Roughly 2,900 kilometers beneath the surface of the planet, along the thermal divide where solid silicate rock meets a churning sea of molten iron, geophysicists have detected six massive, previously undocumented geological structures.
The discovery, detailed in a study published in the Journal of Geophysical Research: Solid Earth by a research team from the Institute of Geology and Geophysics at the Chinese Academy of Sciences (IGGCAS), was made possible by deploying deep learning across three and a half decades of global seismic recordings. By automating the detection of faint, scattered acoustic signals that had eluded human analysts for decades, the researchers processed more than two million earthquake waveforms recorded between 1990 and 2024.
The algorithmic scan yielded 174,929 high-quality detections of rare seismic waves known as PKP precursors. This dataset represents an order of magnitude more detections than all previous geophysical surveys combined.
The resulting global map reveals six distinct, anomalous scattering provinces—designated B1 through B6—situated deep beneath high-latitude Eurasia, Central Asia, and the South Atlantic basin.
CRUST (0 - 35 km)
========================================================================
MANTLE (Silicate Rock)
~2,891 km
------------------------------------------------------------------------
CORE-MANTLE BOUNDARY (CMB) / D'' LAYER
[B1] Eurasia [B2] Central Asia [B3] South Atlantic [B4-B6] Undocumented
* Thermochemical piles * Ancient subducted slabs * ULVZs
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
OUTER CORE (Liquid Iron-Nickel)
~5,150 km
........................................................................
INNER CORE (Solid Iron-Nickel)
~6,371 kmRather than appearing as isolated, random patches of anomalous rock, the data demonstrates that these deep-seated structures form broad, continuous belts along the core-mantle boundary (CMB). The findings fundamentally upend conventional models of deep-mantle dynamics, alter our understanding of heat transfer out of Earth's core, and provide fresh targets for deciphering the planetary engine that drives surface volcanism, plate motion, and the geomagnetic field.
The Physics of Deep Scans: Why PKP Precursors Matter
Humanity has barely scratched the surface of the planet. The deepest hole ever drilled, the Kola Superdeep Borehole in Russia, penetrated just 12,262 meters (7.6 miles) into the continental crust before extreme heat and mechanical stress halted operations. The core-mantle boundary sits more than 230 times deeper.
Because direct physical sampling of the deep Earth is impossible with current technology, geophysicists rely on earthquakes as planetary-scale ultrasound emitters.
When a major earthquake occurs (magnitude 6.0 or greater), it releases energy in the form of seismic body waves that propagate through the interior. Compressional waves, or P-waves, travel through both solid rock and liquid metal. As these waves travel through the interior, their velocities change based on the density, composition, and temperature of the material they traverse.
Earthquake Source (Focus)
*
/ \
/ \ P-waves propagate downward
/ \
v v
[ Mantle: Solid Silicate Rock ]
| |
| |
==============|=======|============== Core-Mantle Boundary (CMB)
: | * | :
: Scatterer -> \ / \ / <- Heterogeneity (B1 - B6)
: v v :
: [ Liquid Outer Core (K) ] : PKP precursor path
: | | : scatters off CMB features
: v v : arriving SECONDS BEFORE
: [ Solid Inner Core (I) ] : the main PKIKP wave
=====================================
| |
v v
Surface Seismometers
(Recording faint precursor arrivals)The study focused on a specific phase of seismic waves known as PKP waves:
- P designates a compressional wave traveling through the crust and mantle.
- K (from the German Kern, meaning core) indicates that the wave entered and traveled through the liquid iron outer core.
- P indicates that the wave exited back into the mantle and crust before reaching a seismic receiver at the surface.
Under normal conditions, a P-wave entering the Earth passes into the liquid outer core, refracts, and travels directly to an observation station on the opposite side of the globe. When a wave travels straight through both the liquid outer core and the solid inner core, it is recorded as a PKIKP arrival (or PKP core phase).
However, if a down-going P-wave strikes a localized patch of unusual material—a pocket of high density, a structural step, or a region of partial melt—just above or at the core-mantle boundary, part of that wave energy scatters.
Because the scattered wave travels along an alternative geometric path through the outer core, it bypasses the slower core-refracting trajectory and reaches surface seismometers seconds before the main PKIKP arrival.
Seismologists call these faint, early-arriving packets of acoustic energy PKP precursors. They act as acoustic beacons. Whenever an instrument registers a PKP precursor, it proves that somewhere along that wave's deep journey, it encountered a physical obstacle—a distinct, small-scale heterogeneity at the base of the mantle.
Overcoming the Seismic Data Bottleneck
Despite their utility for imaging the deep Earth, PKP precursors have historically been among the most difficult seismic signals to study.
First, they are weak. Because they represent scattered energy rather than the direct seismic wavefront, their amplitude is tiny, often buried within the background noise of ocean waves, human activity, and late-arriving crustal reverberations.
Second, identifying them required visual inspection by trained seismologists. A researcher had to inspect individual seismograms, cross-reference travel-time curves, manually adjust filter bands, and evaluate whether a trace represented an actual precursor or ambient noise.
