Trapping and Fracture of Particles in Three-Sphere Ball-Milling Collisions
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.
Kai: We're looking at the paper: "Trapping and Fracture of Particles in Three-Sphere Ball-Milling Collisions".
Mira: Here's the gist of the paper. The study numerically investigates the collision and fracture processes of ball milling based on a minimal model of a three-sphere geometry using discrete element method (DEM) simulations combined with Christensen-type and Weibull-type normalized fracture criteria.
Kai: The outcome of the brittle target particle (ejected versus trapped, fractured versus intact) depends on the impact geometry and velocity.
Mira: Simulation Framework and Model The researchers developed a simulation framework where the collision outcome is separated into two stages that can be analyzed independently.
Kai: The first stage is geometric and frictional, determining whether the target particle is trapped between a pair of grinding balls or ejected before it can be loaded.
Title and authors: Kai: So what they found is that for a fixed size ratio and friction, the most effective way to get fracture is through head-on collisions when the velocity is high enough. They also found that within that trapped regime, the critical velocity needed for fracture is lowest for those head-on impacts and then it rises as the collision angles get bigger.
Mira: That makes sense in a physical sense because you want maximum normal force to cause crushing, and head-on hits give you that directly. They also found that finer particles are easier to fracture once they are trapped, but they're less likely to stay confined between the balls for long.
Lev: That distinction about confinement versus fracture is important for designing systems. If a particle fractures but then gets ejected, the grinding action is less effective in terms of size reduction.
Kai: Exactly, and regarding friction itself, it turns out friction mainly controls the width of that trapping regime, but it doesn't have a big impact on how high the peak compressive stress becomes once you are already trapped.
Mira: That’s a key piece of info because it means you can tune the trapping condition by changing how much friction is there, which is something engineers can actually control in real mills.
The paper's summary: Lev: Now for the proposed improvements, they suggest a few things. First, they propose an automated collision classification system that uses that closed-form geometric condition to decide whether a particle is trapped or ejected based on the impact geometry and friction.
Kai: That sounds like it’s trying to automate the initial decision-making process so you don't have to look at the geometry manually every time. It’s about making that first step faster, right?
Mira: And then they suggest using two different fracture criteria, the Christensen criterion and the Weibull criterion, to evaluate failure. One is based on volume-averaged stress in the particle itself, and the other is based on local contact forces experienced by it.
Lev: That dual approach is interesting for hardware implementation because you could use both metrics to set a safety threshold for fracture, depending on what your sensors can actually measure reliably.
Kai: They also suggest calculating a critical normal force for fracture using the Christensen criterion directly with an equation involving the volume-averaged compressive stress, which avoids having to calculate internal stresses first.
Mira: And they even suggest that you could use AI to predict the required impact velocity for fracture by comparing simulated peak compressive stress maps against boundaries defined by both failure criteria. That moves it from just a simulation to something predictive.
The paper's improvements: Kai: So, wrapping up on "Trapping and Fracture of Particles in Three-Sphere Ball-Milling Collisions," the main implication is that controlling the impact geometry, especially steering towards near head-on hits, is crucial for maximizing fracture probability when you have a fixed size ratio.
Mira: And managing the size distribution of your charge helps too. Keeping a balanced ball mix and continuously removing fine particles through classification can influence how well those particles are trapped and then fractured in the mill.
Lev: From my side, what this paper gives us is a compact mechanistic picture linking single-collision physics to bulk behavior, which is something we really need when trying to build reliable error correction for particle systems.
Kai: It’s a great framework for understanding how to improve grinding efficiency in ball milling applications by tuning the kinematics and material properties of the system.
Mira: Yeah, it gives us concrete rules about when and how particles are likely to fail under these specific collision dynamics.
Lev: The paper sets up a nice path forward by connecting this single-collision physics to measured size-reduction trajectories in real systems, which is exactly what we need next.
Conclusion: Kai: So we looked at the paper "Trapping and Fracture of Particles in Three-Sphere Ball-Milling Collisions," and basically, they’ve created a minimal model—just three spheres—to figure out what happens when balls hit a target particle in a ball mill.
Mira: Right. They use this simulation framework to separate the collision into two stages: first, figuring out if the particle gets trapped or ejected based on geometry and friction.
Lev: And then the second stage is mechanical, checking if that trapped particle actually has enough stress to fracture using two different criteria—a global stress limit and a local crack-based one.
Kai: The main result they show is that for fixed sizes, you get the most effective fracture when the collisions are head-on and happen at a high velocity.
