Under the blacklight

26 minute read

“I see black light.

Victor Hugo’s last words, and likely also mine

While I was taking the Q, Rob Stephen happened to walk through the lab at the beginning of our green grading module. Our instructor invited him to share his wisdom, so he offered: “Grade as if the farmer is watching.” What Rob meant was that we shouldn’t be putting every coffee under a microscope, so to speak—we should be compassionate and reasonable in our assessment of physical defects. We should give the benefit of the doubt, since the difference between 5 secondary defects and 6 might mean little to us, as buyers, but could mean two distinctly different financial realities for a producer. Even if microscopy appears to be my inclination, I do try to take Rob’s advice to heart.

After all, he literally wrote the book on green grading.

His approach is egalitarian, inclusive, and certainly a bit different than the mentality of high-quality focused specialty roasters—who seem to spend as much effort excluding coffees from their cupping table as they do inventing new flavor notes to put on their labels. So I’ve been surprised, impressed—and maybe a little worried—by how many of the coffee samples I’ve received from producers in the last 12-18 months have been free of any green grading defects.

It’s complicated.

Zero-defect coffee typically requires abundant labor (often inexpensive or possibly unpaid), or expensive investments in infrastructure and equipment such as density and color sorters as well as hullers with lower changeover loss and knowledgeable workers to keep the machines calibrated and efficient. The high cost of these investments often effectively consolidates milling power into the hands of those few companies or individuals who can afford it. It’s a plutocracy, borne from the inheritance of colonialism and imperialism.

(My friend and frequent co-conspirator Ben from Crop to Cup notes that there are increasing numbers of small community- or cooperatively-owned mills being built in coffee producing regions. This is true—but it isn’t happening everywhere, is still inaccessible to most smallholders, and typically requires the existence of government support agencies for low-interest financing and technical support)

At the end of that day—even if a producer is selectively picking, sorting, floating in cherry, calibrating and cleaning their pulper, floating in parchment, and picking defects off the drying tables, there will likely still be defects that are only revealed (or created) at the dry mill.

So then we need an army of machines or human beings to cull the defects once the hulls have been removed. In places where human labor is cheap (like Ethiopia), this is typically “women’s work” and conveyors carry kilo after kilo of coffee every minute past skilled hands that pluck defects with the precision of a surgeon (or Barack Obama going after a fly. Must have been all of that practice firing missiles from drones in Yemen). In other parts of the world, where minimum wages for agricultural work have increased somewhat (like El Salvador), green sorting tables sit empty while giant, metal machines vibrate, sort and reject.

A photo I took in Ethiopia in 2014 of defect sorting tables at a private exporter’s dry mill in Addis. Yes: I asked the women in this photograph for permission to take the photo.

Green grading isn’t a perfect assurance of uniformity or quality, even as the SCA and CQI prescribe standards disqualifying coffees from specialty markets based on their green grade. Surely we’ve all, at some point, graded a sample as zero defect only to discover on the cupping table a lack of uniformity between the cups or perhaps a defect.

Is there some other way to detect this?


Let there be light

Sight is entirely dependent on the reflection of light from surfaces. We don’t see objects; rather, we see reflected light bouncing off them into our eyes—and our brains do the heavy lifting to interpret and interpolate the rest (look how much larger the occipital lobe and visual cortex are in humans versus other mammals. We’ve got some serious processing power). Our eyes contain two different types of structures—rods and cones—patterned in our retina to offer both scotopic (low light) and photopic vision (higher light) at different resolutions and spatial acuities, depending how much, from where and what frequency of light strikes the two types of photoreceptors. Colors we perceive are the frequencies of light that reflect into our cones—the frequencies of light that weren’t absorbed by the object.

(Are there any coffee buyers who have tritanomaly or tritanopia? I’m very curious how coffee appears to you under UV light. Can you see differences? Comment below please!)

Surfaces or objects that absorb all light or where light doesn’t reach (black holes, the dark side of the moon, goth kids) appear black to us, signifying the absence of all light reflection. Stealth technology works much the same way, at least with line-of-sight radar—either an object flies below the radar horizon (no reflections) or is made from radar-absorbing material (fewer reflections) and has geometry configured to diffuse reflections:

The F-117 Nighthawk, an example of a stealth aircraft. Notice its faceted geometry, which is designed to scatter radar rather than reflect it back for detection by 1970’s era radar technology. It’s the goth kid of planes.

Thankfully, stealth technology has evolved beyond state-sponsored murder and has been adapted to gift the world the highly fashionable shade of ultrablack.

Even still: Humans can perceive only a portion of the electromagnetic spectrum—known aptly as the “visible light spectrum.”

We’re very good at naming things.