SEISMOGRAM TRACE
Baseline Noise PKP Precursor (Faint Scatter) Main PKIKP Wave Arrival
------------------/\/\--------/\/\/\-----------------------/ \
^ / \
| / \
Faint signal / \
(AI classified) / \
/ \
------------------------------------------------------------------------------------>
Time (seconds)Manual analysis came with steep constraints:
- Subjectivity: Disagreements frequently arose between researchers over whether a subtle deflection in the waveform was a genuine scattering signal or an instrumental artifact.
- Severe Data Scarcity: Over four decades of international manual cataloging, researchers assembled fewer than 15,000 confirmed PKP precursor events globally.
- Geographic Blind Spots: Because manual review took so long, teams focused on small, pre-selected corridors—such as directly beneath the southwest Pacific or parts of North America—leaving vast swaths of the deep Earth uninspected.
To break this bottleneck, the IGGCAS team constructed a deep learning pipeline designed to mimic and surpass manual waveform classification.
The researchers gathered raw seismic recordings from every magnitude 6.0 or larger earthquake documented across international seismic networks between 1990 and 2024, amassing more than two million distinct waveforms.
They built a multi-stage convolutional neural network (CNN) combined with an active learning feedback loop:
- Signal Conditioning and Quality Screening: The network first ingested three-component seismograms, filtered them across bandpass frequencies sensitive to deep scattering (typically 0.5 to 2.0 Hz), and eliminated records with poor signal-to-noise ratios.
- Precursor Identification: The model evaluated the time window directly preceding the theoretical PKIKP arrival, calculating probability vectors to determine if an early deflection met the kinematic criteria of a scattered core arrival.
- Iterative Expert Calibration: When the algorithm flagged ambiguous waveforms, senior seismologists manually reviewed and corrected the classifications. These verified cases were fed back into the training architecture, continuously sharpening the model's feature extraction boundaries.
- Dual-Probability Spatial Mapping: Once the network identified 174,929 authentic precursors, the team ran a dual-probability spatial framework. This algorithm combined the precursor occurrence probability with the geometric scatterer location probability, cross-projecting hundreds of thousands of crisscrossing raypaths to pinpoint the exact 3D coordinates where the scattering events occurred.
The deep learning framework achieved in weeks what would have required centuries of manual human labor, transforming a sparse scatter of data points into a coherent, high-resolution global image.
Characterizing the Six Earth Core Structures
When the team plotted the coordinates of the 174,929 precursors, the resulting global distribution exposed six massive scattering provinces that had never been documented in scientific literature.
These earth core structures, labeled B1 through B6 in the study, reside in the lowermost 200 to 300 kilometers of the mantle—a turbulent transition zone known to geophysicists as the D'' (D-double-prime) layer.
GEOGRAPHIC DISTRIBUTION OF ANOMALIES
80°N +---------------------------------------------------+
| |
| [B1] High-Latitude Eurasia |
40°N | [B2] Central Asia |
| |
0° | (African LLVP) (Pacific LLVP) |
| |
40°S | [B3] South Atlantic |
| |
80°S +---------------------------------------------------+
180°W 120°W 60°W 0° 60°E 120°E 180°E
* B1 - B6: Newly resolved deep scattering belts (Guan et al., 2026)
* LLVPs: Continent-sized Large Low-Shear-Velocity ProvincesBefore this catalog was assembled, prevailing models treated small-scale mantle heterogeneities as isolated, random anomalies—tiny, erratic islands of strange rock surrounded by uniform mantle material. The scale of the new dataset showed that these structures are far more organized.
Rather than appearing as disconnected patches, the newly resolved earth core structures form continuous, elongated belts that stretch for thousands of kilometers along the core-mantle boundary.
The Newly Identified Scattering Belts (B1–B6)
- Zone B1 (High-Latitude Eurasia): Located directly beneath northern Siberia and the Arctic margins of Eurasia. This vast scattering belt sits beneath a region historically characterized as a seismic graveyard for ancient oceanic plates subducted millions of years ago. The PKP precursor density here indicates extreme internal complexity, featuring rapid, short-wavelength shifts in acoustic impedance.
- Zone B2 (Central Asia): Positioned beneath the continental interior of Central Asia, extending southward toward the northern edge of the ancient Tethyan subduction belt. This structure exhibits some of the strongest scattering signals in the catalog, suggesting a sharp density contrast with the surrounding lower-mantle matrix.
- Zone B3 (South Atlantic Basin): Situated beneath the South Atlantic Ocean, flanking the southwestern boundary of the African superplume. This structure appears to form a transitional bridge between the background mantle and the superheated, buoyant interior of the African lower mantle.
- Zones B4, B5, and B6: Distributed across previously under-sampled deep-mantle corridors across the Southern Oceans and the fringes of the circum-Pacific rim. These belts demonstrate that strong seismic scattering is not limited to subduction zones or volcanic hotspots, but forms an interconnected network of lower-mantle features.
