The effect of decaffeination on coffee’s roasting and grinding performance
Shortly after a lecture on the physics of filter coffee in Milan, I bumped into Alex Wallace (QC at Caravan Coffee Roasters) and, together with the advice and support of Mat North (QC at Raw Material Coffee), we fell into the rabbit hole of decaffeination and its mystical effects on extraction.
Before you jump in, I’ll introduce myself. So, hey! I’m Mark Al-Shemmeri. I’m a chemical engineer by training (MEng) and I worked as a barista during my studies, slinging delicious Hasbean coffee from a Slayer (yeh, I know, a pretty awesome intro to coffee, right?). After a few years brewing beer in craft breweries, I started doing an Engineering Doctorate (EngD) at the University of Birmingham (UK) focusing on a data-driven approach to process and product simulations of coffee roasting. The thing that made that project unique was that for four years, I was somewhere between academia and industry, with the EngD part-funded by JDE Peet’s. This means that I had the resources, the time and more importantly, the support to publish my work with open access! Now, whilst also roasting under the pseudonym ‘dialect.coffee’ (using other people’s roasters), I’ve started a new role as a R&D Technology Specialist at JDE Peet’s in the Netherlands, where I have the continued support to share my work whilst I’m busy writing my thesis (it’s almost there…). As a disclaimer, I’m not a chemist (like Anja Rahn, or Samo Smrke), I’m an engineer, and this is reflected in my approach to the study.
So anecdotally, decaffeinated coffees ‘grind finer’, and there’s a few reasons this could be true. Decaffeinated coffees might be physically and/or chemically different and might therefore require more development during roasting to counteract these physical/chemical differences. But let’s also consider how roasters approach decafs. Roasters often schedule decaf coffees as the first batch of the day and more often than not, they use lower batch sizes and start roasting before their machines are properly pre-heated. There’s numerous reasons why decaf coffees might ‘grind finer’ but I’m not a huge fan of anecdotes. Neither was Alex, whose experience was in fact the opposite.
In a café setting, Alex was finding that a lower grinder setting (finer particle size) was needed for decaf coffees to match the brew parameters of non-decaffeinated coffees. Needless to say, this fascinated me. So I figured an explorative study on decaffeination would be the spade we needed to try and dig us out of the rabbit hole and demystify the effects of decaffeination on roasting and grinding performance.
To eliminate many of the factors that will prevent this from being yet another anecdote, Alex sent me 10 kg of green coffee - a regional blend from producer groups in Risaralda, Colombia supplied by Raw Material. The beauty of this was that Raw Material don’t just supply Caravan with the washed processed coffee, they also supply this same coffee but decaffeinated. The team at Descafecol in Colombia do phenomenal things to extract the caffeine and generate a coffee that, when dialled in, is delicious. So I connected with Mat to get his insights and words of wisdom along the way.
Enough of the introduction. Here’s the deal…
The study comprised a Colombian Arabica coffee decaffeinated with EthylAcetate (EA), analysed alongside its non-decaffeinated counterpart (i.e., the reference coffee). The two coffees were roasted via a range of time-temperature profiles, and coffees were ground using different grinder settings. For all green, roasted whole bean and roast & ground coffees, physicochemical analyses were performed to characterise the coffee’s transformation during roasting and grinding. Another disclaimer. Whilst I did taste the coffees with the legends at Caravan, studies led by Morten & Ida (of CoffeeMind) suggest that we just aren’t significant enough :( so I won’t include those here. Maybe next time?
I wanted to include enough detail for replication of the study but if you want to skip the finer details (too soon for a pun?), jump to the so what?
The following is sectioned into the aims, methodology, results, conclusions and outlook.
Aims
The study intended to:
(1) characterise the physical properties of both green coffees to identify the impact of decaffeination on the raw material
(2) roast both green coffees to different time-temperature profiles to explore the viable flavour space
(3) characterise the roasted coffee’s whole bean properties to clarify the decaffeination process’s impact on roasting performance
(4) grind all roasted coffees at different grinder settings to analyse the grinding performance
(5) characterise all ground coffee samples to assess the impact of both decaffeination and roasting on grinding performance.
