Modeling Ostwald Ripening Dynamics in Porous Microstructures

This paper introduces an image-based pore-network model (iPNM) that overcomes the limitations of existing quasi-static models by coupling two-phase flow, solute transport, and Ostwald ripening to accurately simulate the evolution of partially miscible ganglia in porous media, a capability validated against high-resolution microfluidic experiments without adjustable parameters.

The Gist

The Big Picture: The "Bubble Party" in a Sponge

Imagine you have a very complex, bumpy sponge (a porous rock) soaked in water. Inside this sponge, there are tiny bubbles of gas (like hydrogen) trapped in the nooks and crannies.

Over time, these bubbles don't just sit there. They play a game of "musical chairs" called Ostwald Ripening.

• The Rule: Small bubbles are "unhappy" because their surface is curved tightly (high pressure). Large bubbles are "happier" because they are flatter (low pressure).
• The Result: The gas from the small, unhappy bubbles dissolves into the water, travels through the sponge, and re-attaches to the big, happy bubbles. The small ones shrink and vanish; the big ones grow.

Why do we care?
This is crucial for underground hydrogen storage. If we pump hydrogen into a rock formation to store energy, we want to know: Will the gas stay trapped? Will it disappear into the water? Or will it clump together and escape? To predict this, we need to simulate how these bubbles behave.

The Problem: The Old Maps Were Too Simple

Scientists have tried to simulate this before using Pore-Network Models (PNMs). Think of these models as a simplified map of the sponge.

• The Old Way: They treated every hole in the sponge like a perfect sphere or a cube. They also assumed the bubbles were tiny and stayed in just one hole.
• The Flaw: Real rocks aren't perfect spheres. And real bubbles often stretch across multiple holes, like a long worm connecting different rooms in a house. The old models couldn't handle these "worms" or the complex shapes of real rocks. They were like trying to navigate a real city using a map that only has perfect circles for buildings.

The Solution: The "Image-Based" GPS (iPNM)

The authors created a new model called iPNM (Image-based Pore Network Model). Here is how it works, using a simple analogy:

1. The "X-Ray" Vision
Instead of guessing what the holes look like, the iPNM takes a high-resolution photo (an X-ray CT scan) of the actual rock. It doesn't try to force the rock into a perfect shape. It looks at the real jagged, weird shapes of the holes.

2. The "Shape-Shifting" Rules
In the old models, the rules for how a bubble behaves were fixed (like a rigid toy). In iPNM, the rules change based on the specific shape of the hole.

• Analogy: Imagine a video game where the physics engine changes depending on the level. If you are in a round room, the ball bounces one way. If you are in a jagged cave, it bounces differently. iPNM calculates these specific "bouncing rules" for every single hole in the rock.

3. The "Traffic Controller"
The model simulates three things happening at once:

• Flow: How the water moves around the bubbles.
• Transport: How the gas dissolves into the water and moves.
• Ripening: The actual shrinking and growing of the bubbles.
  It acts like a traffic controller, ensuring that when a bubble grows, the water moves out of the way, and when a bubble shrinks, the water rushes in.

The "Aha!" Moments: What the Model Found

The team tested their new model against real experiments where they watched hydrogen bubbles in a sandstone-patterned chip for 24 days. Here is what they discovered:

1. The "Worm" Effect (Multi-pore Ganglia)
Real bubbles often stretch across several holes. The old models couldn't handle this. iPNM could. It showed that these "worms" can snap, merge, or move from one hole to another in ways the old models missed.

• Analogy: It's like realizing that a crowd of people isn't just standing in isolated circles; they are forming long lines that weave through the room.

2. The "Curvature" Surprise
The model found that when a bubble stretches across multiple holes, its shape is more complex than we thought. The old math assumed a simple curve, but the real curve is "bumpy" and changes as the bubble grows.

• Analogy: Think of a balloon. If you blow it up in a round box, it's a perfect sphere. If you blow it up in a box with jagged corners and narrow hallways, it gets squished and weird. iPNM captures that "squishiness."