+---------------+------------------------+------------------------------------+---------------------------------------+
| Belt Label | Geographic Region | Deep-Earth Context | Hypothesized Physical Origin |
+---------------+------------------------+------------------------------------+---------------------------------------+
| B1 | High-Latitude Eurasia | Cold slab burial corridor | Ancient subducted oceanic crust/MORB |
+---------------+------------------------+------------------------------------+---------------------------------------+
| B2 | Central Asia | North Tethyan margin | Deformed slab remnants & high-P phase |
+---------------+------------------------+------------------------------------+---------------------------------------+
| B3 | South Atlantic Basin | African LLVP western margin | Thermochemical pile & partial melt |
+---------------+------------------------+------------------------------------+---------------------------------------+
| B4 | Southern Oceans | Deep oceanic mantle floor | Basal magma ocean cumulates |
+---------------+------------------------+------------------------------------+---------------------------------------+
| B5 | Circum-Pacific Flank | Boundary of Pacific superplume | Ultra-Low Velocity Zone (ULVZ) fringe |
+---------------+------------------------+------------------------------------+---------------------------------------+
| B6 | Sub-Polar Corridor | Low-strain lower-mantle zone | Iron-silicate chemical reaction zone |
+---------------+------------------------+------------------------------------+---------------------------------------+What Lies at 2,900 Kilometers?
What are these structures made of? At a depth of 2,900 kilometers, the physical environment reaches extremes that defy intuitive surface physics.
The pressure exceeds 1.3 million atmospheres (136 Gigapascals), and temperatures swing violently across the boundary, jumping from approximately 2,500 degrees Celsius in the solid mantle to more than 3,800 degrees Celsius in the liquid outer core.
Because seismic scans only measure the speed and scattering of sound waves, identifying the material composition of these earth core structures requires reconciling seismic data with high-pressure mineral physics and numerical geodynamics.
Researchers have put forward five primary hypotheses to explain the presence of these massive scattering belts.
FIVE HYPOTHESES FOR THE STRUCTURES' ORIGIN
[1] Ancient Tectonic Slabs [2] Basal Magma Ocean Relics
- Subducted MORB crust - Primordial dense cumulates
- Hundreds of millions of yrs - Surviving from 4.5 Ga
\ /
v v
+---------------------------------+
| CORE-MANTLE BOUNDARY (D'') |
| Discovered Belts B1 - B6 |
+---------------------------------+
^ ^ ^
/ | \
[3] Ultra-Low Velocity [4] Theia Impact [5] Core-Mantle
Zones (ULVZs) Debris Piles Chemical Leaking
- Partial melting - Giant impactor - Molten iron & silicate
- S-wave drops up to 50% mantle remnants chemical reactions1. The Graveyard of Ancient Ocean Basins
The most direct explanation ties these structures to plate tectonics. Over hundreds of millions of years, oceanic plates dive into the mantle at subduction zones—such as the Marianas Trench or the Peru-Chile Trench.
While geologists once believed that subducted plates stalled in the upper mantle or broke apart at the 660-kilometer transition zone, modern tomographic studies demonstrate that cold, dense tectonic slabs can sink all the way down to the core-mantle boundary.
When an oceanic plate reaches the base of the mantle, its top layer—composed of basaltic oceanic crust (Mid-Ocean Ridge Basalt, or MORB)—is rich in silica and aluminum compared to the surrounding peridotitic mantle. Under the crushing pressures of the D'' layer, this basalt undergoes phase transformations into ultra-dense mineral assemblies.
As the slab warms over tens of millions of years, it deforms, crumples, and pools into dense, laterally continuous belts along the core floor. The locations of zones B1 and B2 align directly with the projected long-term trajectories of ancient subduction systems that once rimmed northern and central Eurasia.
2. Relics of a Primordial Basal Magma Ocean
An alternative hypothesis looks back 4.5 billion years to the formation of the Earth. In the immediate aftermath of planetary accretion, the heat generated by gravitational compression, radioactive decay, and massive planetary impacts melted much of the interior, creating a deep magma ocean.
As this primordial magma ocean slowly cooled and crystallized from the middle outward, the earliest minerals to solidify (such as magnesium-rich bridgmanite) rose or sank depending on their density. Iron-rich melts and heavy, incompatible elements were driven downward, accumulating in a dense, molten layer at the very base of the mantle.
Over billions of years of mantle convection, most of this primordial layer was churned and mixed. However, the most resilient, iron-heavy pockets would resist entrainment by mantle convection currents. The newly discovered belts could represent fossilized reservoirs of this primordial Earth, preserved at the bottom of the planet since its formation.
3. Ultra-Low Velocity Zones (ULVZs) and Localized Partial Melting
A third candidate is that the structures are Ultra-Low Velocity Zones (ULVZs). ULVZs are patches of lower-mantle rock where seismic velocities drop abruptly—shear wave speeds can plunge by up to 50%, and compressional wave speeds fall by 10% to 20%.
ULVZ WAVE VELOCITY PROFILES
Depth (km)
2800 | Mantle Peridotite: Baseline Seismic Velocity (PREM)
| /
2850 | /
| |
2891 | [CMB] |<-- Sharp Boundary
| |
| +--- P-wave velocity drops by 10% - 20%
| +------- S-wave velocity drops by up to 50%
| |
v [ ULVZ / Partial Melt / Iron Enrichment ]Because velocity reductions of this magnitude cannot be explained simply by variations in rock temperature, geophysicists believe ULVZs must contain either partial melt or extreme iron enrichment.
