 1.

708
Fig. 3. Effects of the model parameters on pulse and pile dynamics. Geometry and locations of the
709
pulses generated from a plume head and the pile in different models with varying parameters: (a) B =
710
0.8, μ = 1, and cHPE = 1X; (b) B = 1.2, μ = 1, and cHPE = 1X; (c) B = 0.8, μ = 1, and cHPE = 20X; (d) B =
711
1.2, μ = 1, and cHPE = 20X; (e) B = 0.8, μ = 100, and cHPE = 1X; (f) B = 1.2, μ = 100, and cHPE = 1X; (g)
712
B = 0.8, μ = 100, and cHPE = 20X; (h) B = 1.2, μ = 100, and cHPE = 20X, where X denote cHPE value for
713
the background mantle. Color scale are same as in Fig. 2. Inset of each figure shows the dynamic
714
topography (in km) at the surface for the pulses presented in the respective snapshot.
715
716
717
718
719
720
721
722

723
724
725
726
727
728
729
730
731
Fig. 4. Calculated plots from numerical models of successive pulses for different parametric
values. (a) Variation in the exposed fraction of the core mantle boundary (CMB) for different
model parameters. (b) - (c) Decreasing trends of successive pulse and pile volumes. (d) Varying
plume head locations for successive pulses. The x-axis represents successive pulses, which in
turn reflect progressive time. The symbols stand for the parameter B, and the colors denote μ
and cHPE. Their details are provided in the legend. Also provided are the fields of passive, stable
and unstable piles using dashed curves.

732
733
734
735
736
737
738
739
740
741
Fig. 5. Melt production by partial melting of plume head in the model. (a) Melt localization in
three successive melt pulses (i-iii) at lithosphere-asthenosphere boundary (LAB). They
originate from a single major pulse obtained from the whole mantle model. Colors represent
the temperature and the colored contours represent melt fraction. Black line delineates the
deformed LAB geometry. The slight tilt in the plume axis results from plate movement. The
first two pulses (i-ii) involve intense thermal erosion at the contact between the melt front and
the LAB, resulting in thinning of the thermal boundary layer. The top boundary is deflected to
produce topography during the successive melting events. (b) Calculated plots of melt volume
formed in successive melt pulses. (c) Melt-driven dynamic topography for three successive
pulses. The colors used to represent the pulses in (b) and (c) are shown in the legend.

742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
Fig. 6. African LLSVP and its connection to the Réunion hotspot and the Deccan volcanism.
(a) Global map showing the present-day location of African LLSVP (gray shade) and the
poloidal velocity components at a level 150 km above CMB constructed from Ford and Long,
2015. Strong south-westward velocity can be noticed at the eastern flank of the LLSVP. (b)
Contours of 75% chemical concentration corresponding to a time series, 100 Ma to present day.
The contour plots depict positional changes of African LLSVP through geologic time. The
contours are redrawn from Hassan et al., 2016, 2020 expect that for 66 Ma (dashed contour)
which is interpolated. The figure also shows location of the Tethyan subduction system and
Indian plate (yellow) during the Deccan volcanism at 66 Ma. At this time the western margin
of Indian plate was located directly above the eastern flank of the African LLSVP. The base
map has been produced using S40RTS (Ritsema et al., 2011) depth slice at 2800 km on
SubMachine. (c) Plots of the locations of African LLSVP (solid lines), Réunion plume tail
(dotted lines) and plume head (dashed lines) (i), the rate of southward migration of LLSVP,
and (ii) those calculated from two of our representative models (see text) for each successive
pulsation events.

757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
Fig. 7. A timescale analysis of global LLSVP related volcanic events. Histogram analysis of
the periodic variations of volcanism in Hawai'i (Blue), Réunion (Saffron), and Yellowstone
(Green). Short-term (< 1.5-2 Ma oscillations) and long-term (> 3 Ma oscillations) temporal
variations are distinct in the plots (see discussion).

775
776
Table 1 Physical parameters and their values used for thermochemical modelling
Model Parameters
Reference values
Mantle thickness 
2890 km
Reference density 
3340 kg/
Reference viscosity 
 Pa s
Thermal conductivity 
4.1 W K-1m-1
Specific heat 
1250 J K-1 kg-1
Thermal expansivity 
 K-1
Thermal boundary layer thickness at the CMB ()
100 km
Initial basal layer thickness 
 km
Basal layer density 
3730-3950 kg/
Basal layer viscosity 
 Pa s
Viscosity ratio †

Top temperature 
300 K
Bottom temperature 
3300 K
Reference Temperature 
1600 K
Buoyancy number 

Background Heating rate 
 W/kg
Basal layer heat producing element concentration ()

Initial basal layer volume‡
 km3
Clapeyron slope at 660 km phase transition ()
 Pa/K
Clapeyron slope at 410 km phase transition ()
 Pa/K
† Ratio of viscosity of the basal layer and the ambience
‡ Initial basal layer volume is calculated from the total volume of African LLSVP considering that the eastern flank (corresponds to the initial basal layer) comprises only a
fraction of the total volume.

