Quantifying Injection-Driven Mass Transfer within Porous Media via Time-Elapsed X-ray micro-Computed Tomography

This study evaluates three analytical frameworks for quantifying injection-driven mass transfer in porous media using time-lapse X-ray micro-CT, introducing a volume-ratio filtering technique to mitigate dissolution-driven biases and demonstrating that while all methods yield comparable average mass transfer coefficients, the choice of approach ultimately depends on the trade-off between desired physical detail and available computational resources.

The Gist

The Big Picture: Watching Bubbles Disappear

Imagine you have a jar filled with sand and water, and you've trapped some hydrogen gas bubbles inside the sand. Now, imagine you start pumping fresh water through the jar. Over time, the gas bubbles will dissolve into the water and disappear.

Scientists want to know how fast this happens. This is called "mass transfer." It's crucial for things like cleaning up polluted groundwater or storing energy (like hydrogen) underground.

To watch this happen, the researchers used a special 3D X-ray camera (called micro-CT) that acts like a super-powered time-lapse movie camera. They took hundreds of pictures as the water flowed through, watching the bubbles shrink and vanish.

The Problem: How Do We Measure It?

The tricky part is that the X-ray camera can see the bubbles, but it can't "taste" the water to see how much gas has dissolved in it. So, the scientists had to use math to guess the speed of dissolution based only on how the bubbles changed size.

The paper compares three different ways (or "methods") to do this math. Think of it like three different chefs trying to guess how fast a cake is baking, but they can only look at the oven door, not the cake itself.

Method 1: The "Average Crowd" (Slice-Averaged or SAC)

• The Analogy: Imagine looking at a crowd of people through a foggy window. You can't see individuals, but you can see the "fog" getting thicker in one spot and thinner in another.
• How it works: This method looks at the whole jar in slices (like slicing a loaf of bread). It calculates the average amount of gas lost in each slice and assumes the water is perfectly mixed in that slice.
• Pros: It's fast and easy to compute. It gives you a good "big picture" number.
• Cons: It's a bit blurry. It smooths out all the details. It's like saying "the crowd is moving," but you can't tell if one specific person is running or walking.

Method 2: The "Count Every Bubble" (Non-Classified Per-Cluster or NPC)

• The Analogy: Imagine you are a detective counting every single person in the crowd, regardless of what they are doing. You count people walking, running, standing still, and even people who seem to be teleporting.
• How it works: This method tracks every single bubble individually. It measures how much every bubble changed size and averages them all together.
• Pros: It uses a lot of data.
• Cons: It gets confused by "noise." Sometimes bubbles don't just dissolve; they bump into each other, merge, or break apart (remobilization). This method counts those messy events as "dissolving," which can mess up the math. It's like counting a person who just walked into the room as part of the crowd that left.

Method 3: The "Smart Detective" (Classified Per-Cluster or CPC)

• The Analogy: This is the detective who is very picky. They only count the people who are definitely leaving the building and ignore everyone else. They know that if someone is merging with a group, they aren't "leaving" yet.
• How it works: This method also tracks every bubble, but it's smarter. It filters out the "messy" events (like bubbles merging or breaking). It only calculates the speed for bubbles that are clearly and completely dissolving.
• Pros: It gives the most detailed picture. It can show you exactly where the water is dissolving the gas (the "dissolution front").
• Cons: It is very hard work (computationally expensive) and requires a lot of computer power. Because it looks at fewer bubbles (only the "perfect" ones), it can be sensitive to small errors.

The Big Discovery

The researchers ran all three methods on the exact same data (hydrogen bubbles in plastic sand). Here is what they found:

1. They all agreed on the "Big Number": Even though the methods were very different, they all estimated the overall speed of dissolution to be roughly the same (within the same "order of magnitude"). If you just need a general answer, the easy method (SAC) works fine.
2. They disagreed on the "Details": When they tried to map out exactly how the gas concentration changed in different spots, the methods diverged.
   • The Easy Method (SAC) smoothed everything out too much, making the water look more diluted than it really was.
   • The Smart Detective (CPC) could see the "front" of the water moving through the sand, showing that the gas dissolves unevenly. It even spotted where bubbles were getting pushed around (remobilized) rather than dissolving.
3. The "Remobilization" Issue: They discovered that sometimes, instead of dissolving, bubbles would get pushed by the water and merge with other bubbles. If you don't filter this out, you think the gas is dissolving faster than it actually is. The "Smart Detective" method was best at spotting and removing these errors.

The Takeaway: Which Tool Should You Use?

The paper concludes that there is no single "best" method. It depends on what you need:

• If you are on a budget (or have a slow computer) and just need a general estimate of how fast the gas dissolves, use the Easy Method (SAC). It's quick and gives a decent average.
• If you need to understand the complex physics (like how the water front moves, or where bubbles are getting stuck), you need the Smart Detective (CPC). It requires more computing power and careful work, but it reveals the hidden details that the other methods miss.

In short: You can get a quick, rough sketch of the scene with a cheap camera, but if you want to see the fine details of the action, you need a high-end camera and a lot of editing time. Both give you a picture, but one tells a much richer story.

Technical Summary

1. Problem Statement

Understanding interphase mass transfer (specifically gas dissolution into a liquid solvent) within porous media is critical for applications ranging from groundwater remediation to geologic energy storage (e.g., hydrogen or carbon sequestration). While X-ray micro-Computed Tomography (µCT) allows for non-destructive, in situ visualization of these processes, analyzing the data to quantify mass transfer coefficients ($k$) is challenging due to:

• Resolution Trade-offs: Balancing the spatial resolution needed to see pore-scale features against the temporal resolution required to capture dynamic processes.
• Analytical Limitations: Existing methods often rely on assumptions that obscure microscopic details or fail to account for complex phenomena like cluster remobilization (where gas bubbles move or merge rather than dissolve).
• Lack of Comparative Framework: There is no unified framework comparing different analytical approaches (Slice-Averaged vs. Per-Cluster) on the same dataset to determine their respective accuracies and computational costs.

