On the plate velocity scaling break Authors: Davide Zaccagnino, Marco Cuffaro, Eleonora Ficini, Carlo Doglioni Published: 2025 ABSTRACT AbstractThis research presents the discovery of a kink in the frequency-size distribution of plate angular velocities in different reference frames, a feature not predicted by current physical models of plate tectonics. We analyze global plate motions to investigate the origin of this scaling break and find that larger plates systematically exhibit slower angular velocities than smaller ones. Moreover, we explore the relationship between plate velocities and the physical properties of the lithospheric basal layers, revealing that faster-moving plates correlate with more homogeneous lid rheology and structure, as inferred from reduced P-wave velocity variability. A significant negative correlation between angular velocity and latitude is observed with a persistent west-ward residual drift suggesting the long-term influence of large-scale forcings determining asymmetric mantle flows with respect to the outer layers of the planet. Based on these findings, we propose two possible reasons to explain the observed plate velocity scaling break. The first one suggests different dominant driving mechanisms for large and small plates resulting in a two-tier plate tectonics with a threshold in the range 10 km; an alternative explanation is the presence of a substantial bias in currently assumed hotspot reference frames with respect to the true mantle-reference plate motions to be recovered through a net westward rotation of °/Myr.Keywords: Plate motions, Microplates, Plate velocity, Geodynamic forcesSubject terms: Planetary science, Solid Earth sciences Introduction Plate motions and the spectrum of plate sizes Plate tectonics describes the lithosphere as a mosaic of rigid components moving relative to one another due to the long-term action of several endogenous and exogenous planetary forces1. Since the dawn of geodynamics and large-scale geodesy, the number of recognized plates has grown from a dozen major segments to over than fifty and even more, if sub-plates with still unidentified boundaries are included2. Their area spans scales from the Pacific Plate ( km) to fragments orders of magnitude smaller. This continuum of sizes challenges traditional classifications: while the largest plates dominate Earth’s surface area, smaller plates - usually called microplates - account for a scarce portion of the lithosphere; nevertheless, they play an important geodynamic role and often host quite active tectonic settings. So far, no consensus exists on what physically distinguishes “microplates” from other plates, nor whether their dynamics differ fundamentally3. This ambiguity reflects a deeper open question in plate tectonics: “How does plate size influence kinematics and coupling to the mantle?”. Observations suggest that smaller plates rotate faster on average4, but the mechanisms behind this scaling, as well as its geodynamic implications, e.g., its impact on mantle-lithosphere interaction, remain unresolved. Mantle-lithosphere coupling and the size of plates Relative plate kinematics quantifies the displacement of one plate relative to another along a shared boundary, typically treated as fixed1,5,6, or relative to the Pacific plate7. Angular velocities and rotation poles are directly derived from geological, geophysical, or geodetic data5,8 and are essential for characterizing plate motion, deformation mechanisms, and associated geological phenomena such as mountain building, seismic activity, and volcanism. Plate motions with respect to the mantle can be described in two different kinds of reference frames: the first one is the No-Net-Rotation (NNR)9–11, which assumes zero global torque on the lithosphere, providing a kinematically stable framework for relative plate motions; the second one is the hotspots, which are stated fixed relative to each other (e.g.,4,12–14), or defined as moving ones15. Mantle-reference plate and microplate motions describe lithospheric movement relative to the mantle, providing a global perspective on the driving forces of plate tectonics. While these motions cannot be directly observed, they are inferred by integrating data from relative kinematic frameworks. Such “absolute” reference frames mostly use hotspots to estimate plate motions relative to the deep mantle. While hotspot tracks (e.g., the Hawaii-Emperor Pacific Islands) offer some constraints, their effective reliability as a reference with respect to avowed fixed deep layers of the Earth is debated because it has been shown that they tend to move with respect to each other. Indeed, research works often assume hotspots are either locked relative to one another (e.g.,12–14) or incorporate their motion15. Conversely, a no-net-torque condition between the lithosphere and asthenosphere16 is typically enforced on no-net-rotation (NNR) frames to compute plate velocities. This approach assumes uniform lithosphere-asthenosphere coupling, neglecting lateral variations. Conversely, the hotspot frame accounts for upper-mantle decoupling, making it a more suitable reference for plate and microplate motion reconstructions. Recent studies17 support the fixed-hotspot approximation as the most reliable method for estimating mantle-reference kinematics. Since most hotspot tracks are located near plate boundaries, which themselves move relative to the mantle-fixed frame, intraplate Pacific hotspots are preferred for defining present-day plate displacements in the mantle-reference frame12. Then, the choice of frames and the level of trust we put on them can impact geodynamic interpretations: NNR frames, as well as those referred to specific plates as fixed with respect to the mantle, can reveal inter-plate dynamics, but mask mantle-driven or -mediated motions. This is even more important since the hotspot reference frames suggest a systematic westward net rotation of plates18–20, hinting at planetary-scale mantle flow decoupled from surface rotation. This is the reason why recent works, e.g.,21–23, favor hybrid approaches to take into account of different aspects; nonetheless, discrepancies persist, particularly regarding smaller plates, whose motions are likely more sensitive to local mantle heterogeneity than happens for larger ones. Indeed, the differential motion of tectonic plates relative to the mantle reflects heterogeneity