777
Table 2 Physical parameters and their values used to model partial melting in plumes
Model Parameters
Reference values
Melt density 
2700 kg/
Reference shear viscosity 
 Pa s
Melt viscosity 
10 Pa s
Reference permeability 
 m2
Reference porosity 

Melt weakening factor 

Thermal viscosity exponent 

Thermal expansion coefficient 
 K-1
Solid compressibility 
 Pa-1
Melt compressibility 
 Pa-1
CFL number


Supplementary material for
Periodicity in the Deccan volcanism modulated by plume
perturbations at the mid-mantle transition zone
Dip Ghosh, Joyjeet Sen, and Nibir Mandal*
Department of Geological Sciences, Jadavpur University, Kolkata 700032, India

Fig. S1. Details of the study area in a global perspective. (a) Tomography-Depth slice at 2800 km, showing the
present-day location of the eastern flank of African LLSVP using S40RTS model (data generated using SubMachine).
(b) Satellite bathymetry map of the western Indian Ocean showing the complete Réunion hotspot track (Deccan Traps
to Réunion Island). Aerial extent of the Deccan volcanic province is demarcated in purple within the Indian subcon-
tinent. Crustal age estimates for Réunion plume activity (in Ma) indicate plume positions. Black lines delineate the
plate boundaries. The base map is reproduced from BODC data. www.bodc.ac.uk. The white dashed line represents
the trace of our model section.

Fig. S2. Initial conditions considered for plume model simulations. (a) Initial density profile showing jumps of density
values at the phase transition at 410 km and 660 km boundaries, and steep density increase near CMB due to the pres-
ence of a thermochemical pile. Blue and black lines indicate the maximum and minimum pile density considered in the
modelling. (b) Depth profiles of the initial internal heat-production rate for the ambient mantle and the pile. The pile at
CMB is enriched in heat-producing element (HPE) by up to 20 times relative to the ambient mantle. (c) Initial thermal
structure of the mantle at the onset of convection, characterized by strong thermal boundary layers (TBL) at the upper
200 km (lithosphere) and at 100 km above the CMB. (d) Initial viscosity profile considered in our models. It accounts
for both temperature and depth effects. The pile material is up-to 100 times (blue) more viscous that the ambient man-
tle.

Fig. S3. Evolution of a thermochemical plume in the
reference model (B= 0.8,μ= 100, and cHPE =X).
(a) Piling up of TBL due to forcing by a downwelling
flow in mantle. (b) Growth of a small instability on
the extreme right side of the TBL. (c) Lateral advec-
tion and climb of the instability to the pile crest. (d)
Development of a mature plume from the instability
with increasing buoyancy flux. (e) Perturbation of the
plume head at the mid mantle transition zone to pro-
duce a primary pulse. Note that the pulse in the upper
mantle deflect to the right under the influence of plate
velocity.

Fig. S4. Horizontal variation of the physico-chemical properties in four successive plume pulses produced at the mid-
mantle transition zone. The graphical plots correspond to a depth of 400 km. (a) Variations of internal heat production
showing a maximum peak value for the first pulse (yellow curve). Note that the next pulses consistently reduce their
peak values. The secondary pulse (Brown) contains considerable amount of HPE, as reflected from its high internal
heating production, which weakens with the tertiary pulses (blue and green curves). Their reducing trend indicates
decrease in HPE concentration due to less entrainment of pile materials by the plume. (b) Density profiles. The first
pulse show the highest negative density anomaly reflecting strong buoyancy head. The density anomalies significantly
weaken in the secondary and tertiary pulses. (c)-(d) Excess temperature and viscosity profiles for the pulses, the pat-
terns of which agree with the HPE concentration and the density profiles in (a) and (b), respectively.

Fig. S5. Model simulations of the melt transport processes. (a) Model domain (inset) chosen within the plume model
(left panel). (b) Depth dependent variations of the physical parameters: flow velocity, excess temperature, melt-fraction
and viscosity at the time of melt initiation (i-iv) and at the onset of thermal erosion of the lithosphere by the partial
melts (v-viii).

Fig. S6. Time evolution of models
showing melt production by partial
melting. (a) Melt initiation at the crest
of the plume head. (b)-(c) Progres-
sively increasing melt fraction as the
plume head interacts with the LAB.