The rest—including infrared, on the low frequency side of the visible spectrum, and ultraviolet on the high (lower numbers on the following image = faster wavelengths and higher frequency)—is perceptible in some way to other organisms. Humans need special equipment to perceive it, except, occasionally, in the case of ultraviolet light. Substances called phosphors can make us aware of the presence of ultraviolet light through the phenomenon of fluorescence.

It’s the same principle by which those annoying, buzzing office lights (fluorescent bulbs) work. Electricity passes through a field of gas (like mercury), producing short-wave ultraviolet light (which would normally be invisible to us). This UV light then passes through a coating on the bulb which, essentially, absorbs some of the light’s energy, de-exciting the electrons, downshifting the wave from high-frequency (invisible to humans as ultraviolet light) before re-emitting it back to us at a lower frequency: visible light.

For our purposes, we’ll primarily refer to Ultraviolet A (UV-A, as it might appear on your bottle of Coppertone)—or wavelengths of 315 to 400ish nanometers. More harmful UV rays, those that cause cancer and eye damage—UV-B rays—live in the higher frequency range, from 280-315 nm and constitute about 5% of the UV rays that reach us from the sun (the ultraviolet light produced in fluorescent lights is superfast ultraviolet-C light with a wavelength of 254 nm). 

UV light is really cool. You can use it to detect scorpions before you get into your bed in Guatemala, or look at chlorophyll glowing red in a spinach smoothie, or see where your cute little dog relieved himself on the rug when he was mad at you (a result of jealousy, and the presence of phosphorous derived from phosphocreatine in the urine breaking down). 

By using a $10 UV flashlight (like this one), you can see a history and biography of your coffee that is otherwise invisible to the naked eye. 

It’s CSI, but for coffee! (I did forensics work for a three-lettered agency, albeit for a brief period, and would have killed for the zoom-enhance bullshit they do on that show).

To my knowledge, there are no LED flashlights that produce shortwave UV-B rays. Regardless, use caution when using UV lamps. Even with the flashlight I linked above, which produces UV-A light at 390-395nm, it’s recommended to use protective eyewear such as UV glasses (or even eyeglasses with a standard UV-protective coating on the lenses) and limit exposure times. Take frequent breaks (easy to do, it’s fairly tedious and boring work). And never, ever look directly into the flashlight.


UV analysis as a tool for coffee quality

As a rule, I do not use UV as a rejection tool. Ever.

Period, full stop.

We simply do not understand its application to coffee well enough for it to be reliable in that way, and we are simply too far down the value stream as buyers. In terms of signal detection for defects, the data we gather is so generalized and imprecise that we are equally at risk for false positives and false negatives. But this doesn’t make it entirely worthless. I‘ve observed trends and done research, and I’m happy to share it with you under the following conditions:

By continuing to read this post, you agree that you, too, will only use UV fluorescence data to inform your roast forecasting and planning; deliver feedback to importers, exporters and producers (because of course you already share your cupping notes—don’t you?); and help the industry develop this newly rediscovered and nascent technique by sharing your own findings as they emerge.

And under no circumstances will you use UV analysis as a rejection tool.

Deal? Ok. Let’s move on.


Much of the information in this post derives from my own data and research, but I owe a huge debt of gratitude to the effusively kind, talented and humble Tim Hill, who generously shared some of his ongoing research. Tim turned me on to this inquiry some 3 years ago, and I very much look forward to reading and re-reading his forthcoming handbook for UV fluorescence and green coffee. His work with Getu Beleke on Ethiopian coffee varieties, published by Counter Culture, is essential reading for any coffee buyer.

Like yeast inoculation to control coffee fermentation (seriously, click that link), UV in coffee isn’t new. The coffee board of Kenya published a manual in 1981 referencing UV light sorting, Daterra coffee in Brazil has been using UV sorting for cherries and green at scale for over a decade, and going back to June 1975, the East African Industrial Research Organization in Nairobi presented a paper at the 7th International Scientific Colloquium on Coffee in Hamburg showing that so-called “stinkers” could be identified under longwave UV light (a finding that, according to the EAIRO’s annual report from 1972, had been known by the staff for “some time”). 

On his Instagram, Tim posted a photo of Sortex UV sorting light box from Kenya from 1979. But like he says—everything that’s old is new again. Now, importers like Sucafina/32 Cup have entire sorting rooms dedicated to UV sorting and it’s standard practice for Caravela to perform UV analysis of samples it receives.

(In fact, Caravela published their own blog post about green coffee UV analysis, though my methodology and analysis differs slightly from theirs.)