If heat escaping from the liquid outer core pushes the temperature of lower-mantle minerals past their solidus point, a slush of partially molten rock forms. This molten material gathers in topographic depressions along the core-mantle interface.
The sharp structural margins of these partial-melt zones would produce strong acoustic impedance contrasts, scattering P-waves with the exact signatures captured by the deep learning model.
4. The Planetary Impact Remnant: Shards of Theia
One of the most radical interpretations ties these structures to the birth of the Moon. Under the Giant Impact hypothesis, a Mars-sized protoplanet dubbed Theia collided with the proto-Earth roughly 4.5 billion years ago. The impact stripped away much of Theia's lighter crust, creating the debris ring that coalesced into our Moon.
What happened to Theia's mantle? Recent high-resolution computer simulations published in Nature by Yuan et al. suggest that Theia's mantle was inherently denser and more iron-rich than Earth's.
Following the impact, large slabs of Theia's iron-heavy mantle plunged through Earth's molten interior and settled onto the core-mantle boundary.
While geophysicists have long speculated that Earth's two known continental-sized anomalies—the African and Pacific Large Low-Shear-Velocity Provinces (LLSVPs)—are remnants of Theia, the discovery of zones B1 through B6 raises the possibility that Theia's mantle shattered into smaller fragments that formed a broader network of debris encircling the core.
5. Chemical Reactions and Core-Mantle Leaking
Finally, these structures may be products of chemical reactions occurring across the core-mantle interface. For decades, scientists treated the boundary between the solid silicate mantle and the liquid iron outer core as chemically inert.
Recent diamond anvil cell experiments, however, reveal that at 136 Gigapascals, liquid iron actively reacts with silicate minerals like bridgmanite and post-perovskite.
Iron dissolves into the lower mantle, while silicon, oxygen, and hydrogen from subducted minerals migrate into the uppermost core. Over billions of years, this mutual contamination creates a hybridized reaction layer—a "metallized" boundary layer of iron silicides, iron oxides, and dense mineral alloys.
If mantle currents drag these chemical reaction rinds along the core surface, they would naturally form elongated, highly reflective belts of scatterers that match the observed B1–B6 signatures.
Who Is Affected: The Scientific and Technological Fallout
The identification of these six structures is not just an isolated discovery in geophysics; it creates ripple effects across multiple Earth science and computational disciplines.
CROSS-DISCIPLINARY IMPACTS
+-------------------------------------------------------+
| DISCOVERY: 6 CONTINUOUS CORE-MANTLE STRUCTURES (B1-B6)|
+-------------------------------------------------------+
| | |
v v v
[ Computational Geophysics ] [ Geodynamics & Dynamos ] [ Surface Hazards ]
- Deep learning for legacy - Heat flux across CMB - Plume root anchoring
waveforms (2M+ audited) - Dynamo turbulence & - Hotspot formation
- Re-evaluating 50 years geomagnetic jerks (supervolcano risks)
of seismic archives - Core-mantle torque - Long-term plate driveSeismologists and Computational Geophysicists
For seismologists, the immediate impact is a transformation of methodology. For fifty years, global seismology has faced an archive problem: vast databases maintained by institutions such as the Incorporated Research Institutions for Seismology (IRIS) and the Global Seismographic Network (GSN) were accumulating raw seismic data far faster than human teams could inspect them.
The IGGCAS team's success validates deep-learning-driven waveform analysis at global scale. Researchers are now adapting these neural architectures to re-examine other faint, difficult-to-isolate seismic phases, including:
- ScS and ScP Waves: Shear waves bouncing directly off the core-mantle boundary.
- Pdiff Waves: Compressional waves diffracted along the curved core boundary.
- SKS and SKKS Waves: Shear waves traversing the liquid outer core that provide constraints on deep-mantle anisotropy.
Global seismology is shifting from a data-bottlenecked discipline to an algorithmic recovery field, where legacy recordings from the 1990s and 2000s are re-examined to reveal previously hidden structures.
Geodynamicists and Numerical Modelers
Scientists who run supercomputer simulations of mantle convection are directly affected by these findings.
Previous geodynamic models operated under simplified boundary conditions, treating the lower boundary of the mantle as either uniformly conductive or characterized only by two giant equatorial anomalies (the African and Pacific LLSVPs).
The discovery of the B1–B6 belts demonstrates that the base of the mantle is far more heterogeneous than those simulations assumed.
Modelers must now insert these narrow, high-density, high-roughness belts into 3D mantle circulation simulations. Because small-scale boundary heterogeneities alter how viscous fluid convects, incorporating these structures will change predictions of how heat circulates through the mantle, how thermal plumes originate, and how rapidly subducting slabs descend.
Geomagnetism and Dynamo Theorists
Planetary physicists studying the geodynamo—the convective process in the liquid iron outer core that generates Earth's protective magnetic field—must reckon with altered boundary conditions.