Mira: They also found that within that trapped state, the critical velocity needed to actually cause a fracture is lowest for those direct hits and then it increases as the collision angles get bigger.
Lev: That means if you're trying to design hardware, you want to steer the collisions toward those near head-on impacts to keep the fracture force low enough.
Kai: And they found that friction mainly controls how wide that trapping zone is, but it doesn't really change the peak stress once a particle is already trapped.
Mira: That’s important for tuning things; you can control how much friction there is by changing the surface of the grinding media or by adjusting slurry additives in wet milling.
Lev: If we try to run this on real hardware, we'd need to make sure our sensors can reliably detect those contact forces and stresses within that small three-sphere model.
Kai: So for anyone listening, this paper gives us a clearer picture of how to design the collision kinematics for better grinding efficiency in ball milling processes.
Mira: It shows us that while population balance models are useful for the big picture, you need these kinds of detailed simulations to understand exactly what triggers a fracture at the particle level.
Lev: This work is definitely a stepping stone toward embedding these single-collision physics into larger particle replacement models that predict real size reduction trajectories.
Kai: Next up, we’ll be looking at how those quantum many-body systems behave under extreme conditions, specifically the second law of thermodynamics in closed quantum many-body systems.
Ziyi Cui, Qiong Zhang, Suwei Liu, Richard M. Lueptow
University of Toronto · Northwestern University
cond-mat.soft, physics.app-ph
Submitted: 2026-11-05
Updated: 2026-11-05
Comments: 16 pages, including Supplementary Information; 7 main-text figures and 1 supplementary figure
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 91/100
The gist: The study numerically investigates the collision and fracture processes of ball milling based on a minimal model of a three-sphere geometry using discrete element method (DEM) simulations combined…
Key concepts
- Discrete Element Method (DEM)
- DEM is a simulation technique that treats every particle in the milling system as a separate, discrete entity. It tracks the trajectory, contact forces, and energy dissipation of each particle individually by applying Newton's laws to their motion. This allows researchers to capture the complex mechanics of granular materials during collisions.
- Trapping Condition
- This condition determines if a target particle gets stuck between two grinding balls or is ejected. It is governed by a geometric rule comparing the collision angles ($\alpha$ and $\gamma$) with the friction coefficient ($\mu$). A particle is trapped only if its geometry allows it to be compressed under sustained double contact, defined by a specific mathematical inequality.
- Christensen Criterion
- This is a stress-based fracture criterion that assesses whether the particle will break based on its overall average compressive stress. It calculates a critical force ($F_{Cc}$) needed to cause failure by setting the volume-averaged compressive stress ($\sigma_C$) to a specific material limit, assuming the material is uniform.
- Weibull Criterion
- This is a local, flaw-based fracture criterion that predicts breakage based on random crack development. It looks at the local tensile stresses inside the particle caused by compression and determines failure based on these internal stress patterns rather than just overall average stress.
Terminology
Summary
The study numerically investigates the collision and fracture processes of ball milling based on a minimal model of a three-sphere geometry using discrete element method (DEM) simulations combined with Christensen-type and Weibull-type normalized fracture criteria. p. 0 The outcome of the brittle target particle (ejected versus trapped, fractured versus intact) depends on the impact geometry and velocity. p. 0 p. 0
Simulation Framework and Model
The researchers developed a simulation framework where the collision outcome is separated into two stages that can be analyzed independently. The first stage is geometric and frictional, determining whether the target particle is trapped between a pair of grinding balls or ejected before it can be loaded. p. 0 The second stage is mechanics-based, evaluating whether a trapped particle experiences sufficient compressive loading to fracture. p. 0 A three-sphere collision system serves as a minimal model to represent the local environment of a ball mill while allowing precise control over impact velocity v, impact angle α, obliquity angle γ, size ratio R, and friction coefficient µ. p. 0
Contact and Trapping Conditions
The trapping regime follows a closed-form geometric condition that generalizes the classical angle-of-nip criterion of roll-crusher and high-pressure grinding-roll theory to a three-sphere geometry. The trapping condition is determined by comparing a purely geometric angle, θ = (α + γ)/2, with the half-angle of the Coulomb friction cone, ϕ = arctan µ. p. 3 The target particle is compressed under sustained double contact only if both the contact condition of Equation (2) and the frictional condition of Equation (5) are satisfied. p. 3 This leads to the trapping regime being defined by α + γ ≤ min 2 arctan µ, 2 arccos R / R + 1. p. 3