When I receive a sample of green coffee, I isolate 100g and put it under blacklight (really, it’d be better and more representative to pull a larger sample but this shit takes forever and I have, like, a job to do). In testing various wavelengths of light, I’ve found that 360-390nm produces the brightest fluorescence of coffee and greatest contrast. This makes it easiest to work with for the uses in this post. However, even the flashlight I linked above at 390-395nm will work. According to Tim, 390nm and above doesn’t show contrast of certain types of fluorescent inconsistency well, and 360-365nm may be too much contrast. The sweet spot is in the range of 375-385nm, like that offered by this flashlight.

I then sort the coffee into three piles:

For each of these three categories, I count the individual number of seeds and weigh them to calculate the percentage of the sample that fluoresces in that manner (we’ll get to the reason why I both weigh and count them later).

If you were to analyze the same coffee over time, you’d discover that the amount of fluorescence changes, most often decreasing—this is documented in the literature (Coffee Growing, Processing, Sustainable Production Part III: 2.10.3.2: “UV sorting becomes inaccurate after storage because the fluorescence of the beans decreases drastically.”). This makes sense—in experiments Tim was a part of, the research pointed to the fact that in every case, fluorescence was bound to a lipid. As coffee sits in storage, phospholipids selectively decrease as neutral lipid hydrolysis increases the concentration of free fatty acids.

For this reason, the most useful data from UV analysis can be extracted from fresh coffee—the fresher the better. I typically will analyze mid-harvest type, offer, pre-ship, arrival, and mid-roast cycle samples to track trends, deliver process feedback, and observe how the coffees age. I expect to see reduced fluorescence over that time (starting with the preship), assuming the coffee didn’t suffer damage in milling, shipping or storage.

Our initial population count, though, can give us insights into that coffee.

Unlike in other industries (corn, almonds, peanuts), where UV fluorescence is correlated to the presence of Aflatoxin, in coffee, there is zero correlation, unless you’re using a solvent or chromatographic method designed specifically to show it—but you can’t just look at the green under UV lamp and make a determination. In the spirit of my own contrarian nature (and in response to a particularly needy client), I went as far as to take a “low specialty” coffee that was nearly 100% UV fluorescent—a low-grown, dry-processed coffee from Brazil dried in full sun on patios—and sent it to an independent ISO 17025 accredited and ISO 9001:2008 certified laboratory to test for the presence of molds and toxins. I tested for Aflatoxins B1, B2, G1, and G2 as well as Ochratoxin A—just for shits and giggles.

None were detected at a threshold of 1 part per billion:

Then I had them checked again, just to be sure my results were bulletproof (wink) but this time at a detection limit of 0.5 ppb. Still nothing!

(Sidenote: Fuck that huckster. I’ve linked to this before, but if that sort of thing matters to you, read this report from FAO.)

So UV fluorescence in coffee isn’t correlated with Aflatoxin and can’t be used to detect it. So what is it correlated with (other than, as stated earlier, something to do with lipids)?

Well—though the Coffee Exporter’s Guide says that phenolic coffee can be detected under UV light, that has not been my experience. I think the confusion comes from the fact that there are phenolic compounds in coffee that have emissivity of around 450-480nm under long wave UV-A light (they show as teal/blue) but that doesn’t mean they result in a phenolic cup defect.

Under that advice from the Exporter’s guide, I’d taken samples (in this case, from Brazil) from a single estate that presented with phenol and put kilos of it under black light, sorting out anything that had any sort of fluorescence. Even after this, the UV-sorted coffee still presented with occasional and unpredictable phenol. I have a friend who attempted to sort a Wush Wush from Colombia that presented similarly, but without success.

But there was coffee glowing: This is part of the challenge of using UV sorting as a rejection tool. The images are riddled with false negatives and false positives. To put things back in the metaphor of stealth and radar: a radar doesn’t show you what an object approaching you is; it simply shows you the relative size of its radar signature. It requires the operator to gather additional context (speed, path, and an understanding of how objects on radar appear) to make a judgment call about what is observed.

It’s like TDS—you only know how much you’ve dissolved, but not exactly what you’ve dissolved (or what it tastes like).

In fact, there’s a whole milieu—a shit ton, if you will—of compounds in coffee that reflect blue under UV light, depending on the pH, concentration or context: phenylalanine, tyrosine, proteins, protocatechuic acid, p-Coumaric acid, caffeic acid, ferulic acid, chlorogenic acids, nucleic acids, theobromine, quinic acid, caffeine, and scopoletin, to name a few. And there are other compounds in coffee that absorb or shield from UV (since UV can be damaging at a cellular level, even plants need sunscreen) which can make  it even more confounding. 

Not all coffee glows. So clearly there’s something different about the coffee that does.

What is it?


This post has taken me awhile to write. To be honest, I dreaded it. I get asked constantly by clients and on Instagram how or why I use UV, and I always deflect, saying I’d write about it eventually without ever having any real plans to follow through. So I put out a filler post while I worked on this one, and tugged at the threads until there was no more sweater.

