🔢 mathematics

An implicit octree-based adaptive Material Point Method

This paper introduces the first octree-based implicit Material Point Method that automatically adapts mesh and particle resolution to continuum-structure interfaces, achieving up to a 5.5-fold speedup and significant energy savings in large deformation 3D simulations compared to traditional conforming meshes.

Original authors: Robert E. Bird, William M. Coombs, Charles E. Augarde, Ted J. O'Hare

Published 2026-06-09

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Robert E. Bird, William M. Coombs, Charles E. Augarde, Ted J. O'Hare

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to simulate a massive landslide or a heavy anchor dragging through the ocean floor on a computer. In the past, to get an accurate picture of what happens where the hard object (like the anchor) touches the soft earth, scientists had to build a digital "net" (a mesh) that was incredibly fine and detailed everywhere, even in the empty space far away from the action.

Think of it like trying to take a high-definition photo of a single ant walking on a football field. To get the ant perfectly clear, you'd have to use a camera sensor with billions of pixels covering the entire field, even though 99% of the image is just empty grass. This takes forever to process and uses a huge amount of electricity.

This paper introduces a smarter way to do this using a method called the Material Point Method (MPM), combined with a clever trick called an Octree.

The "Smart Zoom" Camera (The Octree)

Instead of using a uniform net everywhere, the authors built a system that acts like a smart camera with a zoom lens.

The Action Zone: Wherever the rigid object (like a building or an anchor) touches the soil, the computer automatically "zooms in" and creates a very fine, detailed mesh. It's like having a super-high-resolution camera focused only on the ant.
The Quiet Zone: Far away from the action, where nothing interesting is happening, the computer "zooms out" and uses a coarse, blocky mesh. It's like looking at the rest of the football field with a low-resolution, blurry lens.

This "Octree" structure is like a digital tree that keeps splitting branches only where necessary. If a branch needs more detail, it splits into eight smaller pieces (hence "octree"). If it doesn't, it stays big. This means the computer only spends its energy on the parts of the simulation that actually matter.

The "Ghost" Safety Net

When you have a mesh that is zoomed in and out so drastically, you get some tricky gaps where the fine details meet the coarse blocks. In computer simulations, these gaps can sometimes cause the math to crash or become unstable, a problem the authors call the "small-cut problem."

To fix this, the team added a "Ghost Stabilization" technique. Imagine you are walking on a bridge made of planks of different sizes. If a plank is too small or missing, you might fall through. The "ghost" technique is like an invisible safety net placed under those tiny, risky gaps. It doesn't change the physics of the bridge, but it stops the simulation from falling apart when the math gets shaky near the edges of the zoomed-in areas.

The "Smoothie" Blend (GIMP)

Another issue with these simulations is "cell-crossing instability." Imagine a material point (a tiny piece of soil) moving from one grid square to another. In old methods, this movement could cause sudden, jerky jumps in the math, like a video game character glitching.

The authors used a technique called Generalised Interpolation Material Point (GIMP). Think of this as blending the soil particles into a smooth "smoothie" rather than keeping them as distinct, jagged blocks. This smooths out the transitions as the soil moves across the grid, preventing the glitches and making the simulation much more stable, especially for slow, heavy movements like landslides or deep-sea anchors.

The Results: Faster and Greener

The team tested this new "Smart Zoom" method on several problems, including a cone being pushed into sand and an anchor being dragged through soil. They compared it to the old method of using a uniform, high-resolution net everywhere.

The results were impressive:

Speed: The new method was up to 5.5 times faster than the old way.
Energy: Because it was faster and used fewer computer resources, it saved a massive amount of energy. The authors calculated this as a 21-fold reduction in CO2 emissions (carbon footprint) for the same task.

Why This Matters

The paper doesn't claim this will cure diseases or predict the weather. Instead, it solves a specific engineering headache: how to simulate heavy objects moving through soft ground without wasting massive amounts of time and electricity.

By using a "smart zoom" approach that focuses computing power only where it's needed, and adding "ghost" safety nets to keep the math stable, engineers can now run complex simulations of soil and structures much faster and with a much smaller environmental impact. It's a win for both speed and the planet.

Technical Summary: An Implicit Octree-based Adaptive Material Point Method

Problem Statement
The Material Point Method (MPM) is a robust framework for simulating large deformations, particularly in continuum-rigid structure interactions such as soil-structure problems. However, standard MPM implementations rely on structured background meshes, which necessitate high resolution across the entire domain to accurately capture evolving interfaces between rigid bodies and deformable continua. This requirement leads to large numerical systems and excessive computational runtimes, especially in three dimensions. Furthermore, implicit time integration schemes, which are advantageous for long-duration engineering problems (e.g., soil-structure interaction), suffer from poor scaling of sparse linear solvers as the number of degrees of freedom increases. The method also faces well-known numerical instabilities: cell-crossing instability and the "small-cut" problem, where elements are partially cut by boundaries, leading to ill-conditioned systems.

