Topological cell-openness index for porous materials

This paper proposes a topological cell-openness index ($\tau$) based on Betti numbers as a complementary or alternative metric to gas pycnometry for characterizing the open versus closed cell proportions in porous materials, while also demonstrating its correlation with physical quantities and its utility in estimating feature sizes.

The Gist

Imagine you have a sponge. Some sponges are full of holes that all connect to the outside, letting water flow right through. Others have holes, but many of them are trapped inside, like little bubbles sealed in glass, so water can't get in or out.

For a long time, scientists have had a standard way to measure how "open" a sponge is. They call it gas pycnometry. Think of this like blowing into the sponge with a straw. If the air can get in, the hole is "open." If the air can't get in, the hole is "closed." This method gives you a single number: the percentage of open space. It's the industry gold standard.

However, the authors of this paper, Michał Bogdan and Paweł Dłotko, noticed a problem. Imagine a sponge where 99% of the holes are open to the outside, but the remaining 1% are actually a bunch of tiny, isolated bubbles trapped inside the open network. The standard "blow into it" test would say, "Great! It's 99% open!" and stop there. It misses the fact that the open part is actually a messy, disconnected web rather than one smooth highway.

To fix this, the authors created a new tool called the Cell-Openness Index (τ).

The New Tool: Counting Loops and Islands

Instead of just blowing air, the authors use a branch of math called Topological Data Analysis. You can think of this as a super-smart way of counting shapes and connections in a 3D picture of the material.

They use a concept called Betti numbers, which sound complicated but are actually just counters for specific shapes:

• Counting Islands (0D): How many separate chunks of holes are there?
• Counting Loops (1D): How many rings or donut-shapes can you make by walking through the holes?
• Counting Caves (2D): How many completely enclosed bubbles are there?

The authors combine these counts into their new index, τ.

• If τ is 0, the material is like a bag of marbles: every hole is a separate, closed island. Nothing connects.
• If τ is 1, the material is like a perfect honeycomb: every hole is connected to every other hole in one giant, open network.

Why is this better than the old way?

The paper shows that while the old method (gas pycnometry) and the new method (τ) usually agree, they sometimes disagree in a very interesting way.

Imagine two sponges that both test as "99% open" by the old method.

• Sponge A is a perfect, interconnected web.
• Sponge B looks like a web, but it's actually made of 50 separate webs that are all touching the edge of the sponge but not touching each other.

The old method sees both as "99% open." The new method (τ) sees Sponge A as "very open" (high score) and Sponge B as "less open" (lower score) because it detects that the network is broken into disconnected pieces. It's like the difference between a city with one giant highway system and a city with 50 separate cul-de-sacs that happen to all touch the city limits.

Reading the "Fingerprint" of the Material

The authors also found that by looking at how these shape-counts change as they "zoom in" and "zoom out" on the image (a process called filtration), they can guess the physical size of the holes.

Think of it like listening to a song. If you know the rhythm and the notes, you can guess the size of the instruments playing them.

• They found that the "peaks" and "valleys" in their shape-counting graphs correspond to the size of the holes, the distance between holes, and the thickness of the solid walls between them.
• This worked very well for materials with closed, isolated holes (like a block of Swiss cheese where the holes don't touch).
• It was a bit trickier for open, messy networks, but still provided useful clues.

Does it matter for real life?

The authors tested if their new number (τ) could predict how well a material moves heat or fluids.

• Fluids (Permeability): In 2D models, they found a very strong, clear relationship between their new index and how easily fluid flows through the material.
• Heat (Thermal Conductivity): In 3D models, their new index was slightly better at predicting how well heat moves through the material compared to the old method.

The Bottom Line

The paper doesn't claim this will cure diseases or build new rockets immediately. Instead, it proposes a simple, math-based "second opinion" for measuring porous materials.

If you are analyzing a sponge, a rock, or a foam, the old method tells you how much space is open. The authors' new method tells you how well that open space is connected. They suggest that whenever you have a high-quality 3D picture of a material, you should report both numbers: the old one (for tradition) and the new one (to catch the hidden, disconnected bits that the old method misses).

Technical Summary

Technical Summary: Topological Cell-Openness Index for Porous Materials

Problem Statement
Characterizing the openness of porous materials is critical for understanding transport properties, yet current standards rely heavily on gas pycnometry. This experimental method yields a single parameter, the "fraction of open pores" ($\phi_0$), representing the volume fraction of pores connected to the sample boundary. However, $\phi_0$ has limitations: it requires knowledge of the material's true density, struggles with samples where open pores are not surface-connected, and cannot resolve internal structural disconnects within an otherwise "open" phase (e.g., multiple components touching the boundary but disconnected from each other). With the increasing accessibility of 3D microimaging (e.g., X-ray micro-CT), there is a need for complementary, image-based methods that capture topological connectivity information beyond simple volume fractions.