And here we are.

In researching the application of UV fluorescence in coffee, I always felt like I was on a search for dark matter. I could see signs around the edges and detect its presence and significance—but couldn’t see the actual thing I searched for. I’d gathered enough data to build a system of fuzzy signal detection using fluorescence—but when I sat down to write, I hoped to provide a bit more substance.

I wanted to shed some light. Put on a black light and throw a party.

This inquiry has taken me all over space and time—from reports written by the Kenya National Coffee Board retained on record at the New York Public Library; to CENICAFE’s library; to doctoral theses from the UK in the 1980s and Kenya in the 90s; to illycaffè’s work since the ‘70s; and to Counter Culture, Tim Hill and a few Jonnie Walkers (imbibed by me) at a guest house in Santa Ana, El Salvador.

It frustrated the hell out of me to come to a realization part-way through my research: this shit was already known. It’s been in practice for decades, all over the world, and yet most coffee buyers regard it as myth or mysticism, like tarot, or worse, phrenology. But with every frustration, I’d find breadcrumbs and Reese’s pieces of truth leading me toward an answer.

For example, here’s a spectral graph from illy that I uncovered in a 1982 doctoral thesis by Omozoje Ohiokpehai at University of Surrey showing the UV reflectance spectra of high quality coffee, taken using a Perkin-Elmer UV-VIS spectrometer:

Illy has known this shit for decades.

Note the distinct lack of a spike at 480nm? That’s where we’d expect to see a spike if there were emissivity showing blue under UV. The science of this was established years before:

Gibson and Butty (1975) examined the fluorescence of beans under ultra- violet light excitation, and set out the different and complex types of fluorescence they had encountered, especially from defective beans of the ‘stinker’ type found in wet process arabica. Whilst the correlation is not always clear cut, the age of bean playing some role in degree of fluorescence, machines based on this principle have been manufactured by Gunson’s Sortex and are used commercially, particularly in large batteries, in a curing mill in Nairobi, Kenya. The use of UV excitation after electronic sorting along the lines already described has eliminated hand-picking.

R.J. Clarke, Green Coffee Processing (1985).

I hate reinventing the wheel, but so it goes.


We tried using it to detect phenol, we tried using it to detect potato, we tried to show Rio—and it didn’t work.

Some coffees glow green, some yellow, some blue, some purple, some white—and full blacks hardly glow at all (which is why green grading is still important).

And according to Tim, there’s no clear pattern in chromatographic results. 

But there are some things we do know, physically, based on the patterns that appear in the data. 

Fully fluorescent coffee, regardless of process, tends to be:

If you cut into fully fluorescent coffee, sometimes the fluorescence is all the way through; sometimes it’s just on the surface. We can interpret from this that the cause of fluorescence in coffee is external—it occurs from the outside in.

Two forces commonly at play in coffee, particularly in the realm of processing, are heat and moisture. We can produce UV fluorescent coffee if we wanted to, simply by manipulating these two variables. 

For example, if you were to take a lot of coffee from, say, El Salvador, and split it in two—post-washing—and send half of it to dry in a thin layer on raised beds (where it reaches a maximum temperature of 33°C) and half to dry on a tarp on the brick patio (with a maximum temperature of 46°C), you might find—as Tim and I both have—that the coffee dried at lower temperatures will not have fluorescence (or very little) and the coffee dried at higher temperatures will have a fuck ton of it. 

Sidenote: this is one consequence of the advice processing specialists give their clients (use airflow, not more heat to dry coffee—and never have the coffee exceed 35°). I virtually never encounter fully fluorescent coffee that was dried with less heat and more air—although typically, that type of attention to detail in drying systems also means the producers are meticulous about other parts of processing as well, such as tank and pulper hygiene and fermentation protocols.

Poorly dried coffee, too, will fluoresce at higher rates. Often the coffee is dried too quickly (e.g. with excessive heat) as a cause (which of course will kill the embryo and possibly the endosperm tissue), and sometimes in too humid conditions (the saunas I’ve mentioned previously), and microbial and enzymatic activity likely also play a role, as might lipid oxidation (again, going to our “something bound to a lipid” intel from Tim). Drying at high temps causes fused oil bodies and large droplets in the intercellular space and leaking of protoplasm as well as “crystallization” of the surface, which can make it so that the surface of the coffee is “impenetrable to moisture” (Cocoa and Coffee Fermentations p. 386), leading to instability in storage.

And if you took green coffee and sprayed it with water, walked away and came back a couple days later to put it under UV—it’d glow like a Christmas tree (just like some wet-hulled coffees). So there’s something there.