Earth's magnetic field is sustained by thermal and compositional convection within the outer core. The vigor and pattern of that convection depend directly on the rate at which heat is extracted from the liquid core into the cooler, solid mantle above.
If the base of the mantle is insulated by dense, partially molten, or iron-rich belts, heat extraction will be uneven across the globe. Regions with thick, insulating scatterers will trap heat within the outer core, suppressing convection, while regions adjacent to cool, subducted slabs will experience rapid cooling, accelerating core convection currents.
Modulating core heat extraction across these earth core structures directly influences liquid iron convection, offering an explanation for long-standing magnetic mysteries:
- Geomagnetic Jerks: Abrupt, unexpected accelerations in Earth's magnetic secular variation that disrupt navigation systems and satellite instrumentation.
- Magnetic Dipole Decay: Why Earth's global magnetic field strength has weakened by roughly 9% over the past two centuries.
- The South Atlantic Anomaly: The vast depression in magnetic field strength hovering over South America and the South Atlantic Ocean—directly above newly mapped Zone B3.
GEODYNAMO HEAT-FLUX COUPLING
Solid Silicate Mantle (Mantle Convection Upwelling)
============================================================
[ High Heat Flux ] [ Low Heat Flux ]
Cool, subducted slab Insulating belt (B1-B6)
Rapid heat extraction Heat trapped in outer core
------------------------------------------------------------ Core-Mantle Boundary
Vigorous convection Suppressed convection
Strong local field generation Turbulent eddy formation
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Churning Liquid Iron Outer Core (Geodynamo Engine)Volcanologists and Hazard Research Communities
While these structures sit nearly 3,000 kilometers beneath our feet, their thermal footprints reach the surface. The core-mantle boundary serves as the thermal launchpad for mantle plumes—superheated upwellings of rock that rise through the mantle to feed intraplate volcanism and continental flood basalts.
Places like Hawaii, Iceland, Yellowstone, and the Galapagos are not driven by shallow plate boundaries, but by deep-seated mantle plumes.
Field observations demonstrate that mantle plumes almost always originate along the edges of lower-mantle heterogeneities, where sharp temperature and viscosity gradients focus upward thermal flow.
Mapping the precise boundaries of belts B1 through B6 gives volcanologists a clearer look at the root systems of ancient and emerging hotspot tracks.
By understanding where deep-seated structures are anchored, geologists can trace how heat moves from Earth's core to the crust, informing long-term models of supervolcano evolution and mantle-driven intraplate earthquakes.
What Changes: A New Look at the Core-Mantle Boundary
The identification of these six structures alters our conceptual framework for the deep Earth in several fundamental ways.
+------------------------------------+------------------------------------+
| Old Understanding | New Understanding |
+------------------------------------+------------------------------------+
| The CMB is mostly smooth, | The CMB is a structurally complex, |
| interrupted only by two giant | highly dynamic interface flanked |
| blobs (African and Pacific LLSVPs).| by continuous scattering belts. |
+------------------------------------+------------------------------------+
| Small-scale anomalies are sparse, | Anomalies form thousands-of- |
| isolated, and randomly scattered | kilometer-long connected belts |
| pockets of rock. | along convective corridors. |
+------------------------------------+------------------------------------+
| Seismic catalogs are limited by | Machine learning unlocks millions |
| decades of painstaking manual | of archival seismic traces, |
| waveform analysis. | yielding 10x more detections. |
+------------------------------------+------------------------------------+
| Lower-mantle seismic models blur | Precursor scattering reveals fine- |
| fine details below 500-1000 km | scale structures at the 10-to- |
| resolution scales. | 100-kilometer scale. |
+------------------------------------+------------------------------------+1. The Death of the "Homogeneous Mantle" Assumption
For decades, early geophysical textbooks presented the Earth as an onion: a thin crust, a uniform upper mantle, a uniform lower mantle, a smooth liquid outer core, and a solid inner core.
Over time, tomographic imaging added two continent-sized "blobs" (the African and Pacific LLSVPs) to this model. Yet outside those two mega-structures, the base of the mantle was widely treated as an otherwise homogeneous layer.
The new data invalidates that view. The lowermost mantle is a heterogeneous planetary frontier characterized by complex topography, varying mineral phases, and sharp compositional boundaries.
The existence of belts B1 through B6 confirms that lateral mixing in the deep mantle is inefficient.
Dense materials do not get evenly distributed throughout the mantle; instead, they are swept into distinct structural belts that persist over hundreds of millions of years.
2. Moving from Tomographic "Blur" to Structural Precision
Traditional seismic tomography works much like medical CT scans: it measures the bulk arrival times of long-period seismic waves.
While tomography excels at imaging features thousands of kilometers wide—such as entire subducting slabs or broad superplumes—it smooths away features smaller than roughly 500 kilometers.
Analyzing scattered PKP precursors, by contrast, operates on the scale of acoustic diffraction. Because high-frequency P-waves scatter off boundaries with dimensions comparable to their seismic wavelength, PKP precursors are sensitive to heterogeneities measuring between 10 and 100 kilometers wide.
The deep learning model developed by the Chinese Academy of Sciences effectively bridges the gap between planetary-scale tomography and localized high-resolution imaging, resolving the fine structural details of the deep Earth.