Methodology
The authors propose an image-based method utilizing Topological Data Analysis (TDA), specifically persistent homology, to estimate the proportion of open and closed cells in binary 2D and 3D voxelized microstructures.

1. Data Generation and Sources: The study utilizes synthetic binary microstructures generated via a stochastic process involving spherical/circular pores on a grid, connected by channels with varying probabilities ($p$). Four families of datasets were created: partially-connected holes (2D/3D), purely closed cells (2D), purely open cells (2D), and mixed 3D cells for heat-transfer analysis. Additionally, an externally provided 2D dataset of 31 porous images with known permeability values was used.
2. Numerical Pycnometry ($\phi_0$): A digital analogue of gas pycnometry was implemented by identifying connected components of the pore phase touching the domain boundary and calculating their volume fraction relative to the total pore volume.
3. Topological Analysis: The authors compute Betti numbers ($\beta_0, \beta_1, \beta_2$) on sub-level sets of the Signed Distance Transform (SDT) of the binary structure.
   • $\beta_0$: Number of connected components.
   • $\beta_1$: Number of independent loops (1D) or tunnels.
   • $\beta_2$: Number of enclosed cavities (in 3D).
   • The analysis filters out noise by retaining only features with persistence $\ge 1.5$.
4. The Cell-Openness Index ($\tau$): A new scalar index $\tau \in [0, 1]$ is defined based on the Betti numbers at a specific filtration value ($t=-1$):
   • For 2D: $\tau = \frac{\beta_1}{\beta_0 + \beta_1}$
   • For 3D: $\tau = \frac{\beta_1 + \beta_2}{\beta_0 + \beta_1 + \beta_2}$
   • $\tau = 0$ indicates a system of disconnected, closed pores; $\tau = 1$ indicates a fully connected, open network.
5. Physical Simulations: Effective thermal conductivity was computed for 3D structures by solving the steady diffusion equation with piecewise isotropic conductivity. Permeability data was taken from the external reference dataset.

Key Contributions and Results

• Definition of $\tau$: The paper introduces $\tau$ as a topological complement to $\phi_0$. While $\phi_0$ and $\tau$ are highly correlated, $\tau$ captures connectivity nuances that $\phi_0$ misses. Specifically, in regimes where $\phi_0$ saturates near 1 (indicating high openness), $\tau$ can discriminate between a truly well-connected network and one composed of multiple components that individually touch the boundary but are mutually disconnected.
• Correlation with Physical Properties:
  • 2D Permeability: In a dataset of 31 porous images, $\tau$ showed a strong parabolic relationship with $\log_{10}(\text{permeability})$ ($R^2 = 0.95$), despite $\tau$ varying by less than 1% across the dataset.
  • 3D Thermal Conductivity: Across 250 synthetic 3D structures, $\tau$ showed a moderate positive correlation with the trace of the effective thermal conductivity tensor ($\text{tr}(K)$). In all sub-datasets, $\tau$ was a slightly better predictor of $\text{tr}(K)$ than $\phi_0$.
• Length-Scale Estimation via Betti Curves: The authors demonstrate that Betti curves ($\beta_i(t)$) derived from SDT filtrations can recover characteristic geometric lengths:
  • Closed Cells: The points of maximum gradient in $\beta_0$ and $\beta_1$ successfully estimated mean pore radius ($r$), inter-pore half-width ($d/2$), and solid core half-thickness ($h/2$) with high accuracy ($R^2$ up to 0.95).
  • Open Cells: While the method works for idealized periodic structures, the predictive power for throat width ($w/2$) in noisy, realistic open-cell datasets was weaker, likely due to overlapping length scales obscuring topological signatures.

Significance and Claims
The authors position this work as a "deliberately reductive" attempt to determine if a single, interpretable number ($\tau$) can complement the classical gas pycnometry metric ($\phi_0$). They do not claim $\tau$ replaces pycnometry but rather that it offers a topological perspective that resolves structural ambiguities invisible to volume-based measurements.

The paper claims that:

1. $\tau$ provides genuine structural information regarding internal connectivity that $\phi_0$ cannot resolve.
2. The mismatch between $\tau$ and $\phi_0$ in highly open systems is not an error but a feature indicating disconnected sub-networks.
3. Betti curves offer a pathway to estimate characteristic feature sizes (pore radius, throat width) directly from topological signatures, though this is more robust in closed-cell systems.
4. $\tau$ correlates with transport properties (permeability and thermal conductivity), suggesting it encodes physically meaningful structural signatures.

The authors conclude that where high-quality 3D imaging is available, $\tau$ should be reported alongside $\phi_0$ to provide a more complete characterization of porous material microstructures.