Heat, of course, is also introduced before drying, during fermentation. During dry fermentations, coffees can rise in temperature quickly—particularly if they’re done unshaded, in hot climates or if there are additives to accelerate the fermentation—and this is why it’s advisable to mix fermentations often, ferment in shade or indoors, use a probe thermometer to measure the temperature below the surface (where heat can build) and often, to ferment underwater to moderate temperatures (a practice that also helps with cup and drying uniformity, but—as frequently—I digress).

In super trendy fermentation styles like the so-called “anaerobic process” (please stop calling it this) some producers place whole cherries in closed containers like a plastic drum with an airlock or a GrainPro bag for hours or days. I’ve personally recorded temperatures as high as 45° in the maceration—which will definitely kill the embryo and likely also endosperm tissue—and the resulting coffee will often be 100% fully fluorescent (and to my palate taste like rotten cherry, but dear young coffee buyers of the world: you do you).

The same thing, of course, can happen if cherry isn’t processed promptly or the truck breaks down en route to the mill during the afternoon (shit happens). And, of course, I’ve talked about the galaxy of microbes active during a coffee fermentation, which change the chemistry and structure of coffee actively throughout their metabolism, producing new molecules and enzymes as they use others.

Whether because it’s “related to 1: beans with dead/ejected embryos, 2: beans intensely fermented, [or] 3: wet-hulled process,” (Roast Magazine, March-April 2020 p. 50) fully fluorescent coffee—for whatever reason it glows—appears to be less stable.

My data indicates that coffees with a higher percentage of fluorescence show fade faster and show more fade over time than coffees without. This effect tends to be more pronounced in washed coffees and coffees that have not undergone whole cherry fermentations, but I believe that’s more likely due to the masking effect of loud fruit flavors rather than an absence of flavors associated with fade.

So it’s unsurprising, based on the context above, that the evidence of fully fluorescent seeds suggests that a coffee will fade faster and will have greater issues with uniformity, particularly given what we know about their density and moisture. After all, we know that coffees with both lower density and low levels of moisture will have faster rates of degradation.

In general, if I analyze a 100g sample of coffee and find that it has fully fluorescent seeds, because I know that it is likely to fade faster, I’ll schedule it to be roasted sooner than other coffees that don’t show fluorescence—particularly if the glowing lot also is high moisture and/or low density. This way I can present the coffee when it’s tasting its best, and put others that might have a little better shelf life on hold. If we’re in the middle of the harvest, I might give feedback to the producer and, if the relationship is established in this way, might ask what temperature the coffee reaches during drying based on an IR thermometer reading, or what temperature it reaches in the fermentation tank based on probe temperatures (and pay a goddamn fair price).

I’m comfortable having coffees that don’t glow and have their moisture targets in line accumulate a bit of carry over at Conti since they retain their value over time, versus a coffee that does glow and will fall off 3 points before I originally hoped to roast it. Think of it this way: If you buy two gallons of milk, one with a three month shelf life and one that expires in 2 weeks, which are you going to use first? (Not to go too far off-topic in this tome, but suffice to say: milk supply chains should be a post for another time)

Excluding outliers (such as whole cherry macerations I described above, as well as so-called “carbonic macerations,” which is another ridiculous name), the average number of fully fluorescent seeds in my data from 100g samples is 1 (but the mode is 0). This corresponds to an average of 0.16% full fluorescence in aggregate.

Which is good: because we know how to create full fluorescence, we know it’s preventable. And for the most part, the producers I work with have been successful at eliminating it.

But wait—there’s more.

Tim color sorted a batch of roasted coffee that had a high level of fully fluorescent coffee—the difference between the lightest third of the batch and the darkest third was 16 points. That’s huge: roughly the difference between dropping a coffee just before the end of first crack (let’s call that a 72 Agtron) and letting it hit second crack (~56). In separating out coffee that glows versus coffee that doesn’t, if roasted using the exact same profile, those two batches can vary by as much as 10 or 20 Agtron points.

A variance in development that large is highly problematic. If you’re roasting naturals, in particular, working with a dry mill that uses UV color sorting could be an advantage for helping to ensure uniformity in the lot.

But this is all about fully fluorescent seeds. What about the speckled ones?

In green grading, we describe different types of defects and divide them into two categories: primary and secondary. Secondary defects are less impactful on the cup than primary defects—of which a single defect such as a full sour or full black can ruin a cup. And unlike primary defects, secondary defects are not 1:1 — meaning one defect per seed affected. Rather, it takes 5 chipped or broken pieces to yield a single secondary defect, or 2-5 broca affected seeds for a single insect damage defect, or 2-3 partial blacks for a single defect, and so on. And while the industry allows zero primary defects in specialty coffee, we recognize coffees with up to 5 full secondary defects as specialty.

Speckled coffee is kind of like that, too.