in lithosphere-asthenosphere coupling24,25, with plate size emerging as a critical factor in this dynamic relationship. In our work, we mostly focus on the Deep Pacific Hotspot reference frame12 and utilize information retrieved from the LITHO1.0 model for the properties of the lithosphere26. A visual summary is provided in Figs. 1 and 2. Seismic and geodynamic evidence indicates that faster-moving plates typically overlie regions of reduced uppermost mantle viscosity and more homogeneous seismic structure, e.g.27, suggesting enhanced basal decoupling. Such detachment is likely to occur in the weakest layer of the asthenosphere, mostly in its upper part in a region called lithosphere-asthenosphere boundary (LAB)28,29 and embodied within a shallow low-velocity (and viscosity) zone (LVZ) mostly located around 120-150 km at depth. The efficiency of coupling depends on several factors; among them, the type of plate, i.e., if mainly composed of continental or oceanic lithosphere, and also the size of the plate itself30. Indeed, it appears to be scale-dependent: larger plates cover broader mantle domains, averaging out local rheological contrasts, while smaller plates respond more sensitively to regional basal viscosity anomalies. The transition between these regimes may be modulated by lithospheric thickness variations, with thinner and uniform oceanic lithosphere facilitating localized shear zones with respect to thicker lithosphere due to orogens. Notably, latitudinal velocity trends persist across plate sizes, suggesting planetary-scale influences superimposed on this size-dependent coupling behavior. These observations collectively highlight how plate dimensions interact with mantle heterogeneity to govern tectonic fragmentation patterns, though the precise scaling relationships remain unresolved. Global map of lithosphere thickness (a) and P-waves velocity at the depth of the bottom lid according to26 (b). a Plate velocities with respect to the hotspot reference frame obtained adding the Pacific plate angular vector from12 to the relative plate motions in the database by7. b Plate boundaries by7. Open questions and research aim On the basis of the evidence described above, it is clear that the area of plates is a key parameter in order to understand how the machine of plate tectonics works. Nevertheless, although research has been already done on the scaling properties of the surface of plates31,32, no systematic study exists in the literature on this important feature of the Earth’s lithosphere in relation with plate kinematic and underlying layers dynamics. By integrating plate kinematics, upper mantle properties, and reference frame theory, we aim to better understand the mechanisms behind how plate size and mantle coupling jointly govern the motions of the lithosphere - a step toward unifying the physics of plates over different scales. This study analyzes global plate motions across reference frames, with particular attention to the deep hotspot reference12, to test whether the size-velocity scaling reflects universal dynamics or frame-dependent artifacts. Jointly, we evaluate how lid heterogeneity (inferred from global seismic tomography models26) correlates with plate motions. Our approach also enables us to overcome the microplate concept ambiguity: can “microplates” be physically defined, or do they simply occupy the small-plate end of a continuum ranging from lithospheric fragments to the largest Pacific one? Indeed, although no strict size criterion exists, microplates are typically defined as lithospheric fragments significantly smaller than major tectonic plates (e.g., Pacific, Eurasia, African) but large enough to behave as coherent units bounded by distinct plate boundaries. These boundaries may be divergent (spreading ridges), convergent (subduction or collision zones), transform faults, or a combination thereof. Their defining feature is kinematic independence; indeed, their motion is not entirely controlled by adjacent major plates. Empirically, plates smaller than km could be classified as microplates but no physically justified reason exists. Microplates originate through various geodynamic processes, including lithospheric fragmentation between major plates, rift zone jumps, or backarc basin formation in subduction settings (e.g., Adriatic, Aegean, Anatolian and Iberian microplates). Also, most microplates cluster at mid- to low-latitude regions where plate interactions are diffuse and deformation zones are broader and more dynamic8,33. Then, it is quite difficult to collect consensus on how plates and microplates can be defined in an objective and physically sound way. If confirmed in the future, our work may also introduce a new criterion to reassess the need for special “microplate” classifications in light of physical mechanisms affecting plates kinematics. Results The kinematic behavior of lithospheric plates, ranging from major plates to smaller fragments, represents a fundamental but incompletely understood aspect of plate tectonics. While relative motions between these fragments are well-documented, the underlying controls on their size distribution and velocity variations remain a subject of investigation. The tectonic system emerges from the complex interaction between lithospheric motion and mantle dynamics, where plates move both relative to each other and with respect to the underlying mantle. Current understanding suggests that plate dimensions and velocities may be governed by two potentially complementary mechanisms: boundary forces including slab pull, ridge push, and trench suction, and basal drags resulting from variable lithosphere-asthenosphere coupling. Observations present challenges to purely boundary-force driven models; indeed, several major plates (e.g., Eurasia, Africa) exhibit significant motion without being attached to subducting slabs34. This has led to alternative hypotheses emphasizing the role of heterogeneous basal shear. Upper mantle properties and plate angular velocity Here, we present our results starting from the topological properties of the bottom lid (the layer at the boundary between the lithosphere and asthenosphere, where most of the mechanical decoupling is expected to occur hosting the lowest viscosity). The lithosphere thickness and its cumulative distribution are shown in Figure S1, displaying variations according to different tectonic