3. Better than the "Average" Model
They compared iPNM to a "Continuum Model" (which treats the rock like a smooth, average block of cheese).

• The Continuum Model: Good at telling you the total amount of gas left (like knowing the whole cake is half-eaten).
• The iPNM: Good at telling you exactly which crumbs are left, where they are, and how big they are.
• Result: iPNM was much better at predicting the detailed behavior, especially the "pre-equilibrium" phase (the messy middle part before everything settles down).

Why This Matters

This isn't just about bubbles; it's about energy security.

• If we want to store hydrogen underground to power our future, we need to know exactly how long it will stay there.
• If the bubbles dissolve too fast, we lose our energy. If they clump together and escape, we lose our energy.
• This new model gives engineers a much sharper tool to design safer, more efficient storage sites.

Summary in One Sentence

The authors built a super-smart computer simulator that uses real photos of rocks to predict how gas bubbles grow, shrink, and move underground, replacing old, simplified guesses with a detailed, realistic map of the action.

Technical Summary

1. Problem Statement

Ostwald ripening is a thermodynamic process where partially miscible ganglia (trapped bubbles) in a porous medium evolve due to differences in interfacial curvature. High-curvature ganglia dissolve and feed low-curvature ganglia, altering the population statistics and topology of the trapped phase. This phenomenon is critical for applications such as underground hydrogen storage (UHS), geological carbon storage, and fuel cells.

Key Challenges in Existing Models:

• Single-Pore Limitation: Existing Pore-Network Models (PNMs) typically restrict ganglia to single pores, failing to capture the reality where ganglia span multiple pores.
• Geometric Idealization: Traditional PNMs rely on idealized pore shapes (e.g., cubes, spheres) to derive analytical capillary pressure curves, which do not reflect the complex geometry of real rocks.
• Quasi-Static Assumptions: Many models assume instantaneous capillary equilibrium, ignoring the finite timescales of fluid flow and the discrete capillary events (invasion, snap-off, fragmentation, coalescence, dislocation) that occur during ripening.
• Lack of Flow Coupling: Previous ripening models often neglect the coupling between two-phase flow, solute transport, and the ripening process itself.

2. Methodology: Image-Based Pore-Network Model (iPNM)

The authors propose a novel Image-Based Pore-Network Model (iPNM) that overcomes these limitations by operating directly on binarized images of porous microstructures.

A. Network Extraction and Geometry

• Image Processing: The model extracts a computational graph (nodes = pores, links = throats) from a 2D or 3D image using watershed segmentation on a distance map.
• Pore Morphology Method (PMM): Instead of assuming ideal shapes, iPNM computes unique curvature-saturation ($\kappa_b - S_b$) curves for each individual pore using PMM. This captures the specific geometric influence of the pore body on capillarity.
• Pore Classification: Pores are dynamically classified into three types based on their connectivity to the non-wetting phase:
  1. Singleton: Isolated in a single pore.
  2. Terminal: Connected to one invaded throat.
  3. Bridge: Connected to multiple invaded throats.
     Each type has a distinct $\kappa_b - S_b$ curve that adapts as the ganglion topology changes.

B. Governing Physics and Equations

The model couples three balance equations solved via Lie–Trotter operator splitting:

1. Wetting Phase Conservation: Tracks volumetric flow driven by pressure gradients.
2. Non-Wetting Phase Conservation: Tracks bubble saturation changes due to flow and mass transfer.
3. Dissolved Species Conservation: Tracks the diffusion of the non-wetting species (e.g., $H_2$) in the wetting phase, driven by concentration gradients established by the Kelvin equation.