Most often, speckled coffee is caused by some sort of physical trauma to the seed—typically pulper damage, and occasionally dry mill damage, as well as insect damage. Though you might see some signs of subtle bruising with the naked eye, it’s often otherwise undetectable. And like secondary defects, there is some difference in the cup, though it’s subtle and depends on how much there is. If you were to separate out speckled coffees, roast and cup them alongside their non-glowing counterparts, you’d likely find that speckled coffees are slightly more fruited and have a less pointed acidity, both indicative of some sort of change to the acid profile that could be attributed to microbial activity.

But just because it doesn’t impact the cup as much doesn’t mean you shouldn’t care about speckled coffees: the percentage of speckled seeds is an incredibly valuable piece of data to share with producers.

By giving the percentage of speckled coffees to producers (particularly for coffees with a high amount of speckled fluorescence), it can help them to address outturns related to pulper damage through calibration. This can dramatically improve yields and overall profitability at the wet mill level. This is because, as Tim says, there is a correlation between a high level of pulper damage showing under fluorescence and full removal of parchment by the pulper.

What turns coffee into a disco?

In the literature, I found various explanations for fluorescence. Gibson and Butty, in their seminal study on the use of fluorescence in green coffee sorting, which was presented in 1975 at the 7th international Colloquium on the Chemistry of Coffee, wrote “This phenomenon has been inferred to be due to intra- and inter-cellular physical and chemical changes in the coffee bean tissue. Since physical cellular damage is an observed feature of over-fermented coffee.”

And I don’t know if this extends to coffee seeds (waiting on Sam Knowlton’s fountain of wisdom here) but I found a couple of studies demonstrating that in plants under water stress and in presence of heat, the leaves of the plant glow increasingly blue under ultraviolet light as UV-absorbing components called flavonols build up in the stressed leaves, absorbing red bands of light and increasing blue fluorescence. Water stress and presence of heat and light sort of sounds like coffee processing and drying. 

If I were to hazard a guess (lol remember how I never took organic chemistry?) it would be phenolics such as ferulic acid (which is derived from the much-less-fluorescent amino acid phenylalanine and biosynthesized from the much-less-fluorescent caffeic acid) as well as, importantly, scopoletin, increasing blue fluorescence in coffee:

The presence of a natural fluorescent constituent of coffee, scopoletin [emphasis mine], LVI, hinders a rapid screening of green coffee beans for the presence of aflatoxin B1, and some of the early reports of toxin contamination may have been partly due to the presence of interferents.

R.J. Clarke, Coffee: Physiology (Page 23)

Scopoletin, which is present inherently in green coffee, can also be produced by microbial infection. In other plants, such as cassava, tissue dehydration and oxidative stress after harvesting lead to the accumulation of scopoletin, especially at the sites of mechanical damage—which kind of sounds like coffee that has been damaged by a pulper before drying, or drying in high heat (or tumbling in a guardiola, for that matter).

Drying appears to be implicated in the generation of fluorescent compounds in other ways as well, possibly due to its effect of converting protochlorophylls to 440-460nm reflectant fluorescent chlorophylls:

One assumes that the pigments extracted from the silverskin have been synthesised within the coffee cherry in the absence of light. Accordingly they may resemble the protochlorophylls that Jones (1965, 1966) found in the seed coat of Cucurbita pepo. However, Boardmann (1966) reported that a protochlorophyll can be converted to chlorophyll by light and heat. The rate and extent of conversion is temperature dependent, reaching a maximum at 40°C. Such conversions may occur during sun drying or mechanical drying of coffee, yielding a more complex mixture of pigments.

Omozoje Ohiokpehai, “Chlorogenic acid content of green coffee beans” (1982)

And allagochromes, which are glow under ultraviolet light after oxidation, are formed between proteins and and quinones, such as chlorogenoquinone (the quinone derived from chlorogenic acid) during the enzymatic oxidation of phenols—such as during the harvesting and decay of plant material. At green coffee’s pH of around 5.5-5.75, any allagochromes formed in green coffee would appear blue.

In other words: fluorescence, however it’s summed up, whatever it chemically indicates, whatever its cause—appears to be indicative of decay and deterioration: of life, of quality.

But like I said: illycaffè already knew this shit, which is why they’ve used it in Brazil for decades, where much of the coffee is dry processed.

I looked in the book Andrea Illy co-edited (he’s the grandson of illy’s founder, Francesco Illy, and the son of Ernesto, who led illy’s expansion and contributed massive amounts of research to the coffee industry), Espresso Coffee, and sure enough—I found context, and what is likely illy’s initial attempted use case:

It has been found that certain non-visible defects, such as moulds or bacteria, fluoresce when irradiated with long-wave ultraviolet light (360nm). This property may be used as a basis for sorting. This technique was originally developed for removing ‘stinkers’ from green arabica coffee beans, but has found applications in sorting other material as well. However, the fluorescence effects can be best sorted on freshly harvested coffee. In the case of light beans after a long storage period following the harvest, this technique faces more difficulties.