regions. Two peaks are observed corresponding to the oceanic lithosphere (depth km) and close to 250 km due to the presence of thicker orogens. We systematically calculate the distribution of seismic P-wave velocity below each plate according to the definition by Bird7 at the bottom lid depth. Details are described in the section devoted to methods. Then, we correlate its value with the angular velocity of plates. Angular velocity represents the most appropriate physical quantity for this type of analysis. Plate motions on the Earth’s surface are described as rigid-body rotations about Euler poles, and angular velocity therefore provides the most direct and general description of this motion. On the other side, surface velocity depends on the distance from the rotation pole and cannot be consistently compared among plates with different geometries. For this reason, we chose to analyze plate behavior in terms of angular velocities, which more coherently reflect the global kinematics of the lithosphere. In agreement with this, Figure S2 in the Supplementary Materials presents plate angular velocities in the Deep Pacific Hotspot frame4 with respect to their Euler poles plotted against the minimum and maximum values of P-wave velocity at the bottom lid beneath each plate, revealing correlations between these parameters. Especially, the angular velocity is positively correlated to the minimum value of seismic wave velocities and negatively related to the largest one, in agreement with the idea that the faster moving plates are those with the stronger basal decoupling, i.e., the lowest viscosity in the low velocity zone. A deeper insight is provided by the connection between the angular velocity of plates and the heterogeneity of seismic velocity and bottom lid depth: more homogeneous values are found to be positively correlated with faster motions. The relationship between angular velocity and P-wave velocity variability is demonstrated in Fig. 3, which includes both raw and area-rescaled standard deviations of P-wave velocity, along with correlations to the variability of bottom lid depth. The surface rescaling is performed in order to take into account of the effect of stochastic fluctuations. Indeed, the standard deviation of the wave velocity and depth decreases as of the number of measurements used to calculate its value which is proportional to the square root of the size of each plate. We also note that Fig. 3b supports the idea that upper mantle rheology and lithospheric thickness are two sides of the same coin, being almost perfectly linearly dependent once the size rescaling is performed (). While these correlations between P-wave velocity heterogeneity and plate motions provide useful insights, we note that their interpretation still remains qualitative. The use of P-wave velocity anomalies is intended as a first-order proxy for lateral variations in mantle rheology and not as an accurate measure of basal coupling, which depends on the integrated viscosity structure of the asthenosphere. The primary aim of this analysis is to examine whether systematic patterns exist between upper-mantle heterogeneity and plate behavior and particularly in the context of global scale observations discussed in the next paragraphs and modeled in the section devoted to discussions. The distribution of Euler pole orientations in the Deep Pacific Hotspot reference frame appears in Figure S3, showing asymmetric patterns of plate motion directions. Together with the results in Fig. 4, which displays plate angular velocities in the Pacific hotspot frame4 versus average plate latitude, with dot sizes representing plate areas and colors indicating average bottom lid depth according to the provided color scale, it is an evidence of western-plate motion polarization with respect to the lower layers associated with a latitude dependent large-scale forcing in agreement with35–37. Anyway, we stress that the spatial distribution of tectonic plates shown in Fig. 4, revealing a concentration of plates at lower latitudes, may reflect, in part, geometric constraints: high-latitude regions offer less surface area for plate formation and detection compared to equatorial zones. Moreover, the identification of smaller plates near the poles is limited by both reduced surface area and typically slower relative motions, which complicate kinematic resolution. While these factors introduce a potential bias in the observed latitudinal distribution, the persistent correlation between angular velocity and latitude suggests an underlying geodynamic significance beyond spurious geometric effects. a Plate angular velocities in the Deep Pacific Hotspot reference frame4 with respect to their Euler poles vs the variability of P-wave velocity (standard deviation and standard deviation rescaled with the plate area) at the bottom lid below each tectonic plate. b, c Correlation between the variability of P-wave velocity and the standard deviation of the bottom lid depth rescaled with respect to the plate area. Plate angular velocities in the Deep Pacific Hotspot reference frame with respect to their Euler poles vs average plate latitude. The size of dots represents the plate area, while the color stands for the average depth of the bottom lid according to the color bar on the right. Fast rotating plates are smaller and mostly located at low latitudes. Lithosphere self-organization and its connection with the inner Earth Seismic and geodetic evidence indicates that plate velocities correlate with asthenospheric viscosity structure, particularly within the low-velocity zone (LVZ). The previous paragraph is consistent with this view. The Pacific plate, exhibiting the fastest absolute motion, overlies a region of exceptionally low seismic velocities interpreted as reduced viscosity. Conversely, slower-moving plates appear associated with more viscous mantle regions. This suggests that lateral viscosity variations in the LVZ may modulate plate motions through differential basal coupling. Then, the size distribution of plates might reflect hierarchical mantle heterogeneity. Major plates could correspond to first-order viscosity domains in the asthenosphere, while smaller fragments (microplates) might form where secondary viscosity anomalies create localized strengthened decoupling. Alternative formation mechanisms support the idea of stress concentration at plate boundary or