Fracture Criteria and Stress Analysis
The study evaluates fracture within the trapping regime using two failure criteria: a global stress-based limit (Christensen criterion) and a local, flaw-based criterion (Weibull criterion). The Christensen failure function is based on the stress tensor averaged over the entire particle, assuming homogeneity of the material. p. 5 The Weibull failure function is based on random crack development due to local tensile stress inside the particle under compression. p. 5 The critical force for fracture based on the Christensen criterion is determined by setting the volume-averaged compressive stress to σC, yielding F C c = 2/3 π r squared t σC. p. 7
Key Findings on Collision Effects
The results demonstrate that for fixed sizes and friction, the most effective fracture occurs in head-on collisions at sufficiently high velocity. Within the trapping regime, the critical velocity is lowest for head-on impacts and rises as the collision angles grow. p. 9 Finer particles fracture more easily once trapped but are less likely to remain confined, and friction mainly determines the width of the trapping regime while having little role in the magnitude of σpeak when a particle is trapped. p. 9
Comparison of Fracture Criteria
The Christensen and Weibull criteria for fracture are qualitatively similar. The crossover between the two criteria occurs at R ≈ 4. p. 96, consistent with the reversal between the Christensen and Weibull boundaries between R = 4 and R = 6 in Fig. p. 9 7. The Christensen threshold F C c applies to the normal component at either contact, F n = F cos θ, whereas the Weibull threshold F W c(R) applies to the resultant F along the na axis. p. 9
Practical Implications
In practice, fracture probability is raised by steering collision kinematics toward near-head-on, lowoffset impacts and by managing the size distribution of the charge. The former can be achieved by operating at rotation speeds that promote cataracting and by lifter and liner designs that favor direct impacts; the latter by keeping a balanced ball mix and continuously removing fine particles through classification. p. 4 The friction that holds the trapping regime open can be tuned through the surface condition of the grinding media or, in wet grinding, the slurry and its additives. p. 5 The three-sphere model presented here is clearly a minimal model. p. 0 In future work, it would be reasonable to follow Professor Mojtaba Ghadiri’s approach of uncovering fundamental links between particle-scale interactions and bulk particle behavior in real systems. p. 0 A particle replacement model can then be embedded, connecting the single-collision physics established here to measured size-reduction trajectories. p. 0
--- Page 1 ---
Trapping and Fracture of Particles in Three-Sphere Ball-Milling Collisions Ziyi Cui, Qiong Zhang, Suwei Liu, and Richard M. Lueptow 2 University of Toronto Schools, Toronto, Ontario M5S 2R7, Canada 2 Department of Mechanical Engineering, Northwestern University, Evanston, Illinois 60208, USA 3 Department of Chemical and Biological Engineering, Northwestern University, Evanston, Illinois 60208. p. 0
Abstract. We numerically investigate the collision and fracture processes of ball milling based on a minimal model of a three-sphere geometry using discrete element method (DEM) simulations combined with Christensen-type and Weibull-type normalized fracture criteria. p. 0 The outcome of the brittle target particle (ejected versus trapped, fractured versus intact) depends on the impact geometry and velocity. We develop a reduced three-sphere collision description in which trapping is governed by impact geometry and friction, while fracture inside the trapping regime is evaluated from DEM output using stress-based and contact-force-based normalized criteria. p. 0 The trapping regime follows a closed-form geometric condition, while the critical conditions for fracture vary with impact angle, obliquity angle, impact velocity, and particle size ratio. p. 0 Head-on collisions with larger size ratios require lower critical impact velocities for fracture. p. 0
Keywords: ball milling; discrete element method; fracture; trapping condition; Weibull theory; Christensen criterion
--- Page 2 ---