Espresso Coffee (p. 107-108)

And:

The surface discoloration of whitish beans is due to fermentation by Streptococcus bacteria that in the most severe cases may also reach the cells under the epidermis (Dentan, 1991). The attack can occur if storage is too long or in conditions of excess humidity [remember my post about fade and microbes?]. White beans are not very aromatic and give slightly bitter and woody cups. Under ultraviolet radiation they show a blue fluorescence probably attributable to chlorogenic acids and caffeic acid, which makes them difficult to sort out from stinkers because of the fluorescent backgrounds they produce [emphasis mine].

Espresso Coffee (p. 132-133)

And:

Under examination using ultraviolet light, stinker beans show a white or white-blue fluorescence [again, mine], probably associated with larger quantities of free caffeic acid compared with healthy beans. Nevertheless, other defects and old crop beans are also fluorescent, making it difficult to sort stinker beans solely on this method.

Espresso Coffee (p. 130)

(Remember that caffeic acid is a precursor to highly fluorescent ferulic acid?)

Curiously, though,

Stinker beans can contaminate a large batch, even when present at very low levels, turning health beans into stinkers at a lower level of intensity (Gibson and Butty, 1975). Beans that have become stinkers only by contact are not fluorescent.

Espresso Coffee (p. 130)

There’s that Gibson and Butty study again.

This is yet another reason why you cannot make purchasing decisions based on UV alone, but UV indicating likely presence of stinkers might make you more cautious or want to cup a sequence of five cups instead of just three.

And Illy confirms that for Rioy coffees “UV-VIS reflectance spectrum [is] identical to that of a normal bean.” (p. 132).

The point is: when coffee glows, it’s showing us something. It’s not always clear what, but by building out a more thorough history of the coffee, how it was handled, processed, dried, stored, milled, shipped and stored, you can infer details that can not only lead to year-over-year improvements in cup quality, but improvements in profitability for its producer, and the best possible experience of that coffee by your customers.


100g of green coffee under ultraviolet light

Try it at home

UV fluorescence is a very simple phenomenon to explore on your own, particularly if you have an IKAWA or other sample roaster capable of roasting very small batches. Simply place a coffee under UV-A light, separate out anything that glows, roast and cup the reference lot separately from your glowers and your non-glowers, and evaluate your findings.

For demonstration purposes of this phenomenon, I took two coffees from my warehouse and conducted my standard UV analysis, separating them into three piles. Neither featured many full fluorescents, so I kept sorting and combined fully fluorescent and speckled coffees until I had enough coffee to conduct moisture content analysis using my moisture meter (about 80g total fluorescent coffee). I maintained a reference sample from each (unsorted, straight from the bag), leaving me with three total preparations from each sample:

  1. Reference
  2. Non-glowing
  3. Any fluorescence

The first lot contained 9.6% total fluorescence now, four months after landing, a decrease from 13.9% as a preship sample. The moisture content of the fluorescent coffee was 11.1%, versus 11.2% in the non-glowing separation and reference (a pretty narrow spread, really—it tends to be greater than this). The glowing coffee had a bulk fill density of 0.68 g/mL, significantly below the non-glowing, at 0.72 g/mL, and reference at 0.71 g/mL.

The second showed a total of 22.4% fluorescence, nearly 8 months after landing, but unfortunately I don’t have data from a preship of this coffee to compare against (it’s the rare lot I purchased spot). However, the UV glowers had a moisture of 10.2% and density of 0.70 g/mL versus the non-glowing coffee at 11.0% moisture and 0.73 g/mL. The reference sat at 10.7% and 0.72 g/mL.

I roasted each of the three preparations of both coffees using an identical profile in an IKAWA sample roaster and presented them in a cupping flight to experienced cuppers blind but in the following order:

Reference
Glowing
Non-Glowing

The older lot was poured first.

In the older lot, the Reference cup initially presented greater complexity than the non-glowing cup, but as the coffees cooled, the wheels fell off: the non-glowing cup remained consistently vibrant, juicy and sweet, while the initial character of the Reference lot fell off into a papery, dry finish with little structure. The second cup of the set—the cup of Glowing coffee—presented papery, thin, and dry throughout, with low aromatic intensity.

In the lot that landed more recently, the differences were immediate and stark: the UV glowers presented with muted fragrance and aroma and the cup tasted vegetal, thin, and astringent. The Reference cup, while sweet and clean, lacked the winey acid profile, silky body and structure of the non-glowing cup. This contrast grew more pronounced as the cups cooled.