fragmentation of larger plates due to inherited lithospheric weaknesses38. For sure, the energy budget of plate tectonics likely involves multiple driving mechanisms. While boundary forces contribute locally, global-scale processes may provide the primary energy source. Such mechanisms, ranging from small-scale polarized mantle convection generating large-scale flow patterns, long-standing tidal forces and integration of thermal-mechanical-chemical heterogeneities are likely to empower lateral viscosity variations together with more conventionally recognized contributions. They are likely not mutually exclusive, and their relative importance may vary spatially. The observed westward drift of the lithosphere and the planetary-scale distribution of seismic energy release suggest that plate motions integrate both global driving forces and local effects. Studying how the size of plates is related to their velocity in suitable reference systems can provide useful information about the physical connection between rigid lithospheric structures and underlying layers. In this section, we investigate the size distribution of tectonic plate and their angular velocity. In Fig. S4 the complementary cumulative distribution (not normalized) on a log-log plot is displayed, with the power-law trend and deviations at extreme size ranges evident in the plot, as in7. The deviation from the power law observed on the left, corresponding to very small plates, suggests that below a size 1/1000 the Pacific plate the detection of new microplates becomes difficult and we still have to include unknown kinematically independent lithospheric fragments in the catalog of microplates. Indeed, roughly 200 plates with angular area larger than are expected to exist, while only 50-60 are listed. On the right of the distribution, the deviation from the power law is due to a finite size effect: the surface of the Earth is limited and the area of the largest fragment of the lithosphere is strongly depleted from the interactions with its neighbors. While this feature is well known by at least two decades, here we present an unprecedented result: while plate surfaces are power-law distributed, the angular plate velocities associated to the same plate database are not power low distributed. A double power-law scaling of angular plate velocity is observed and shown in Fig. 5, where plates are separated into distinct regimes based on their rotation speed, with the transition occurring at 1.1-1.4°/Myr in different reference frames. The different systems shown in Fig. 5 are the fixed Pacific plate from7, the Deep Pacific Hotspot frames from4, the obtained NNR frame as described in the methods, and the shallow hotspot framework39, where the depth of the hotspot source was chosen at 225 km, following40. It is worth noticing that, although the NNR frame neglects lateral viscosity variations in the asthenosphere, implicitly treating mantle coupling as uniform and resulting in a lack of physical justification, we decided to include it in the kinematic description because the plate velocity kink is also emphasized in that reference frame, and it becomes useful for comparison (Fig. 5). Moreover, a deeper look into the propagation of uncertainties from the geodetic datasets obtained by41 was also explored (Fig. 5b), using their proposed NNR geodetic framework to compute plate kinematic models with respect the fixed Pacific plate, and the deep and shallow (225 km) hotspot frames (Fig. 5b). a Double power-law scaling of angular plate rotation in the Deep Pacific Hotspot reference system integrating geodetic plate kinematics from8. The scaling exponent of plates rotating faster than 1°/Myr is roughly 1.0, while slower plates have a completely different regime featured by a scaling exponent of about 0.3. b The power-law scaling breaking is observed at the same angular velocity value ( 0.9° to 1.5°/Myr) in all the different considered reference systems. The blue points represent the empirical scaling in the case of the Deep Pacific Hotspot Reference Frame as in (a), the yellow ones the Fixed Pacific Frame, the violet dots stand for the Shallow Hotspot Frame - we choose 225 km as a reliable example - and red points for the NNR reference system. See methods for detailed definition. c The same of (b) using data from4. The picture is finally completed by Fig. 6, which directly shows the relationship between angular plate velocity and plate area using Bird’s dataset7, displaying an overall linear decreasing trend in log-log plot, although quite scattered. In Fig. 6, we did not show the analysis using the Kremeer’s dataset because it employs a different definition of plate area with respect to Bird’s one; moreover, results in Fig. 6 and Table 1 show that the scaling properties of the angular velocities retrieved from the two models are statistically compatible. Therefore, even plate area and angular velocity are power law dependent on each other. These results strengthen the idea of an extensive-property control on plate tectonics where small and large plates play different roles in organizing the same interacting system. Distribution of angular plate velocity in the Deep Pacific Hotspot reference frame4 as a function of the area. A clear decreasing trend is found compatible with a power law; although scattered, the best fit is featured by a scaling exponent equal to – 1.1. Table 1Double power law fitting results for tectonic plate angular velocity distributions in different plate motion frames and models from8 (upper part) and from4 (lower lines). The break point separates fast plates (right branch) from slow plates (left branch), with corresponding scaling exponents for each regime.ModelBreak Velocity (°/Myr)Deep Pacific Hotspot1.48 ± 0.020.39 ± 0.030.99 ± 0.02No Net Rotation1.84 ± 0.030.43 ± 0.040.94 ± 0.03Pacific Plate0.85 ± 0.060.24 ± 0.071.05 ± 0.03Shallow Hotspots (225 km)1.55 ± 0.040.37 ± 0.051.01 ± 0.02Average (Geological)1.43 ± 0.210.36 ± 0.051.00 ± 0.05Deep Pacific Hotspot1.12 ± 0.020.35 ± 0.040.97 ± 0.03No-Net Rotation1.52 ± 0.030.49 ± 0.041.00 ± 0.01Pacific Plate0.85 ± 0.040.15 ± 0.081.04 ± 0.02Shallow Hotspot (225 km)1.11 ± 0.030.43 ± 0.041.00 ± 0.03Average (Geodetic)1.15 ± 0.280.36 ± 0.061.00 ± 0.04Total Average1.29 ± 0.220.36 ± 0.041.00 ± 0.03 Double power law fitting results for tectonic plate angular velocity distributions in different plate motion frames and models