Methodology
Green Coffee
Two coffees from the same producer groups in Colombia were sourced by Raw Material Coffee and supplied by Caravan Coffee Roasters. Both coffees were: (i) from the Risaralda region, (ii) a blend of Castillo, Colombia and Caturra varieties with a screen size of 18+ and (iii) subject to washed post-harvest processing. The coffee was then split into two lots, wherein one underwent decaffeination with sugarcane derived Ethyl Acetate (EA), whilst the second was not (i.e., underwent no further processing). A visual of the two green coffees is presented here for reference.
Decaffeination with sugarcane derived EA was performed by Descafecol, Colombia, whereby the green coffee is first steamed to improve mass transfer rates, and subsequently steeped in an EA solution to extract caffeine. After steeping, the coffee is again steamed to remove any residual EA solution and then dried until the original moisture content is achieved. RM provide a beautiful poster detailing the decaffeination process here.
Roasting Conditions
500 g batches were roasted in a pilot-scale spouted bed roaster (RFB-S, Neuhaus Neotec) with moderate airflow (48 Hz) under different time-temperature profiles.
The experimental design consisted of a rotated 2-factor (time, temperature) design, where each factor had three levels (-1 (low); 0 (centre); +1 (high)). Factor 1 (Temperature) was the constant inlet air temperature (245, 250 and 255°C). Factor 2 (Time) was the roasting time (250, 275 and 300 s). The roasting conditions (both coded and natural values) according to the experimental design are given in given below. Both the decaffeinated and reference coffees were roasted via these 5 different roasting conditions, yielding 10 roasted coffee samples.
Grinding
Roasted whole beans were ground with a flat burr grinder (EK43, Mahlkönig) at 3 different grinder settings (0, 4, 8) corresponding to particle sizes appropriate for espresso and filter (in the range 200–800 μm). Particle size analyses were performed via laser diffraction (HELOS, Sympatec) for each coffee to create a grinder calibration protocol, detailing the median particle size (x₅₀) at corresponding grinder settings.
Coffee Characterisation
Green coffee analyses included mass, density (bulk, packing & intrinsic), moisture content via recirculating oven, electrical conductivity, colour, water activity, size (principal dimensions, volume and surface area) and porosity (via MicroCT). Roasted whole bean coffee analyses included mass, bulk density, moisture content and colour (whole). Roast & ground coffee analyses included colour and particle size (via laser diffraction).
Results
Green Coffee Analysis
Measured green coffee properties are provided here for reference. Although there were no significant differences in size (principal dimensions, surface area, volume and sphericity), density or porosity, there were differences in moisture, water activity, colour and electrical conductivity. Differences in water activity are attributed to distinct moisture contents (Pittia, et al. 2007), despite efforts to dry the decaffeinated coffee to its original moisture content.
A significant difference in green coffee colour was apparent — similar effects were seen in Ramalakshmi & Raghavan (1999) and Klikarová, et al. (2022) for various decaffeination methods. The impact of EA decaffeination on the green coffee’s measured colour shows that the colour distribution’s span is typical for both coffees, but distinct.
No differences between the coffee’s micro-porous network could be resolved, suggesting that microscale diffusion is not affected by decaffeination and that the steaming, soaking and drying processes do not impact the bean matrix on the micrometer scale.
Time-Temperature Profiles
Time-temperature profiles for both coffees and all roasting conditions reveal some interesting phenomena.
For the first 30-60s, the temperature response is attributed to the thermometric lag of the thermocouple, and is of no interest here. At around 60s, the temperature response for all reference roasts begin to decrease. This could be due to chaff removal (the decaffeinated coffee is polished and waxed, so chaff removal does not occur during roasting for the decaffeinated coffee) and the need to remove a greater volume of moisture (ca. 1% more).
Temperature responses for roasts of decaffeinated coffee were higher or equal to roasts of the reference coffee under the same process conditions. Prior to first crack, this phenomenon is attributed to the reference’s greater green coffee moisture, where more energy was initially required for water removal. Around first crack, temperature responses for all roasts of the reference coffee begin to decelerate (i.e., rate of change of temperature decreases). For high inlet air temperature conditions, the temperature response then increases rapidly, but this effect of lowering and increasing the rate of change of temperature is less dominant when roasting with an inlet air temperature of <250°C, likely because the required activation energy was not attained.