Methodology
This paper presents the first implicit octree-based Generalised Interpolation Material Point Method (GIMPM). The core of the approach is an adaptive background mesh that automatically refines and coarsens based on the position of the rigid body interacting with the continuum.

Octree Discretisation: The background mesh utilizes an octree structure to concentrate computational effort spatially. Elements are refined only in regions requiring high fidelity (near the interface), while coarse elements are retained elsewhere. This introduces hanging nodes (nodes on the face or edge of a coarser neighbor), which break standard $C^0$ continuity.
Discrete GIMP Basis Formulation: To handle the non-conforming octree mesh, the authors formulate a discrete GIMP basis. Instead of constructing complex piecewise functions for all possible intersection variations, the basis functions are computed via a single Gauss point quadrature rule over the intersection of the material point domain and the background elements. This approach naturally incorporates the constraint equations required to link hanging nodes to their supporting vertex nodes, maintaining $C^1$ continuity of the basis functions.
Ghost Stabilisation: The small-cut problem, which causes ill-conditioning in unfitted domains, is mitigated by extending ghost stabilisation techniques to octree meshes. This involves adding a ghost bilinear form to the stiffness matrix across "ghost faces" (interfaces between active and inactive or boundary elements), ensuring stability without the need for aggregation techniques that might complicate the framework.
Adaptive Strategy: The adaptivity is driven by the rigid body's position. The strategy involves three steps: identifying elements intersecting the rigid body, refining elements within a specified distance ( $dx_{region}$ ), and refining the material points (GIMPs) themselves based on the local element size. Material data is transferred from coarse to refined GIMPs, though the authors note that for position-dependent material properties, this transfer relies on reference coordinates rather than an energetic framework.

Key Contributions

First Implicit Octree GIMPM: The paper introduces the first implicit GIMPM utilizing an octree background mesh, enabling efficient solution of large-deformation continuum-structure interaction problems.
Discrete Basis Formulation: A novel discrete integration method for GIMP basis functions on non-conforming meshes, avoiding the need for hierarchical mesh coupling or complex piecewise definitions.
Extended Ghost Stabilisation: The adaptation of ghost stabilisation to octree meshes with hanging nodes to resolve the small-cut instability.
Automatic Adaptivity: A systematic strategy that adapts both the computational mesh and material point discretisation based on the contact interface, eliminating the need for separate derefinement algorithms.

Results
The method was validated through five numerical examples:

Convergence Studies: A 1D column under self-weight and a column under constant contact stress demonstrated convergence of stress solutions with mesh refinement, validating the hanging node formulation and contact penalty approach.
Dynamic Contact: A sphere rolling down a slope validated the friction implementation and contact formulation in the presence of hanging nodes, showing excellent agreement with analytical solutions for various friction coefficients.
Geotechnical Applications:
- Cone Penetration Test (CPT): The adaptive octree mesh successfully reproduced experimental CPT results. The study highlighted that while refinement increases stiffness in this specific setup (due to Young's modulus definition at GIMP centers), the results converged with refinement.
- Anchor Penetration: An anchor dragged through a sand bed was simulated and compared to experimental data and a previous structured-mesh simulation.

Performance and Environmental Impact
In the anchor penetration case, the adaptive octree approach achieved results comparable to or better than a high-resolution conforming mesh.

Speed: The adaptive method achieved up to a 5.5-times speedup compared to the structured mesh simulation.
Efficiency: The reduction in degrees of freedom and material points led to a significant decrease in computational resources.
Carbon Footprint: The paper reports a 21-times reduction in CO2 emissions for the adaptive simulation compared to the structured mesh equivalent, calculated based on High-Performance Computing (HPC) resource usage.

Significance
The paper claims this work represents a significant step forward in the efficiency of implicit MPM for large-deformation problems. By integrating octree adaptivity with the GIMPM, the method focuses computational resolution strictly on the evolving interface, reducing the number of degrees of freedom and material points. This not only accelerates simulations but also offers substantial environmental benefits by lowering the energy consumption required for high-fidelity geotechnical modeling. The authors emphasize that the method requires minimal user input (defining minimum element size and refinement region) and provides a robust framework for complex soil-rigid structure interactions.