from8 (upper part) and from4 (lower lines). The break point separates fast plates (right branch) from slow plates (left branch), with corresponding scaling exponents for each regime. All figures and the output of the double power-law analysis of the plate velocity scaling in Table 1 collectively present the relationships between plate motions, physical properties at the lithosphere base, and geometric characteristics of tectonic plates. They show systematic variations in lithospheric thickness, seismic velocity patterns, latitudinal distributions, and scaling behaviors of plate sizes and angular velocities. The data demonstrate correlations between plate rotation rates and various underlying structural parameters without implying specific physical mechanisms. In the next section, we compare the observed scaling relationships to theoretical predictions based on crustal fragmentation models and lithosphere-mantle mechanical interaction. Discussion Why is the lithosphere moving with fragments of different dimensions? Why plates or microplates have different velocities? What is the primary control on plate velocity gradients: heterogeneity in the driving forces or variability in lithosphere-asthenosphere decoupling? The fastest-decoupled plate is the Pacific, which is characterized by the presently known lowest viscosity within the top asthenosphere, i.e., the low-velocity zone (LVZ). The existence of a given plate could be linked with the presence of a roughly homogeneous underlying LVZ viscosity, different from the surrounding mantle. The major plates are associated with several microplates: why? Our results, in agreement with previous research, show that the dynamics of tectonic plates exhibits fundamental relationships between their kinematic behavior and physical properties that reflect underlying geodynamic processes. This study investigates five key characteristics that show systematic connections with plate motions represented by plate angular velocity: the depth of the lithosphere-asthenosphere boundary (LAB, that we associate to the location of the bottom lid as in26), seismic P-wave velocities beneath plates, spatial variability of these P-wave velocities, plate latitudes and Euler poles, and variability in LAB depth and P-waves velocity. These properties collectively provide insights of how lithospheric structure interacts with mantle flow to generate observed plate motions. The distribution of plate areas follows a power law, consistent with scale-invariant fragmentation processes31, while their angular velocities display a more complex double power law distribution. This peculiar distribution suggests distinct dominant dynamical regimes affecting plate motions. Smaller plates ( km) typically move faster, with their velocities showing stronger dependence on area (), while larger plates ( km) exhibit slower, less area-dependent motions. This transition in scaling behavior correlates with measurable changes in lithospheric-asthenospheric properties. Beneath faster-moving plates, the lithosphere-asthenosphere system shows remarkable homogeneity. The depth of the LAB varies less significantly and the result is even reinforced once the variability is rescaled considering the plate area, with standard deviations below 10 km compared to over 50 km up to 100 km beneath slow plates. Similarly, P-wave velocity anomalies beneath fast plates display reduced spatial variability, suggesting more uniform mantle rheology. These observations suggest that mechanical decoupling occurs more efficiently when the mantle transition zone is structurally homogeneous, allowing less restricted plate motion (See Figure S2 in the Supplementary Materials and 3). Anyway, it is challenging to assert whether such features are connected by a causal mechanism or are rather produced by a feedback mechanism, i.e., correlation generated by further hidden dynamics. Latitudinal patterns additionally constrain possible driving mechanisms (Figure S3 and 4): plates at higher absolute latitudes tend to move more slowly, with a correlation coefficient between angular velocity and latitude. This relationship persists at different plate sizes, suggesting a global influence possibly related to polar wander or large-scale mantle flow patterns42,43. Additionally, a systematic westward net rotation component appears in all plate motions, suggesting that top-bottom drivers or tidal effects may break rotational symmetry44,45. The variability in LAB depth seems to be diagnostic, showing an inverse relationship with plate velocity that is robust over different plate sizes. Fast plates not only have more homogeneous LAB depths, but also tend to have shallower average lithospheric roots. This pattern agrees with theoretical expectations for plates that are effectively decoupled from the underlying mantle, where reduced basal mantle drag would permit faster motion with respect to the adjoining plates. Conversely, the thick, variable lithosphere beneath slow plates suggests stronger coupling to heterogeneous mantle structures. Then, such empirical relationships between plate motions and their physical properties provide useful constraints for geodynamic models. The systematic differences between small/fast and large/slow plates may essentially be interpreted in two alternative ways (or a combination of them). They may point to fundamentally distinct extreme dynamical regimes - one dominated by boundary forces and local interactions, the other by mantle coupling. The transition between these regimes manifests not only in the scaling of angular velocities but also in measurable lithospheric characteristics, offering multiple observational handles onto the underlying physics. On the other hand, our results may suggest that the reference systems we deal with routinely are essentially biased being poorly informative about the actual motion of the lithosphere with respect to the underlying deeper layers of the mantle assumed to be fixed. Hereafter, we try to provide reasonable explanations for the scaling exponents we found in the previous sections and displayed in Figs. 5 and 6. The observed distributions of tectonic plate areas and angular velocities reveal scale-invariant behavior with distinct dynamical regimes. Plate areas A follow a power-law distribution consistent with fragmentation processes. It is interesting to notice that, since , either the average and the