The time and temperature at which first crack occurs (data here) depended on process conditions, wherein higher constant inlet air temperatures yield greater heat transfer rates, decreasing time to first crack and increasing the corresponding temperature at time of first crack. Small differences in the recorded time and temperature of first crack are due to the subjectivity of first crack’s measurement, here taken as the time and temperature with the mode first crack audio. Interestingly, decaffeinated coffees ‘cracked’ at higher temperatures than the reference under all roasting conditions. Although end of roast temperature was lower for the reference coffees than their decaffeinated counterparts, end of roast temperatures were largely dependent on roasting conditions, whereby greater end of roast temperatures were attained due to longer roasting times and higher air temperatures. Due to the dominant effect of roasting conditions, there was no statistical impact of decaffeination on end of roast temperature.
Roasted Coffee Analysis
Coffee’s mass loss during roasting depends on roasting conditions and decaffeination. Greater air temperatures and longer roasting times both increased mass loss, although decaffeination had no statistically significant impact on mass loss due to the dominant impact of roasting conditions. For each process condition, the decaffeinated coffee had lower mass loss than the reference coffee. This was also observed for dry matter mass, so differences are not directly attributed to green coffee moisture.
Bulk density decreased as a function of both longer roasting times and greater air temperatures. Differences incurred by decaffeination are difficult to discern due to large standard deviations associated with the measurement method. This phenomenon is due to distinct temperature responses, where at higher temperatures the reference coffee’s thermal response decelerates before rapidly accelerating through first crack as exothermic reactions dominate. These reactions produce a significant thermal load within the bean that quickly transforms coffee’s physical properties.
The coffees’ roasted moisture content was not significantly influenced by applied air temperature, or by decaffeination, even though the initial moisture content differed between the two coffees. Roast time, however, had a significant impact on moisture. Data suggest that the final moisture content is invariant to thermal loading rates in the tested range.
The influence of decaffeination on green bean colour is documented in Ramalakshmi & Raghavan (1999) and Klikarová, et al. (2022), although physicochemical pathways for colour development during roasting are not. Data here has shown that whole bean colour was significantly impacted by coffee’s decaffeination, where decaffeinated coffees were darker than the reference for all roasting conditions.
Batch Homogeneity
Consideration of batch homogeneity is critical for whole bean products, particularly as changes in pigment and hue induced by the decaffeination process might mislead consumers and their affective perception of the coffee. Colour distributions were used here to characterise whole bean uniformity for the decaffeinated and reference coffees roasted under different time-temperature conditions.
Data show that decaffeinated coffees are darker (have a higher colour value) and more uniform (have a smaller span) than the reference coffees for all roasting conditions. Although decaffeinated coffee’s uniformity is greater than the reference, the two coffees are visually distinct.
Grinding & Particle Size Analysis
Particle size distributions were determined via laser diffraction and performed for all roasted coffee samples and three grinder settings.
Grinder setting (i.e., burr gap) had the most significant impact on median particle size (x₅₀), increasing x₅₀ approximately 50 μm per integer grinder setting. For specified roasting conditions and grinder setting, the x₅₀ of decaffeinated coffees were lower than the reference, with the difference ranging from 8–38 μm.
For a specified coffee and grinder setting, roasting conditions modulated x₅₀ by 18–38 μm. Similarly, grinder setting had the most significant impact on the percentage of fines (particle size < 100 μm, denoted Q₁₀₀), decreasing Q₁₀₀ approximately 2% per integer grinder setting. For specified roasting conditions and grinder setting, the Q₁₀₀ of decaffeinated coffees were up to 4% higher than the reference. For a specified coffee and grinder setting, roasting conditions modulated Q₁₀₀ by up to 4%, whilst the span of possible Q₁₀₀ values decreased as grinder setting increased.
All studied factors had a significant impact on ground colour. Longer roasting times and higher air temperatures darkened the coffee for both coffees and all grinder settings. A greater grinder setting, producing larger particles, increased the measured colour value (ColorTrack scale); for reflectance methods, larger particle sizes increase light scattering and decrease the reflected response. Relative to the colour of coffee at the lowest grinder setting (i.e., smallest particle size), ground coffee colour darkens by 1.5 colour units (ColorTrack), or 5 colour units (Colorette), for every 100 μm increase in median particle size. Decaffeination darkened green coffee colour and this propagates throughout roasting, as for all roasting conditions, the ground colour of decaffeinated coffees were darker than the reference. The difference between the whole bean (i.e., surface) and ground (i.e., average, core) colour is indicative of heat and mass transfer rates during roasting. The whole bean colour (cw) was darker than ground colour (cg) across all samples and processing. The colour difference (cw-cg) decreased as roasting progressed (i.e., as time increased), yet the impact of air temperature on the colour difference was unclear. For the reference coffee, the colour difference was lower for higher applied air temperatures, but results are unclear for the decaffeinated coffee due to sample variation.
