Coffee Data Science
A Peer Review of the Origin of the Fines Migration Theory
Reviewing a 30 year old paper
In discussing fines migration, all roads lead back to a single paper whose experiment is seen as the proof of the theory of fines migration. That paper is called “Espresso Coffee Brewing Dynamics: Development of Mathematical and Computational Models” by Petracco and Liverani [1]. I finally found a copy, and I have some serious questions.
It is important to understand the author’s methods because so much has been based on what they concluded with their experiment and mathematical model of espresso. When we look at data without proper context, we can see what isn’t there. This is inherently a major challenge in the field of data science where one can intentionally or unintentionally misconstrue data.
First, I will summarize what the paper says, and then I will discuss some critiques of their methods and conclusions. While people take their experiment as hard facts, the authors ask for contributions to the field with hard data to prove the model because they claim their model and experiments to be “soft” results.
The Paper
They started out with some experiments to help understand parameters of espresso. These experiments looked how flow is affected by pressure and temperature. They concluded:
- Flow is not constant, and it has a transient and steady state response.
- Flow isn’t proportional to pressure but related.
- Lower water temperature causes a higher flow rate.
They hypothesize that the filter bed is modified during extraction, and they more specifically focus on the hypothesis that fines migrate. They define fines as fragments of the coffee cell wall (typically less than 100um in diameter). They theorize that all or most of the fines migrate to the bottom of the coffee puck during the shot resulting in an effect on the water flow.
They then build a overturnable percolation setup from which they push cold water (4 C) through the puck one direction, rotated the chamber, and pushed flow through the other way. The cold water explains why there was such a high flow rate (we shall come back to that point).
They claim this is indirect confirmation of fines migration because the curve for forward flow and reverse flow have a similar shape. However, they did not have direct observation of fines migration nor did they do analysis on the puck afterwards.
The authors then move on to designing a model with the assumption that fines migrate by having a variable to describe the amount of migration. From this model, they make a computer simulation where they have fine particles and coarse particles with water flowing through as seen below.
This model, as designed, shows fines migrating.
The authors conclude by saying that they made a model to make soft experiments, and they hope others will contribute to making hard experiments to validate the model or improve it.
Critique
The top level critiques are:
- They used cold water (4 C) for experiments.
- Figure 7 is not interesting and doesn’t sufficiently isolate variables.
- Their simulation doesn’t account for cake compression.
Water Temperature
The authors used cold water in these experiments, but not just tap water cold. They used near freezing water at 4 C. They used this cold of water because the transient and steady state curve of flow was a similar shape as with using hotter water.
As a result, the water will do two things:
- Extract coffee very slowly.
- Not release CO2 trapped in the coffee.
Both of these variables are hugely important in understanding water flow through coffee because they both interfere with any movement of fines. CO2 inhibits flow as bubbles have to make their way through in the liquid.
Extracting of solubles allows the puck to change under the water pressure, and their experiment doesn’t clearly state if they use spent grounds or fresh coffee, so we don’t have better information on a key part of the experiment.
Figure 7 is Not Surprising
The authors look at this figure to say the two shapes are similar, but I don’t think that says anything about fines migrating. The transient flow is high until the puck is pressurized, and then the flow hits a steady state.
If you are pumping water into a puck at constant pressure, then to achieve that pressure, you need a high flow rate until the puck pressurizes. That’s a better and simpler explanation for what’s occuring.
Cake Modeling
Their simulation model shows fines migrating during water flow which is how the model was designed. However, the model assumes larger particles are immobile under high pressure and water flow. This is a large assumption that isn’t seen in experimental data. Instead, larger particles are pushed downwards and compressed thus removing easy channels for fines to migrate.
Concluding Remarks
The authors had some interesting research, and at that time, it was quite good considering technical limitations in studying espresso. However, so many people in coffee have believed fines migrate based on this single paper when even the authors didn’t believe their results strongly or directly proved the theory. Additionally, these critiques should be sufficient to re-evaluate this paper’s influence and basis.
This paper collected a large amount of data at the time, acknowledged by the authors. Espresso is a highly complex, multi-variable problem, and in the coffee community, specially the coffee data science community, we should question the underlying assumptions of data so we don’t science wrong.
I would love for someone to reproduce this experiment and their results using hot water or better still as well as find direct evidence that fines migrate significantly.
My data has shown that fines migrate but only very slightly, and I doubt this has such a disproportionate affect on the puck.
If you like, follow me on Twitter, YouTube, and Instagram where I post videos of espresso shots on different machines and espresso related stuff. You can also find me on LinkedIn. You can also follow me on Medium and Subscribe.
Further readings of mine:
- Petracco, M., and F. Suggi Liverani. “Espresso coffee brewing dynamics: development of mathematical and computational models.” Colloque Scientifique International sur le Café. Vol. 15. ASIC ASSOCIATION SCIENTIFIQUE INTERNATIONALE, 1993.