variance of the distribution of plate sizes exist, being controlled by the largest one. In contrast, angular velocities exhibit a double power law: suggesting two distinct physical regimes. This duality mirrors the scaling between and A, where with , while a double scaling would be expected, but it is not observed. Hereafter, we provide an explanation for this value of γ and a discussion on possible physical mechanisms that could theoretically affect it depending on plates' area. Such an exponent emerges from torque balance between driving forces and viscous drag. The driving torque scales with plate boundary area , while the basal drag torque derives from integrating viscous stress over the plate area. The shear stress at distance r from the rotation axis is , where h is the asthenosphere thickness. The infinitesimal drag torque is giving total drag torque . Considering a characteristic plate size , the integral scales as , yielding . Balancing torques we get , which returns the scaling The observed indicates slight deviations from this ideal case, but uncertainties are large and the present theoretical prediction provides a possible explanation of the exponent without further hypotheses. Now, we would like to explain the dynamics of small plates. We model them as rigid disks of radius R () driven by boundary stress from larger neighbors, the torque scales as: Our “flat disk” hypothesis is reasonable since we only deal with plates smaller than km, hence, much smaller than the Earth’s surface. We assume a constant stress and a moment of inertia , where stands for the lithospheric surface density. Indeed, the moment of inertia of a rigid disk rotating about its central axis is given by , where m is the disk mass, that can be written as , so that . Then, the angular velocity, under the hypothesis that the torque is constant and we are looking at a steady state plate dynamics, the relationship between angular velocity and torque reads matching the observed with (Fig. 5a, Table 1) if the plate area scaling exponent is 1. The observed shallower area distribution, assumed to be valid for microplates but possibly underestimated, suggests the angular velocity-area scaling exponent be larger than 1, potentially consistent with large data scatter and missing tiny plates (Fig. 6) and potentially explicable with second order effects such as complex plate geometry. This implies small plates behave like a gear system, where kinematic coupling enforces at boundaries, and neighbor-driven torques dominate inertia. If basal drag provides a resistance torque , where is the basal viscosity, the equilibrium condition yields: which is still inconsistent with observations. Thus, viscous drag cannot explain the steep scaling unless stress itself depends on plate size. This possibility could be feasible, but it requires detailed investigations and it is definitely beyond the goal of the present discussion. Anyway, this discrepancy suggests that inter-plate interactions, not basal friction, may govern small-plate dynamics. For large plates, the shallower scaling may imply different physics. Here, mantle drag and planetary forcing likely dominate, with torques scaling sub-linearly with A. If a basal drag force , being D the mantle thickness where shear occurs, balances driving forces , then hence independent of plate size. Then, the observed may arise from a weakly decreasing with A (e.g., being the plate decoupled by the weakest interfaces within its heterogeneous mantle) or correspond to a crossover from the small-plate scaling toward a flattened distribution at low angular velocities. So, the double power law in may thus reflect a different contribution of nearest-neighbor-driven rotation with stress governed by local plate-plate interactions for small fast-moving plates and global planetary forces with weaker or null size dependence acting on most of the lithosphere. Our alternative explanation invokes reference frame effects: if the hotspot frame rotates at relative to the true mantle frame, the corrected velocities could collapse the double power law into a single one. Testing this requires minimizing the residuals from the power-law scaling under different reference frames with rotations. We perform this analysis using the Deep Pacific Hotspot as a starting point; then, we applied the same algorithm as in40 to add a uniform angular velocity component . This addition actually corresponds to locate an average thin ultra-weak decollement layer within the asthenosphere with shallower depth as the added angular velocity increases. The model which best reproduces the single power law gives corresponding to a net westward angular rotation velocity with respect to the mantle reference frame °/Myr and an average basal hotspots depth at about 185 km beneath the Earth’s surface. This result is compatible with previous research works posing the magma source of hotspots within the astenosphere, e.g.,46. °/Myr assumes the current westward net rotation is °/Myr compatible with20,37, while °/Myr is based on the hypothesis that °/Myr12 is the best model for the average global westward lithosphere rotation. See Fig. 7. These calculations are based on the hypothesis that the intra-asthenospheric decoupling occurs approximately as a linear shear flow reflecting our ignorance about physical processes focusing it within thin layers (e.g., exponential channeling). Analysis of the angular plate velocity4 scaling using the Deep Pacific Hotspot as a starting point and modified adding different westward velocities as in the shallow hotspot reference system40 with average decollement layer varying from 125 to 250 km below the Earth’s surface41. The best model reproducing the power law returns corresponding to a net westward angular rotation velocity °/Myr (according to the assumptions described in the main text) and an average basal hotspots depth at about 185 km. Supporting evidence comes from correlations between and physical properties since fast plates (with small area) also exhibit lower P-wave velocity variance, suggesting homogeneous mantle coupling. At the same time, slow plates (large area) correlate with higher latitudes, possibly reflecting polar wander or large-scale flow. In summary, the interplay between plate size, driving mechanisms, and interactions generates the observed scaling laws. Small plates follow due to neighbor-driven rotation, while