Conclusions
The impact of EA decaffeination on coffee’s roasting and grinding performance has shown significant differences in green coffee properties, the time-temperature response during roasting, roasted whole bean properties, the resultant particle size during grinding, and roast and ground properties.
In the green coffee, distinct visual differences in colour were observed, as seen in previous studies (Ramalakshmi & Raghavan (1999) and Klikarová, et al. (2022)), but the significance of consumer affection toward decaffeinated coffees is expected to surpass their selectivity based on colour. Despite intense physical processing during decaffeination, differences in the coffee’s micro-porous network were not resolved, indicating that the bean’s matrix was unchanged at the micrometer scale, although inspection of the matrix at a cellular level is recommended to confirm these observations.
The time-temperature response of the decaffeinated coffee differed from the reference at early times due to differences in moisture content (ca. 1%) and the absence of chaff. Through first crack, the rate of change of the reference coffee’s temperature decreased moments before exothermic reactions dominated and the rate of change of temperature increased, yet the rate of change of the decaffeinated coffee’s temperature was unperturbed by apparent exothermic or endothermic contributions. The temperature response beyond first crack provides developers with an invaluable tool to modulate coffee’s development for different in-cup flavour profiles, although it’s important to realise that the temperature response at late times (i.e., after first crack) is highly dependent on the coffee’s post-harvest processing and the developer’s control strategy should reflect this.
Decaffeination had a significant impact on coffee’s roasted whole bean properties, with roasting conditions amplifying these differences in some instances. For the same roasting conditions, the decaffeinated coffee had a lower mass loss, darker colour and greater colour homogeneity than the reference coffee, but density and moisture were not influenced by decaffeination. These observations indicate that physical differentiators are insufficient to describe the impact of decaffeination on coffee’s roasting performance, and so, to fully demystify the effect of decaffeination on roasting performance, cellular inspection and chemical analyses are required.
Regarding grinding performance and the resulting ground coffee properties, decaffeination, roasting time and constant inlet air temperature each had a significant effect. For the same roasting conditions and grinder setting, the decaffeinated coffee produced granules with a lower median particle size and greater percentage of fines than the reference coffee. Roasting conditions also modulated the grinding performance, showing that developers can manipulate roasting parameters to affect the particle size distribution of ground coffees. For the tested grinder, the grinder calibrations indicate that similar particle size distributions are attainable for both decaffeinated and reference coffees using the specified grinder. However, if a café’s grinder is at the lowest setting and cannot obtain an appropriate espresso-friendly particle size distribution, which is often the case for decaffeinated coffees, developers must extend the time, or increase the temperature, of the roast to ensure a lower median particle size and greater percentage of fines in the resulting granules. If equivalency of the particle size distribution is obtained, and similar extraction parameters are used, the physical extraction responses (extraction time, total dissolved solids, etc.) are expected to be similar, although chemical and sensorial differences are anticipated. Chemical assays to complement grinding and extraction performance would yield interesting data for developers to couple with the physical data presented here.
Outlook
Decaffeination is a complex process affecting several physicochemical pathways and generating the plethora of precursors that yield a similar but distinct coffee. Whilst this study has addressed the physical aspects of decaffeination’s impact on roasting and grinding performance, many questions remain that will likely be answered via complementary extraction, chemical and sensorial studies.
Acknowledgements
This work was conceptualised during conversations between the myself (Mark Al-Shemmeri) and Alex Wallace (QC at Caravan Coffee Roasters) and with the support of Mat North (QC at Raw Material Coffee). I’d like to acknowledge that this work was performed during study for an Engineering Doctorate (EngD) with funding from EPSRC through the Centre for Doctoral Training in Formulation Engineering (grant no. EP/L015153/1) at the University of Birmingham and from Jacobs Douwe Egberts.
Data Availability
Tabular data for green and roasted coffees, ground coffee particle size distributions and time-temperature profile data are available here.