Dr. Monika Fekete explores the evolution of a coffee bean under an electron microscope and uncovers how roast levels affect grinding.
Watching green beans turn into aromatic roasted coffee is something of a magical experience.
Have you ever stood there wishing you could see what’s going on inside those beans as fundamental chemical and physical changes take place?
This is the journey I would like to take you on. To look closer and systematically track how each bean is ground and extracted, our friends at Vacation Coffee Roasters roasted some Colombian washed process beans at six different stages of development for an experiment at Bureaux Collective.
The first sample was green beans, followed by a roast stopped pre-first crack on an Ikawa sample roaster. The rest of the samples were roasted on a Probat Probatone 12 roaster in four-kilogram batches. Each followed the same roast curve, and dropped at times corresponding to filter, light espresso (espresso 1), dark espresso (espresso 2), and past second crack. For roast details, see Table 1.
To reveal what happens inside the bean’s cell structure as the roast develops, I took a single bean from each roast level and cut a cross-section under the electron microscope at La Trobe University’s BioImaging Platform (see Figure 1).
Green beans have a compact cell structure, with the vacuole (the bag of nutrients inside a plant cell, see BeanScene August 2018) still intact. The cells begin to puff up and water starts to evaporate even before first crack, which is why you can see hollow cells in the second image. Non-enzymatic browning, such as caramelisation and Maillard reactions, are also underway at this stage. Still, the beans are very hard to grind and the coffee doesn’t become drinkable until after first crack. At this point, steam, CO2 and volatiles inside the cell build up enough pressure to cause the cell walls to press against each other and eventually rupture.
Looking closely at Figure 1, we can observe small fissures in the cell walls after first crack that were not present before. These are essential for opening up the cells so that the water can more easily reach inside and dissolve the plant nutrients, which, by now, are fast turning into characteristic flavour compounds.
The cracks continue to become more prominent as the roast develops. The last micrograph in Image 1 shows that cell walls are strongly fractured past second crack. Second crack also opens up tiny channels towards the surface of the bean that allow oils to migrate to the surface with the help of capillary forces.
How do structural changes relate to grinding and extraction performance?
So far, I’m not aware of any scientific reports systematically exploring the relationship between roast development, grinding, and extraction variables.
Let’s look at grinding first. As the roast progresses further, beans keep losing weight and become more dehydrated. The cell walls are increasingly weakened by fractures, as we saw in the microscope images. It seems reasonable that these would lead to darker roasts becoming more brittle and possibly fracturing more easily in the grinder. To test this idea, I ground all five samples past first crack on a Mahlkonig EK43 grinder, keeping the grind setting constant at 1.9. Then I measured their grind particle size distribution (PSD) in triplicate runs on a MasterSizer 2000 laser particle sizer at La Trobe University.
The results, plotted as the percentage each size range contributes to the total sample volume, are shown in Figure 2. The trend is very clear and in line with our expectations: darker roasts grind finer. The median particle size shifts from 305 micrometres just past first crack to 207 micrometres for filter, 160/153 micrometres for the two espresso samples and 120 micrometres for the sample past second crack. We can expect that this very significant shift in PSD will affect extraction variables too.
The insert shows the median particle size of all roasts past first crack. Data was recorded on a Mastersizer 2000 at Department of Environmental Geoscience, La Trobe University.
The lighter espresso roast was dialled in to yield a 42 ± 0.5-gram beverage from a 21.0-gram dose in around 28 seconds on a La Marzocco Linea PB. Ten espresso samples of each roast level (past first crack) were prepared in a fully randomised order. The grind setting (1.9 on the EK43 grinder) and the brew ratio were kept constant, allowing the shot times to fluctuate. Shot times of each roast level group were then compared by statistical analysis.
The results, shown in Figure 3, once again follow a very clear trend: darker roasts extract significantly more slowly than lighter ones under the above described conditions. Shot times ranged from 11 seconds for the roast just past first crack to 57 seconds for the roast past second crack. This is in agreement with the observed prominent shift in PSD as the finer grinds take longer to yield the same target beverage weight. Interestingly, total dissolved solids (TDS) levels only showed a significant increase between the first two samples, just past first crack and the filter roast (see Figure 3, insert).
Past this stage, no statistically significant difference in TDS percentage could be measured between the four samples. A possible explanation is that the quantity and availability of soluble components is still low until the roast has developed to a level at least equal to a typical filter roast. Past this, even though the exact make-up of the flavour compounds could differ, their total amount remains roughly equivalent as long as the same brew ratio is maintained.
In practice, these results suggest that darker roasts, rather than being more soluble, just break more easily. The observed behaviour in extraction and grinding all relate back to the structural changes the beans go through roasting – something to keep in mind next time you carefully watch your beans develop.
I would like to thank Vacation Coffee Roasters, Peter Lock at the LaTrobe BioImaging Platform, Melissa Reidy from the Department of Environmental Geoscience, and Dr David Hoxley, lecturer in physics at La Trobe University, for their kind assistance with this research.
This article appears in the October 2019 edition of BeanScene. Subscribe HERE.