large plates exhibit weaker scaling from mantle-coupled dynamics. We infer that large plates exist because they lay over mantle zones with viscosity values higher or lower with respect to the adjacent major bigger plates, while the motion of smaller plates may be almost unaffected by the properties of the underlying mantle and produced by the interaction with the surrounding larger plates. In the case of a perfectly homogeneous LVZ viscosity and lithosphere/asthenosphere decoupling, we could virtually expect the entire lithosphere to rotate relative to the mantle as a single outer shell, without fragmentation into separate plates. However, mantle tomography and magmatic geochemistry illuminate a heterogeneous mantle both vertically and horizontally. Therefore, lateral variation in the geochemical composition, temperature, and fluid content can easily generate viscosity variations in the LVZ, determining the variable decoupling and different velocities among larger plates. A hierarchy in the viscosity within the LVZ could help justify the nature of major plates related to the first-order viscosity values in the asthenosphere, while microplates could be associated with the occurrence of second-order local interactions, i.e., marking the self-organizing nature of plate tectonics. They may, in turn, promote second-order mantle heterogeneities over geological times in more restricted upper mantle zones. It is important to emphasize that the proposed “two-tier” framework should be interpreted as a conceptual model highlighting a dominant regime change over spatial scales in global geodynamics, rather than a physical dichotomy. The transition in scaling exponents likely reflects a continuous change in the balance of driving and resisting forces with plate scale, where the relative contributions of mechanisms vary gradually. In addition, the correlations identified between plate velocity, size, and mantle properties indicate a strong mutual influence, but do not establish unambiguous causality. The observed patterns may arise from complex, feedback-driven interactions between the lithosphere and the underlying mantle. Finally, the empirical scaling laws presented in this study, particularly the double power-law distribution of angular velocities, are statistically robust and found in multiple plate motion models and reference frames. It is important to note that our analysis explicitly incorporated the uncertainties in plate angular velocities (as shown in Fig. 5), and the observed scaling break remains a statistically significant feature within these error bounds. However, we acknowledge that the theoretical framework used to explain the transition between the two identified scaling regimes may require further development. The hypotheses proposed (e.g., a shift in dominant driving mechanisms or a reference frame bias) are plausible interpretations consistent with the data, but they are not definitive; better, they are proposals based on first order analytical physical analysis which requires to be confirmed by simulations and further investigations. Future work should focus on developing quantitative geodynamic models that can physically connect the observed velocity “kink” to specific mechanical processes, such as changes in the lithosphere-asthenosphere coupling efficiency or the force balance at a given plate scale. While the statistical evidence for the scaling of small plates is significant, it is important to consider potential influences from data resolution limits. The identification and kinematic characterization of very small plates remain challenging. The deviation from the power-law trend for the smallest plate sizes (Figure S4) likely reflects, in part, an observational limit where current geodetic and geological networks cannot resolve all kinematically independent lithospheric fragments. This does not invalidate the observed scaling break in angular velocities, which is defined by plates well above this resolution threshold, but it suggests that the full spectrum of microplate dynamics may be richer than currently documented. Future improvements in global geodetic coverage will be crucial to refine the lower end of the plate size-velocity distribution and to test the universality of the scaling laws proposed here. Conclusions Our findings reveal an apparent dichotomy in plate dynamics, where the observed double power law for angular velocities ( for slow large plates and for fast small plates) emerges from two distinct extreme physical regimes and with an average behavior arising from a competition of viscous basal drag and boundary effects compatible with the observed γ = 0.9. The steep scaling of small plates may be consistent with boundary-driven rotation, where torque balance predicts neighbor-dominated motion when stress remains size-independent, while their correlation with homogeneous mantle properties suggests decoupling from asthenospheric heterogeneity. In contrast, large plates follow shallower scaling, reflecting mantle coupling where planetary forces saturate with plate size, consistent with their systematic westward drift and latitudinal dependence. This bimodality implies that current geodynamic models must simultaneously incorporate both local boundary interactions (dominating small plate dynamics) and global mantle flow patterns (governing large plates), with the reference frame dependence suggesting additional deep-mantle contributions. The scaling break thus represents an important constraint for unified theories of plate-mantle coupling, requiring explicit treatment of both plate fragmentation processes and large-scale mantle heterogeneity. This integrated perspective emphasizes that plate tectonics emerges from the interaction between lithospheric strength heterogeneities and mantle flow variations, with plate dimensions and velocities reflecting the scale-dependent nature of this coupling. Alternatively, our work would imply that currently utilized reference frames are biased and we should look for the one removing the double-power law allowing the system to recover a possible plate velocity scaling law for all the scales corresponding to a self-organized plate tectonics. This would imply a “true” plate motion reference frame with respect to the deep mantle with net westward rotation °/Myr. Methods We utilize the digital plate and microplate boundaries and the angular vectors proposed