In other words: the coffees that contained fluorescent coffees also contained flavors we associated with past crop and fade, while removal of fluorescent seeds improved the structure, sweetness, acidity, and finish of the cup.

Based on this, I can imagine that it’s easy to feel the temptation, as a buyer, to ignore the sacred pact we’d made earlier and adopt a rejection mentality.

But What Would Rob Stephen Do?


Like I said earlier: zero-defect coffee requires either abundant, inexpensive labor from human beings or expensive machinery. This is true of UV sorting as well—even more so. Hand-sorting using UV is slower and even more of a nightmare than typical defect sorting and presents the additional challenge of eye safety.

I can imagine a buyer acknowledging this and buying a color sorting machine or a UV sorting machine for their roasteries. There are cases of well-known roasters doing this to increase the uniformity and consistency of their coffees.

And to be fair—you’re still following the rule we agreed to. You’re buying the coffee, you say, you’re just cleaning it up when you get it so that it’s usable by your standards. And you’re buying more coffee from the producer—that’s a good thing too! But you’re sorting out coffee that you’ve already paid to have someone bag, export, ship, inspect, import, store, finance and truck, and whose contract price was likely established based on a lower quality evaluation than the one you give it post-sorting. 

Think of implementing UV sorting at the roastery as increasing your roast loss percentage from 12% to 20%—but with more carbon emissions and more money going to a bank. You’re literally doing less to mitigate climate change while claiming that the invisible hand of conscious capitalism is helping the people who climate change will most directly impact. (“Quality above all else” is the trickle-down-economic theory of specialty coffee.)

And, meanwhile, the fluorescence you’re sorting is entirely preventable.

Paying $30,000 or more for a UV sorting machine (or color sorter) for your roastery is a less efficient utilization of resources than spending that $30,000 on higher premiums for producers with whom you work toward investing in their own operations—whether through a training program such as CQI’s Processing certification, or through working with private consultants, or just simply not needing to borrow at usurious interest rates between harvest cycles. Every dollar has more value the further you go up the value stream.

By returning data about what you see under black light to producers and exporters you can meaningfully deliver value to producers from whom you buy, and whom you rely on. If a producer reduces drying temperatures and dries to 10.5%, for example, or calibrates their pulper and reduces outturns—both of which would reduce fluorescence—the shelf life will be longer, the cup will be cleaner and more uniform, and there will be more coffee for them to sell. And likely, in a specialty context, they will get paid a better price (you’d pay a better price, wouldn’t you?)

As an exporter (or specialty importer), implementing UV sorting at a dry mill might help a coffee retain value in the marketplace overtime by reducing fade; reduce rejections due to defect or uniformity issues (which would, critically, reduce strain on the middle of the supply chain—rejections are a huge problem); and help give producers a doorway into the specialty market, as their coffees might cup higher post-sorting at a dry mill and thus have significantly greater value than previously determined. 

All of this is likely true—but it still makes more sense to do in Cauca, and not in Cleveland. (Alas—it’s March, and we could probably use the extra ultraviolet light)

I need to acknowledge the fact, plainly, that sometimes glowing lots perform better in the cup. It’s true: especially if you’re into trendy so-called carbonic macerations (I’m not) or so-called anerobic lots (nope!), or wet-hulled Sumatrans (not my style, but! People love them).

Especially if these coffees are fresh.

In the words of the great Joe Marrocco: “Taste is an aesthetic choice, not a moral one.” (sorry if you don’t remember saying that, Joe—I’ve been repeating it for years because it’s fucking great).

And the phenomenon of fluorescence relating to moisture happens to intersect with the phenomenon of fresh, high-moisture coffee tasting delicious in just the right way such that a high-moisture coffee that is shipped quickly might look like shit under black light but taste fucking great (I’d still roast it soon since it’ll likely fade quickly). If we take in mind that nicked and bruised coffees that glow speckled have a little more fruit and softer acid in the cup, then it makes sense that having a high percentage of “damaged” speckled seeds could round out or flatten the vegetal character of hybrids or Catimors, and could increase the perceived aromatics or fruitiness of other lots. So why would you reject it just based on how it looks under UV, if you can account for everything else?

Looking at my own data, my favorite coffees that I’ve roasted from the last 2-3 years have been anywhere from 1.1-14.6% total fluorescence. Less is not necessarily more.

The point is this: see what there is to see, but also taste what there is to taste. After all, even though they say “You eat with your eyes,” in my super humble opinion, this isn’t the best way to enjoy a cup of coffee.

Tags:

That's just, like, your opinion, man

    1. It’s not perfect – but more often than not in my experience pulling out anything that fluoresces fully will remove most of the quakers. Give it a try and let me know what you discover!

Leave a Reply to christopher Cancel reply

Your email address will not be published. Required fields are marked *

Back to top