1 Introduction Grinding is a critical process applied in a wide range of industries, including mineral processing [1], cement [2], ceramics [3], food [4], pharmaceuticals [5], and the beneficiation of rare-earth metals [6]. Despite its broad application, ball milling is among the least efficient industrial operations [7]. p. 0 Comminution alone consumes up to 4% of global industrial electricity [8] and approximately 36% of a mine’s total power usage [9], yet less than 1% of the input energy actually produces new surface area, with the remainder dissipated as heat, noise, and vibration [10–12]. p. 0 This inefficiency raises operating costs and magnifies environmental impacts, including increased carbon emissions [13–15]. p. 0 Because grinding is so widely adopted, even modest efficiency improvements would translate into substantial energy and cost savings while reducing greenhouse gas emissions. p. 0 Among his many contributions to the field of particle and powder technology, Mojtaba Ghadiri has studied various aspects of milling operations and particle mechanics [16–24]. p. 0 To connect the physics of individual collisions with large-scale milling, previous studies have used experiments to measure breakage. p. 0 Single-particle indentation [20], impact [17, 25], and drop-weight experiments [26] measure the strength and probability of breakage under controlled loading. p. 0 These studies reveal real material behavior but cannot practically span the wide range of impact geometries, size ratios, and frictional conditions seen in industrial mills, and scaled-up experiments are costly, time-intensive, and often difficult to reproduce. p. 0 Theoretical approaches such as population balance models describe particle size distribution changes across milling times and are computationally efficient [12], but they rely heavily on empirical fitting and do not resolve the mechanics of individual collisions, which are the actual triggers of fracture. p. 0 In short, experiments provide realism but narrow scope, while population balance models provide generality but limited physical detail. p. 0 Numerical simulation offers a way to overcome some of these limitations. p. 0 The discrete element method (DEM) has become one of the most versatile tools for studying granular materials because it explicitly resolves particle trajectories, contact forces, and energy dissipation by treating each particle as a discrete entity and integrating its motion under Newton’s laws. p. 0 Unlike population balance models, DEM does not require empirical fitting to capture local collision mechanics; unlike experiments, DEM can systematically sweep impact velocity, angles, size ratio, and friction over a range of values that is difficult to achieve in the laboratory. p. 0 At the same time, simulation outcomes are most powerful when interpreted together with theoretical models that encode the governing physics. p. 0 This study develops a simulation framework in which the collision outcome is separated into two stages that can be analyzed independently. p. 0 The first stage is geometric and frictional, concerning whether the target particle is trapped between a pair of grinding balls or ejected before it can be loaded. p. 0 The second stage is mechanics-based, assessing whether a trapped particle experiences sufficient compressive loading to fracture. p. 0 A three-sphere collision system is used as a minimal model to represent the local environment of a ball mill while allowing precise control of impact velocity v, impact angle α, obliquity angle γ, size ratio R, and friction coefficient µ. p. 0 We show that trapping is governed by a closed-form geometric condition that generalizes the classical angle-of-nip criterion of roll-crusher and high-pressure grinding-roll theory to a three-sphere geometry. p. 1 We then evaluate fracture within the trapping regime with two failure criteria applied to the DEM output, one a global stress-based limit and the other a local, flaw-based criterion: a Christensen failure function based on stresses in the target particle and a Weibull failure function based on the contact forces experienced by the target particle. p. 1 By combining DEM with analytic contact geometry, this study provides both detailed numerical results and a compact mechanistic picture that sheds light on improving grinding efficiency in a broad range of ball milling applications. p. 1
--- Page 3 ---
2 Geometry of Impacts Fig. 1 shows the planar three-sphere collision studied here. p. 1 Two identical grinding balls have radius rg, and the smaller target particle has radius rt, resulting in a size ratio R = rg/rt. p. 1 One grinding ball is fixed, so it can neither translate nor rotate, and centered at Cf, initially tangent to the target particle at Ct. p. 1 The incoming grinding ball, centered at Ci, begins a short distance away and travels with initial in-plane velocity v until it collides with the target particle. p. 1 Simultaneous contacts, Cf − Ct and Ci − Ct, result in the contact forces F1 and F2 along the corresponding axes, the lines of centers: the n1 axis passes through Cf and Ct, and the n2 axis through Ci and Ct. p. 1 The impact angle α is measured between −v and the n1 axis (α ∈ [−90◦, 90◦]), and the obliquity angle γ is between v and the n2 axis (γ ∈ [0◦, 90◦]). p. 1 Depending on the impact geometry, the target particle is either not contacted at all, ejected from between the two grinding balls, or trapped between them. p. 1 In the trapped case, the target particle is captured between the incoming grinding ball and fixed grinding ball so that it is compressed until the incoming grinding ball rebounds (Fig. p. 1 2(a)). If the compressive stress is high enough, the target particle will be crushed and break up into smaller pieces. p. 1 At lower compressive stresses, the target particle will remain intact and is released, as shown in the figure. p. 1 Changing the impact and obliquity angles so that the incoming grinding ball hits the target particle less directly results in the target particle being ejected as it is squeezed out between the two balls (Fig. p. 1 2(b)). The ejection depends on friction at contact points between the grinding balls and the target particles. p. 1