by7, which have been incorporated into the intraplate fixed Deep Pacific hotspot frame using methods proposed by4. These correspond to add the Pacific plate angular vector12 to the relative plate motions in the database obtained by7 (Fig. 2b). Moreover, using methods and plate geometrical factors reported in4, the resulting net-rotation of the lithosphere corresponds to a top-down right-handed rotation of rate 0.429°/Ma about an Euler pole at 55.31°S, 69.33°E, and describes the NNR framework rotation with respect to the hotspot frame9, Thus, we subtract the obtained net-rotation angular velocity, , to the plate angular vectors to obtain the plate angular velocities in the NNR frame (Fig. 2a). Specifically, we analyzed plate motions across four distinct reference frames: No-Net-Rotation (NNR) Frame defined by , where and are the angular velocity and area of plate , with N the number of tectonic plates considered to compose the Earth’s lithosphere9. Plate velocities are computed by removing the net lithospheric rotation component : 11 where 4. No-Net-Rotation (NNR) Frame defined by , where and are the angular velocity and area of plate , with N the number of tectonic plates considered to compose the Earth’s lithosphere9. Plate velocities are computed by removing the net lithospheric rotation component : 11 where 4. Deep Pacific Hotspot Frame, which is anchored to fixed Pacific hotspots12, with transformations: 12 where is the angular velocity of the plate i relative to an arbitrary other plate taken as a reference. This reference frame can be generalized to other deep hotspot frames ( km depth) using the formula 13 where incorporates hotspot motions from deep mantle plumes15. Deep Pacific Hotspot Frame, which is anchored to fixed Pacific hotspots12, with transformations: 12 where is the angular velocity of the plate i relative to an arbitrary other plate taken as a reference. This reference frame can be generalized to other deep hotspot frames ( km depth) using the formula 13 where incorporates hotspot motions from deep mantle plumes15. Shallow Hotspot Frame ( km depth): 14 accounting for plate kinematic models obtained by40 and asthenospheric decoupling through a viscosity jump at less than 250 km depth17. In this article, we explore a range of possible average depths for the decollement layer in between 125 and 250 km and we use such simulations to interpolate the best model (See Fig. 7). Shallow Hotspot Frame ( km depth): 14 accounting for plate kinematic models obtained by40 and asthenospheric decoupling through a viscosity jump at less than 250 km depth17. In this article, we explore a range of possible average depths for the decollement layer in between 125 and 250 km and we use such simulations to interpolate the best model (See Fig. 7). The obtained angular velocities and areas are then compared to the seismic P wave velocities at the bottom lid retrieved by26. For each point of the grid, we associate it with a different plate according to the digital database provided by Bird7 and calculate average, variance and other quantities of interest. Geodetic datasets As described for the previous points, we used those combination properties of plate angular vectors and we explored the geodetic datasets obtained by8, using their proposed NNR geodetic framework to compute plate kinematic models with respect the fixed Pacific plate, and the deep and shallow (225 km) hotspot frames, combining the available variance-covariance matrices as described in47 and references therein, to obtain uncertainties of plate poles and angular velocities. Double power-law modeling The angular velocity distribution of tectonic plates was analyzed using a double power law approach to characterize the distinct scaling regimes of fast and slow plates. For each plate motion model, the survival function representing the number of plates with angular velocity greater than was computed and plotted on log-log axes. The double power law was parameterized as: where is the break velocity separating the two regimes, and are the scaling exponents for the slow and fast plate branches, respectively, and , are normalization constants. Parameter estimation The break velocity and scaling exponents , were determined using maximum likelihood estimation. For each candidate break point sampled from the 20th to 80th percentiles of the angular velocity distribution, the data were split into slow () and fast () plate velocities. Then, linear regression in log-log scale was performed on each segment to obtain the scaling exponents The optimal break velocity point was selected by maximizing the combined log-likelihood of both segments. In doing this, we ensured each branch contained sufficient data points (minimum of 5 plates per segment). For each parameter (break velocity or scaling exponent), the uncertainty was computed as the standard deviation: where models, are the parameter values from each model, and is the mean calculated using all the different models. See Table 1 for results. Supplementary Information Supplementary Information. Acknowledgements The authors thank Vincenzo Nesi for valuable discussions, the editor and two anonymous reviewers for valuable comments which improved the manuscript. Eleonora Ficini acknowledges the support of the European Union-Next Generation EU, Mission 4 - Component 2 - Investment 1.2 THAUMASS – CUP B53C22007490006. Author contributions D.Z. conceived the analysis, performed the investigation, formal analysis, physical modeling of results and wrote the first draft of the manuscript. M.C. and E.F. took care of plate kinematic modeling. C.D. supervised the project and realized the geological interpretation of results. D.Z., E.F., M.C. realized the figures and contributed to data curation. All authors reviewed the manuscript. Data availability The datasets analyzed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Accession codes Scripts and data are accessible upon reasonable request to M.C. and D.Z. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Davide Zaccagnino, Email: zaccagnino@sustech.edu.cn. Carlo Doglioni, Email: carlo.doglioni@uniroma1.it. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-31301-w. References Associated Data Supplementary Materials Supplementary Information. Data Availability Statement The datasets analyzed during the current study are available from the corresponding author on reasonable request.