Updated: Feb 14, 2018
Welcome to the second part of our mini-series on coffee microscopy! This time we look at two optical microscopes in the laboratory that both show coffee grinds in a different light.
Plus, did you know that microscopy can be used to estimate grind particle distribution? It’s quite easy with a little help from Fiji, just read on!
Introducing Olymus and Leica
Both the Olympus and Leica brands manufacture a whole range of exciting microscope technology, but today we picked out a transmission optical microscope by Olympus and a reflected light optical microscope by Leica to examine the two approaches.
OK, so what is the difference…?
Simply put, in a transmission microscope, the light comes from below the sample (bottom up), while in the case of the reflected light (or incident light) microscope, the sample is illuminated from the top down. The results look completely different, even though both samples are espresso grinds from a Mazzer Robur, zoomed in to the same level (the scale bars are very similar).
Similar coffee grinds look very different under a transmission (left) and a reflected light (right) optical microscope. The difference is due to the way the samples are illuminated.
The top down illumination in the Leica results in images similar to the DinoLite Captures. We can see the fine particles quite clearly, and a good depth of field means that particles on top and further below are all quite nicely in focus. Once again, just like with the DinoLite, we can observe that the larger particles are almost completely covered in a layer of fines.
In contrast, the transmission image from the Olympus microscope shows the silhouettes of the particles without much surface detail. The light shines from below the sample, and the coffee particles -being opaque- block the light from reaching our eyes.
That’s right, it works the same way as shadow puppets!
The hand blocks the light and the shadow is projected on the wall. In a similar way to our coffee grinds, lots of surface detail gets lost in the process- but it’s not necessarily a bad thing. We can suddenly observe the larger particles behind the layer of fines and see their shapes more clearly. The fines can still be seen as small dots sitting around the edges.
Shadow Puppets: a transmission image
Now Let’s Count!
If we could measure the sizes of all the particles in a representative image, we could calculate particle size distribution more accurately then using sieves and without the need for bulky, expensive instruments, such as laser diffraction particle sizers.
Counting and measuring particles one by one sounds tediously time-consuming! Now’s the time to turn to Fiji for help. This is exactly what the Fiji ImageJ software does: it automatically sizes and counts particles in an image. (It's free to download from here)
Sounds great, but let’s consider a few issues first…
How does the software know what is a particle and what is a blob of a few particles stuck together?
Unfortunately, it doesn’t. We have to separate the particles as much as possible before taking a microscopic image to get a realistic result. This can be done by blowing the clusters apart in a confined space using compressed air. Otherwise, we could try floating them on a liquid and let surface tension separate them. New ideas welcome!
How do we get a representative sample?
In other words, how do we make sure that what we measure for one image is true for the whole coffee sample? The best way to make sure the sample is representative is to take and measure more images, more samples and average them at the end. This means more work of course, but gives much more accurate results.
What about grey in-between colours, small disturbances in the background etc, can we get rid of those?
Luckily, yes. ImageJ has a fairly sophisticated way of separating the background from sample. After a process called thresholding, we are ready to count!
Fiji Image Processing: the background is "cleaned up" and the shape boundaries are identified (thesholding)
The particles are counted, measured and sorted by ImageJ, now we only have to convert pixels to micrometers (using the scalebar we referred to before in Part 1). Flat shadows also need to be converted to equivalent sphere diameters, so that we can display the results the usual way:
Particle distribution by optical microscopy
Here’s our particle size distribution, based on optical microscopy and Fiji ImageJ. Of course, it’s not 100% perfect, as some particles are still counted as one, rather than two or more. At the same time, the results are remarkably similar to the particle distribution of a Mazzer Robur grind measured by separation on a M100 electric sieve kit:
Particle distribution measured by separation on a sieve kit
The sieve kit measures particle distribution by weight, but, assuming uniform density it can be thought of as distribution by particle volume too. We will later return to more particle sizing issues in the future, keep checking our blog!
Next time we will look into the most exciting microscopy method so far: Scanning Electron Microscopy (SEM) will bring us to a real close-up with ground coffee particles up to 2000x magnification!
Would you like any of our methods applied to your products or instruments, to learn more about how your coffee and your equipment performs? We offer a completely personalized consulting service on all your coffee science issues.
Give us a call to get started with a free consultation session and learn what we could do for your coffee quality!