IMAGING RESOLUTION COMPARISON
Traditional Tomography (Travel-Time Delays)
[================== 500 km to 2,000 km Resolution ==================]
* Captures giant LLSVPs ("blobs")
* Blurs away fine boundaries, melt pockets, and sharp slab walls
Precursor Scattering Analysis (Guan et al., 2026 Deep Learning Model)
[=== 10 km to 100 km ===]
* Captures discrete scattering belts (B1 - B6)
* Pinpoints structural margins, thermal rinds, and partial melt zones3. The Interconnection of Deep Reservoirs
Previously, geophysicists who discovered a localized patch of low velocity beneath northern Siberia or the southern Pacific assumed they were observing an isolated quirk of deep-Earth geology.
By analyzing nearly 175,000 arrivals at once, the new global map reveals that these anomalies align along continuous corridors.
This continuous structure indicates that mantle convection operates like a planetary conveyor belt, sweeping dense chemical fragments, subducted crustal remnants, and partial-melt pockets along predictable dynamic corridors at the base of the planet.
Short-Term Consequences (1 to 5 Years)
The publication of this catalogue sets off immediate empirical and logistical initiatives across international research institutions.
SHORT-TERM ROADMAP (1-5 YEARS)
[ Year 1 - 2 ] ------------------------> [ Year 3 - 4 ] ------------------------> [ Year 5 ]
* Global deployment of * Temporary broadband OBS arrays * New 3D reference models
machine learning across deployed over Pacific/Atlantic replace 1981 PREM.
ScS, Pdiff, and SKS traces. geometric shadow zones. * Diamond anvil cell tests
* Open-source validation * Integration of deep-mantle belts synthesize deep-mantle
of CAS neural network code. into geodynamo simulations. scattering rock alloys.1. Algorithmic Re-Auditing of Legacy Seismic Archives
The immediate priority for global seismology teams is verifying and expanding the IGGCAS findings.
Research groups at institutions including Caltech, Cambridge, the University of Science and Technology of China (USTC), and the Institut de Physique du Globe de Paris (IPGP) are mobilizing similar neural models to audit other wave phases.
By training neural classifiers to identify precursors within core-reflected shear waves (ScS) and diffracted P-waves (Pdiff), geophysicists expect to assemble complementary global catalogues within the next 24 to 36 months.
These cross-phase datasets will test whether the B1–B6 belts display sharp reductions in shear wave speed, which would provide definitive evidence of partial melting or high iron content.
2. Targeted Seismometer Array Deployments
Because earthquakes occur primarily along tectonic boundaries and seismometers are heavily clustered on continental landmasses, global seismic data coverage remains uneven.
The identification of zones B1 through B6 gives observational seismologists specific geographic targets.
Over the next three years, international seismic consortia are adjusting their deployments:
- Repositioning mobile broadband sensor networks to capture geometric illumination corridors passing through zones B1 and B2.
- Deploying ocean-bottom seismometers (OBS) in the South Atlantic to resolve the fine boundaries of zone B3.
- Coordinating seismic monitoring arrays to focus on earthquakes whose raypaths traverse the geometric margins of these newly documented structures.
3. Diamond Anvil Cell Experiments at New Extremes
In high-pressure mineral physics laboratories—such as the Advanced Photon Source (APS) in Illinois, the European Synchrotron Radiation Facility (ESRF) in France, and SPring-8 in Japan—researchers are designing new experiments to simulate conditions within the B1–B6 belts.
Using laser-heated diamond anvil cells, physicists will compress synthetic samples of basaltic crust, peridotite, and iron-nickel alloys to 136 Gigapascals and heat them to 3,500 Kelvin.
By firing synchrotron X-rays through these microscopic, superheated samples, scientists will measure how sound waves travel through specific mineral phases, identifying the exact rock recipes that match the scattering signatures cataloged by the CAS deep learning pipeline.
Long-Term Consequences (5 to 50 Years)
Over longer timeframes, the existence of continuous, belt-like earth core structures alters core principles of planetary science, geophysics, and Earth evolution.
LONG-TERM GEOLOGICAL IMPACTS
[ Geomagnetic Stability ] [ Supercontinent Cycles ]
- Explains superchrons - Governs how mantle plumes rise
(eras with no magnetic reversals) - Predicts continental breakup and
- Maps persistent field weaknesses assembly of Pangea Ultima (250 Ma)
\ /
v v
+----------------------------------------------------+
| NEW CONSERVATIVE DYNAMICS OF DEEP-EARTH EVOLUTION |
+----------------------------------------------------+
^ ^
/ \
[ The Primordial Volatile Budget ] [ Planetary Habitability Models ]
- Isolates primordial Helium-3 - Thermal regulation of outer core
and Xenon reservoirs sustains atmosphere-protecting
- Resolves early accretion timelines magnetic shield over billions of yrs1. Solving the Mystery of Superchrons and Magnetic Reversals
Earth's magnetic field periodically reverses its polarity, swapping magnetic north and south.