Key Simplifications:

• Timescale Separation: Flow (fast) and mass transfer (slow) are decoupled.
• Dilute Limit: The dissolved concentration is negligible compared to the bubble density, simplifying the transport equation.
• Capillary Stability Check: An iterative algorithm (Algorithm 1) checks for discrete events at every time step:
  • Snap-off: Throat fills with wetting phase if capillary pressure drops below a threshold.
  • Invasion: Ganglion enters a new pore if pressure exceeds entry pressure.
  • Retraction/Fragmentation/Coalescence: Topological changes are handled dynamically.

C. Numerical Implementation

• Solver: Uses backward Euler for time integration and Newton's method for non-linear flow equations.
• Adaptive Time Stepping: The time step adjusts dynamically based on proximity to critical events (e.g., approaching critical saturation or snap-off thresholds).
• Visualization: Simulated saturations are mapped back to the original image using PMM sub-images, allowing for high-fidelity visualization of phase distribution, including connectivity correction for multi-pore ganglia.

3. Key Contributions

1. Removal of Geometric Idealization: iPNM is the first ripening model that requires no idealization of pore shapes, deriving capillary properties directly from the actual microstructure image.
2. Multi-Pore Ganglia Dynamics: It successfully models ganglia spanning multiple pores, capturing complex topological changes (invasion, snap-off, dislocation) that single-pore models miss.
3. Unified Framework: It couples two-phase flow, solute transport, and Ostwald ripening in a single dynamic framework, resolving pre-equilibrium dynamics.
4. Validation against Experiments: The model is validated against high-resolution microfluidic experiments of hydrogen ripening over 15–24 days, a timescale rarely addressed in pore-scale simulations.

4. Results and Validation

The model was validated against microfluidic experiments involving hydrogen ripening in a sandstone-patterned micromodel at 40°C and 80°C.

• Verification: iPNM was first verified against a prior quasi-static PNM, successfully reproducing key phenomena like dissolution-induced snap-off, bubble dislocation, and ganglion growth in heterogeneous networks.
• Macroscopic Agreement: iPNM showed excellent agreement with experimental data for total non-wetting phase saturation ($S_b^{tot}$) and mean curvature ($\bar{r}$) without using any adjustable parameters.
• Microscopic Resolution: Unlike continuum models, iPNM accurately resolved:
  • Curvature Distributions: It captured the shifting probability density functions (PDFs) of ganglion curvatures over time.
  • Spatial Evolution: It correctly predicted the formation of depletion zones near concentration sinks and the spatial distribution of trapped gas.
• Comparison with Continuum Models: While continuum models captured macroscopic saturation trends, they failed to resolve individual ganglion statistics, population distributions, and pre-equilibrium dynamics. The continuum model tended to overestimate mean curvature because it could not account for small, high-curvature, out-of-equilibrium bubbles.

Discrepancies and Limitations:

• Boundary Conditions: Deviations at 80°C were attributed to non-ideal boundary conditions in the experimental channels (irregular bubble shapes acting as stronger sinks than modeled).
• Condensation: The model did not account for water condensation droplets observed at 80°C, which affect local curvature and mass transfer.
• PMM Curvature Bias: For multi-pore ganglia, the PMM slightly overestimates curvature because it erodes pore sub-images independently, ignoring the volume contribution of occupied throats.

5. Significance and Impact

• Theoretical Advancement: iPNM provides a rigorous pore-scale foundation for developing improved theories of ripening in confined geometries, bridging the gap between microscopic physics and macroscopic continuum descriptions.
• Practical Application: The model is computationally efficient compared to Direct Numerical Simulation (DNS) but far more accurate than traditional PNMs. It is suitable for guiding the design of underground hydrogen storage strategies by predicting the long-term fate of trapped gas ganglia.
• Generalizability: The framework is applicable to 3D rock images (from X-ray CT) and can be extended to non-zero contact angles and multi-component gas mixtures, making it a versatile tool for energy and environmental engineering.

In summary, this paper presents a breakthrough in pore-scale modeling by integrating image-based geometry with dynamic multiphase flow and transport, offering a predictive tool for understanding complex interfacial phenomena in porous media.