--- Page 4 ---
2.1 Contact Condition Whether the incoming grinding ball can reach the target particle is determined solely by the opening angle α + γ and the size ratio, with a limiting case shown in Fig. p. 2 3(a). Contact with the target occurs before contact between the two grinding balls only if 180◦ − α + γ ≥ 2 arccos rg / (rg + rt), or equivalently, in terms of R = rg/rt α + γ ≤ 2 arccos R / (R + 1). p. 2 For larger values of α + γ, the incoming grinding ball hits the fixed grinding ball before it hits the target particle, and the outcome is no contact between the incoming grinding ball and the target particle. p. 2
--- Page 4 ---
2.2 Trapping Condition Whether the target particle is trapped between the two grinding balls or ejected before it can be sufficiently loaded is determined by the collision geometry and the friction coefficient. p. 2 The construction described below generalizes the classical angle-of-nip condition of roll-crusher and highpressure grinding-roll theory [33–36], in which a particle is captured only if the half-angle subtended at the two loading surfaces does not exceed the friction angle, ϕ. p. 2 Here, the condition involves two independent angles and emerges in the combination α + γ ≤ 2ϕ. p. 2 The trapping condition can be met only if a single resultant direction can lie within ϕ of both normals, i. p. 2e., only if θ ≤ ϕ = arctan µ. p. 2 The trapping condition therefore compares a purely geometric angle, θ = (α + γ)/2, set by the collision geometry through α and γ, with the half-angle of the Coulomb friction cone, ϕ = arctan µ. p. 3 The target particle is compressed under sustained double contact (trapped) only if both the contact condition of Equation (2) and the frictional condition of Equation (5) are satisfied. p. 3 Combining the frictional bound α + γ ≤ 2 arctan µ with the contact condition of Equation (2) yields the trapping regime α + γ ≤ min 2 arctan µ, 2 arccos R / R + 1. p. 3
--- Page 5 ---
3 Simulation Methods With this background, we now simulate the simple collision of an incoming grinding ball on a stationary target particle already in contact with a fixed grinding ball (Fig. 1) over the parameter ranges v ∈ [0. p. 35, 5] m/s (varied in steps of 0. p. 31 m/s), typical of ball mills [37, 38]. p. 3 The three-sphere dynamics are simulated with the open-source LAMMPS package [39]. p. 3 Contacts are modeled using Hertz’s model with a contact restitution coefficient in the normal direction and the history-dependent Mindlin model limited by Coulomb sliding friction in the tangential direction. p. 3 The equations of motion are integrated with the synchronized velocity-Verlet scheme with a timestep of ∆t = 1. p. 30 × 10−7 s, resolving each binary contact with hundreds of timesteps over the investigated material and size ranges. p. 3 Output is recorded every 100 timesteps; at this interval, the recovered force and stress peaks are converged to within 0. p. 313%. For cases where the incoming grinding ball collides with the target particle, we obtain the peak volume-averaged compressive principal stress, σpeak, as follows. p. 3 The volume-averaged stress tensor of the target particle is defined as [46, 47]. p. 3
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Automated collision classification system can determine whether a target particle is trapped or ejected based on impact geometry and friction using
a closed-form geometric condition
derived from Equation (6). -
Fracture prediction modules can utilize two independent criteria, the Christensen criterion and the Weibull criterion, to assess failure by comparing the volume-averaged stress tensor with material properties or contact forces.
-
The system can calculate a critical normal force for fracture based on the Christensen criterion using Equation (18), providing a direct method to determine if a collision exceeds this threshold without calculating internal stresses.
-
The AI can predict the required impact velocity for fracture by comparing simulated peak compressive stress maps with the boundaries defined by both failure criteria, such as
the dashed black curve indicates the fracture boundary for the Christensen criterion (fC = 1, Equation (10)).
-
The system can analyze how friction controls trapping effectiveness in real-time simulations by observing how
the high-stress region broadens with increasing µ following the trapping boundary ±2 arctan µ, according to Equation (5).
Related papers
- Coarse-graining dense, deformable active particles
- Three-body corrections to the Brownian viscosity of hard spheres without hydrodynamic interactions
- Thin active nematohydrodynamic layers: asymptotic theories and instabilities
- Origins of suppressed self-diffusion of nanoscale constituents of a complex liquid
- Soft material mechanical property degradation at high strain rates
- Avalanche criticality emerges by thermal fluctuation in a quiescent glass