Throughout Earth's history, the reversal frequency has fluctuated dramatically. During the Cretaceous Normal Superchron (roughly 120 to 83 million years ago), the magnetic field maintained a single polarity for nearly 40 million years without a reversal. During other geological eras, the field flipped every several hundred thousand years.
What governs these swings in geodynamo stability? Magnetohydrodynamic simulations show that magnetic reversals are triggered when core convection becomes chaotic and asymmetric.
The primary stabilizing force is the pattern of heat extraction across the core-mantle boundary.
Over geological timescales (tens to hundreds of millions of years), the slow movement of subducting slabs and the accumulation of insulating belts like B1–B6 alter heat distribution across the CMB.
By reconstructing the migration of these structures across deep time, geophysicists will finally test whether superchrons coincide with periods when cold, subducted slabs covered the core, or when insulating thermochemical piles dominated, changing our understanding of long-term geomagnetic field stability.
THE GEODYNAMO REVERSAL TIMELINE
Past (Hundreds of Millions of Years) Future (Deep Time)
========================================================================>
Cretaceous Normal Superchron High-Frequency Reversals Next Polarity Shift?
(No reversals for 40 Myr) (Flips every ~300,000 yrs) (Governed by moving
Deep slabs blanket CMB floor; Insulating belts split; B1-B6 belt thermal
Symmetric, stable geodynamo Turbulent, chaotic dynamo flux footprints)2. The Supercontinent Cycle and Plume Triggering
Surface continents continuously assemble and break apart in the 500-million-year supercontinent cycle (from Rodinia to Pangea, and onward to the future supercontinent Pangea Ultima or Amasia).
This dance is driven by the sinking of cold oceanic lithosphere and the upwelling of hot mantle plumes.
Because zones B1 through B6 represent massive accumulations of dense material at the base of the mantle, they act as deep-seated dams. As cold slabs sink to the bottom of the planet, they cannot easily penetrate these dense belts; instead, they are deflected along their flanks.
This channeling focuses thermal stresses along the edges of the belts, turning their margins into preferred launchpads for mantle superplumes.
Over the coming decades, integrating these structural boundaries into plate tectonic reconstructions will provide predictive models for where continental rifting will occur and how future supercontinents will assemble.
3. Preserved Reservoirs of Primordial Earth and Atmospheric Evolution
A long-standing debate among geochemists is the "primitive mantle" paradox.
When basalts erupt from ocean-island volcanoes like Hawaii or Iceland, they carry primitive isotopic ratios—including elevated ratios of Helium-3 to Helium-4, as well as distinct Xenon and Tungsten signatures—that reflect reservoirs that have remained largely unmixed since the planet formed 4.5 billion years ago.
Where are these ancient reservoirs stored? The standard mantle has been churned by convection for billions of years, which should have mixed those primordial signatures away.
The discovery that deep-mantle structures form organized, laterally continuous belts offers a long-term solution.
If belts B1 through B6 contain dense, iron-rich cumulates left behind by the basal magma ocean or remnants of the protoplanet Theia, their high density would keep them settled on the CMB floor, protecting them from the convective currents that churn the rest of the mantle.
These belts could serve as the long-sought geochemical storage vaults of the deep Earth, slowly leaking ancient noble gases and primordial volatiles into rising plumes over billions of years.
THE PRIMORDIAL RESERVOIR PARADOX
[ Ocean-Island Basalt Eruption: Hawaii / Iceland ]
- High 3He/4He, anomalous 182W signatures
- Primitive volatiles from the dawn of the Solar System
^
|
(Rising Mantle Plume)
|
============================================================
Upper & Lower Mantle (Vigorously Convecting)
============================================================
^
| Entrained volatiles slowly escape
|
+-------------------------------------+
| Discovered Belts B1 - B6 |
| * High density resists entrainment |
| * Preserves 4.5-billion-year-old |
| primordial geochemical signatures|
+-------------------------------------+
------------------------------------------------------------ Core-Mantle Boundary
Molten Liquid Iron Outer Core4. Planetary Habitability and Exoplanet Modeling
Earth is the only known planet in the Solar System that maintains both active plate tectonics and a sustained, high-strength magnetic dipole.
Mars lost its magnetic dynamo roughly four billion years ago, allowing solar wind to strip away its atmosphere and desiccate its surface; Venus possesses a dense atmosphere but lacks both plate tectonics and an internally generated magnetic field.
Deciphering the balance of these newly identified structures will help explain why Earth remained habitable while its sister planets diverged.
The continuous extraction of heat across the core-mantle boundary—mediated by the interplay between sinking slabs and deep-seated structural belts—acts as Earth's central thermal regulator.
As the James Webb Space Telescope and future space observatories analyze the atmospheres of rocky exoplanets, astrophysicists will use these refined deep-mantle models to evaluate whether an alien world possesses the internal thermal structure needed to sustain a planetary magnetic shield over evolutionary timescales.
Unresolved Questions and What to Watch For Next
The discovery made by the Chinese Academy of Sciences resolves an observational bottleneck, but it opens a series of fundamental questions about Earth's internal architecture.