2. Methodology

The study utilized a re-analysis of existing time-lapse µCT data from Patmonoaji et al. (2023), focusing on the dissolution of trapped Hydrogen ($H_2$) gas in a water-wet granular plastic packing. The experiment involved four solvent injection rates (0.10, 0.25, 0.50, and 1.0 mL/min).

The authors applied and compared three distinct analytical frameworks:

1. Slice-Averaged Concentration (SAC) Approach:

   • Basis: Uses a 1-D Advection-Diffusion equation.
   • Method: Calculates the average aqueous solute concentration across a cross-sectional slice based on the change in gas saturation over time.
   • Assumption: Assumes pseudo-steady state and radial uniformity; treats the system as a 1-D column.

2. Non-Classified Per-Cluster (NPC) Approach:

   • Basis: Tracks individual gas clusters ("bubbles") across time intervals.
   • Method: Applies mass transfer equations to all observed volume changes (growth, shrinkage, dissolution) without filtering.
   • Assumption: Relies on the "Law of Large Numbers" to average out noise and non-mass-transfer events (like mobilization) to estimate bulk parameters.

3. Classified Per-Cluster (CPC) Approach:

   • Basis: Tracks individual clusters but applies strict morphological classification.
   • Method: Categorizes clusters by their evolution (e.g., completely dissolved, partially dissolved, grown, snapped-off). Mass transfer calculations are restricted to clusters that completely dissolved, assuming these events occur at the "dissolution front" where the concentration gradient is maximal.
   • Refinement: Introduces a volume-ratio filtering technique to exclude time intervals where cluster remobilization (merging/growth) dominates over dissolution. This ratio ($\Delta V_{gained} / |\Delta V_{lost}|$) is used to flag and remove intervals with high mobilization bias.

Key Technical Innovation: The study introduced a volume-ratio filter to mitigate bias from dissolution-driven cluster remobilization, ensuring that mass transfer estimates reflect true dissolution events rather than physical displacement or merging of bubbles.

3. Key Contributions

• Unified Comparison: First study to apply SAC, NPC, and CPC methodologies to the exact same time-lapse µCT dataset, allowing for a direct comparison of their outputs.
• Mobilization Filtering: Development and application of a quantitative threshold (volume gain/loss ratio $\leq 0.2$) to identify and remove time intervals contaminated by cluster remobilization, significantly improving data fidelity.
• Resolution vs. Cost Analysis: A comprehensive evaluation of the trade-off between computational intensity and the level of physical detail (macroscopic vs. microscopic) provided by each method.
• Visualization of Mesoscale Phenomena: Demonstrated that the CPC approach can visualize complex, non-uniform phenomena, such as the propagation of the dissolution front and localized cluster remobilization, which are averaged out in other methods.

4. Results

• Mass Transfer Coefficients ($k$):

  • All three approaches yielded mass transfer coefficient estimates within one order of magnitude of each other for the same injection rate.
  • Trends: Generally, $K_{CPC} > K_{NPC} > K_{SAC}$ at lower injection rates. However, at the highest injection rate (1.0 mL/min), the NPC estimate surpassed the others.
  • Dimensionless Analysis: When converted to Sherwood ($Sh$) and Reynolds ($Re$) numbers, the differences between approaches decreased proportionally as the injection rate increased.

• Concentration Profiles:

  • SAC: Produced dilute concentration profiles due to the assumption of radial uniformity (dilution effect). It failed to capture spatial heterogeneity.
  • NPC: Produced highly variable distributions with many values exceeding physical bounds ($C/C_{sol} > 1$ or $< 0$) because it included "grown" clusters (mobilization) in the calculation, violating the assumption of a dilute system.
  • CPC: Produced concentration values closer to physical bounds but still yielded some negative values. These negative values were spatially localized to the leading edge of the dissolution front, indicating that the front is underestimated at higher injection rates.

• Impact of Filtering:

  • Mobilization filtering significantly altered the results for the CPC approach (removing up to ~40% of volume in some sequences) but had a smaller impact on NPC and SAC when applied uniformly.
  • Filtering was essential to prevent the overestimation of mass transfer caused by assuming all volume loss was due to dissolution.

5. Significance and Implications

• Method Selection Framework: The study provides a decision matrix for researchers:
  • Choose SAC for computationally efficient, macroscopic system parameters (e.g., bulk $k$) where microscopic detail is not required.
  • Choose NPC for a balance of detail and statistical robustness, provided the system has a high number of clusters to average out noise, though it struggles with complex mobilization.
  • Choose CPC for high-fidelity analysis of complex phenomena (e.g., dissolution fingering, front propagation) and when the ability to filter specific physical events (remobilization) is critical, despite higher computational costs and sensitivity to outliers.
• Data Quality Awareness: The results highlight that µCT-based mass transfer estimates are highly sensitive to image processing pipelines and the specific assumptions made about cluster behavior.
• Future Directions: The findings suggest that while different methods converge on bulk parameters, they diverge significantly when resolving spatial profiles and complex dynamics. Future work requires standardized comparisons across different porous media and gases to fully constrain the variability observed.

In conclusion, the paper establishes that while all three methods can estimate bulk mass transfer coefficients with similar accuracy, the CPC approach offers superior resolution of physical mechanisms and complex dynamics, provided that rigorous filtering (like the volume-ratio filter) is applied to handle cluster remobilization.