KEY MILESTONES TO WATCH NEXT
[ Distributed Acoustic Sensing ] [ Global Machine Learning ] [ Diamond Anvil Synchrotrons ]
Using subsea telecom fiber Expanding beyond PKP to Subjecting bridgmanite
as planetary-scale seismic ScS, Pdiff, and SKS phases and post-perovskite to
arrays across ocean basins. for 3D phase tomography. 136 GPa and 3,800 Kelvin.As the global scientific community digests this dataset, several milestones will indicate how our understanding of Earth's deep interior is evolving:
- Confirmation via Independent Wave Phases: Researchers will look to see whether independent teams verify zones B1 through B6 using different seismic phases, such as core-reflected shear waves (ScS) and transmitted SKS waves. If these belts show matching drops in shear-wave velocities, the case for partial melting or extreme iron enrichment will become definitive.
- Constraining Absolute Velocity Contrasts: While the precursor catalogue pinpoints where the scattering happens, it cannot easily untangle trade-offs between a structure's thickness, shape, and absolute velocity drop. High-performance waveform-inversion modeling on supercomputers will be required to calculate the exact acoustic properties of belts B1 through B6.
- Distributed Acoustic Sensing (DAS) Innovations: Seismologists are beginning to convert trans-oceanic fiber-optic telecommunication cables into massive seismic arrays via laser reflectometry. Over the next decade, oceanic DAS networks will provide thousands of new virtual seismic stations across the Pacific and Atlantic basins, illuminating shadow zones beneath regions B3, B4, and B5 with unprecedented spatial density.
- Integration into Global Reference Models: Since 1981, geophysics has relied on the Preliminary Reference Earth Model (PREM) as the standard 1D blueprint for how density and seismic wave speed vary with depth. The discovery that the lowermost mantle contains continuous, heterogeneous belts marks a key step toward replacing PREM with a high-resolution, three-dimensional reference model of Earth's interior.
- The Movement Question: Are these structures stationary anchors like the deep LLSVPs appear to be, or are they slowly swept along the core surface by mantle convection currents? Determining whether belts B1 through B6 migrate across geological eras will settle long-running disputes over how the core and mantle exchange momentum and heat.
The interior of our planet is not an inert ball of cooling stone, but an active, layered engine whose deepest features shape the continents, drive volcanism, and sustain the planetary magnetic shield that protects our atmosphere.
By deploying deep learning to decode faint echoes hidden within three decades of seismic noise, researchers have illuminated an interconnected subterranean frontier at Earth's core—one that will occupy the next generation of geophysicists, mineral physicists, and planetary scientists.
Reference:
- https://www.indoneo.com/science/china-deep-learning-earth-core-mantle-boundary-seismic/
- https://fomoera.com/una-ia-hallo-174-929-senales-en-el-fondo-del-manto-y-los-parches-sueltos-resultaron-cinturones/
- https://www.thenews.com.pk/latest/1414093-ai-just-found-175000-hidden-signals-inside-earth
- https://www.sciencealert.com/six-never-before-seen-mystery-structures-found-lurking-deep-inside-the-earth
- https://www.researchgate.net/institution/Institute-of-Geology-and-Geophysics-Chinese-Academy-of-Sciences
- https://www.sdpnoticias.com/sorprendente/descubren-seis-misteriosas-estructuras-en-las-profundidades-de-la-tierra/
- https://futurism.com/science-energy/strange-new-structures-deep-inside-earth-precursors
- https://mycharisma.com/news/scientists-discover-6-mysterious-structures-deep-inside-earth/
- https://futurism.com/science-energy/strange-new-structures-deep-inside-earth-precursors
- https://www.sciencealert.com/six-never-before-seen-mystery-structures-found-lurking-deep-inside-the-earth
- https://dailygalaxy.com/2026/08/scientists-discover-six-structures-earth/
- https://www.indoneo.com/science/china-deep-learning-earth-core-mantle-boundary-seismic/
- https://www.facebook.com/popularmechanics/posts/researchers-trained-an-ai-model-to-identify-signals-in-the-earths-core-mantle-bo/1441714634482435/
- https://news.kagi.com:443/science/2026083012/scientists-identify-six-deep-earth-structures-using-ai
- https://pubs.geoscienceworld.org/ssa/tsr/article/4/2/111/644002/Evidence-for-Ultra-Low-Velocity-Zone-Genesis-in
- https://endtimeheadlines.org/2026/09/scientists-discover-six-never-before-seen-mystery-structures-lurking-deep-inside-the-earth/
- https://yes-dtn.ac.uk/research/the-deep-unknown-imaging-earths-deep-interior-with-massive-seismic-datasets-and-machine-learning/
- https://www.dur.ac.uk/media/durham-university/departments-/earth-sciences/msc-projects-2021/MScR_Aleutians_ULVZ.pdf
- https://en.wikipedia.org/wiki/Ultra-low_velocity_zone
- https://en.ustc.edu.cn/info/1007/4345.htm
- https://asu.elsevierpure.com/en/publications/ultralow-velocity-zones-at-the-core-mantle-boundary-near-the-caro/
- https://www.facebook.com/Abovethenormnews/videos/-something-strange-is-hiding-near-earths-core-earthquake-waves-are-hitting-myste/1044428675122812/
