Research Article

Received: 11 October 2019 Revised: 5 March 2020 Accepted article published: 26 April 2020 Published online in Wiley Online Library: 20 May 2020

(wileyonlinelibrary.com) DOI 10.1002/jsfa.10440

Swelling properties of roasted coffee particles

Verena Bernadette Hargarten, Michael Kuhn and Heiko Briesen*

Abstract
BACKGROUND: In this study, the swelling behavior of roasted coffee particles in water and particularly its impact on particle
diameter is examined by applying laser-diffraction analysis and microscopy. Several potential influencing factors are investi-
gated: initial particle size, roasting degree, and temperature. Additionally, the time dependency of swelling and particle shape
is evaluated at two different temperatures.
RESULTS: We verify that particle erosion occurs – as observed by an increase of the fine particle fraction after wetting – and it is
revealed that this effect is more pronounced with a rise in temperature. The total relative increase in particle size is determined
as approximately 15% based on a broad range of different sized coffee grounds. It is demonstrated that the degree of swelling
is independent of both the initial particle diameter and the roasting degree. The particle shape is found to be unaffected by
swelling. This research reveals that swelling is initially quick, with 60–80% of the final steady-state diameter being reached after
30 s and completed after 4 min of wetting, i.e. within the timescale of conventional coffee brewing methods.
CONCLUSION: This work provides a better understanding of the impact of wetting as part of the coffee brewing process, thus
aiding the design, modeling, and optimization of coffee extraction. It clarifies the strong deviation of previous results on coffee-
particle swelling by considering particle erosion and degassing and provides a robust method for quantification.
© 2020 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of
Chemical Industry.

Keywords: erosion; extraction; imbibition; particle size; wetting

INTRODUCTION et al.;10 they also discovered that extraction kinetics, and thus,
In recent decades, researchers have become increasingly inter- the ratio of compounds in the final beverage, are influenced by
ested in coffee extraction. In particular, the highly complex pro- the particle size distribution.
cess of percolation conducted for espresso preparation has not It has often been stated in the literature that swelling of coffee
been fully explored, a process that combines the mass transport particles has a considerable impact on coffee extraction. This
of multiple volatile and non-volatile substances with flow through swelling process is described as an effect occurring with colloidal
an inhomogeneous particle packing. imbibition, which means the uptake of water by plant tissue fol-
Variations in brewing parameters strongly influence the extrac- lowed by chemical hydration of the biocolloids present in the cell
tion yield, the composition, and hence the sensory profile of the and an expansion of the insoluble cell wall structure.11,12 Com-
coffee beverage.1 Besides water pressure,2 temperature,3 coffee/ pounds that are abundant in roasted coffee and are assumed to
water ratio,4 extraction method,5 and water quality,6 it is known contribute to swelling due to their hydrophilicity13 are water-
that the particle size distribution of ground roasted coffee beans insoluble polysaccharides, referred to as holocellulose.
is an essential variable affecting not only the diffusion of solutes Three principal polysaccharides have been identified in green
into water but also the flow through the particle bed. Voilley coffee: arabinogalactan,14 mannan,15 and cellulose.16 Their struc-
and Simatos7 showed that a finer coffee ground leads to a higher ture and contents were analyzed by Bradbury and Halliday,17
dissolved solids content. Spiro and Selwood8 revealed an effect of who found similar contents of mannan and cellulose in arabica
the particle diameter on the extraction kinetics of caffeine for cof- and robusta beans: 2.2 g kg−1 mannan and 0.8 g kg−1 cellulose.
fee infusion by demonstrating that with decreasing particle size, The arabinogalactan content is slightly higher in robusta beans
the rate constants of caffeine extraction increased by two orders at 1.7 g kg−1, with 1.4 g kg−1 in arabica beans. During roasting,
of magnitude. Furthermore, it applies to percolation that the size holocellulose polysaccharides are thermally degraded.18 The total
of coffee particles is positively correlated to the permeability of
the particle bed according to the Kozeny–Carman equation; parti-
cle size distribution and particle shape also affect permeability by * Correspondence to: H Briesen, Chair of Process Systems Engineering, TUM
means of the pore size distribution and tortuosity.9 According to School of Life Sciences, Technical University of Munich, Gregor-Mendel-Straße
4, D-85354 Freising, Germany. E-mail: heiko.briesen@tum.de
Darcy's law,9 the flow velocity rises with increased permeability
and results in a shorter residence time corresponding to a lower Chair of Process Systems Engineering, TUM School of Life Sciences, Technical
extraction yield, as verified for espresso extraction by Kuhn University of Munich, Munich, Germany
3960

© 2020 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
Swelling properties of roasted coffee particles www.soci.org

content decreases from approximately 3.2 g kg−1 on a dry basis to penetrated into the cell lumen, which supports the hypothesis
2.5 g kg−1 when light roasted, 1.9 g kg−1 when medium roasted, that imbibition occurs.
and 1.7 g kg−1 when dark roasted, whereby cellulose remains Finer particles were analyzed by Corrochano,32 who compared
the most stable throughout the roasting process.18 Asante and laser-diffraction measurements of dry and wet particles by dis-
Thaler19 demonstrated that the water solubility of mannan persing them in an air stream and in tap water at 15 °C. He
increases during the roasting process, leading to a decrease in revealed that the fine particle fraction is higher when using the
the insoluble fraction of the bean and an increase in the viscosity wet method compared with the dry and that the wet method
of the beverage.20 Due to this change in the amount of insoluble detects finer particles. These fine particles are assumed to be fine
polysaccharides, it can be assumed that the roasting degree influ- cell fragments that are either detached from the coffee-particle's
ences swelling. Furthermore, it is understood that the swelling of pores or are lumps of substances that are insoluble at the measur-
pulp fibers composed of cellulose, lignin, and hemicellulose is ing temperature, but which are possibly soluble at higher temper-
influenced by pH and salt content,21 lignin content,22 atures. Comparing the d4,3 of two roasted coffee grounds,
temperature,23 hemicellulose,23 and fibrillar content.24 Cuissinat Corrochano found a volume increase in the d4,3 of 12% for
and Navard25 revealed that plant fibers mainly composed of cellu- medium-coarse particles with an initial d4,3 of 363.6 ± 3.8 μm,
lose and hemicellulose and small contents of lignin and pectin but no significant increase in the d4,3 for fine particles with an ini-
exhibited homogeneous swelling without dissolution in tial d4,3 of 198.8 ± 0.9 μm. According to Corrochano, the strong
0.76 g kg−1 sodium hydroxide (NaOH) and in most aqueous deviation from Mateus et al.'s13 results might be attributed to a
N-methylmorpholine-N-oxide (NMMO) solutions. The ratio of different dispersant or the roasting degree. The absence of signif-
swelling decreased with lower concentrations of the solvent. icant swelling of fine particles in Corrochano's work may lead to
According to Wolfrom et al.,15 mannan is partially soluble in a the assumption that the degree of swelling depends on the initial
1.8 g kg−1 aqueous NaOH solution. Based on these findings on size of coffee particles. Accordingly, this aspect is addressed in this
the individual insoluble polysaccharides comprising the coffee study.
matrix, the occurrence of swelling in water is basically presumable It is clear from the widely differing values and contradictory
but possibly depends on the composition of coffee ground and statements mentioned earlier that no consensus exists about
solvent. the occurrence and degree of coffee-particle swelling. Addition-
Rivetti et al.26 concluded that swelling is influenced by water ally, relevant factors – such as the initial particle size, the roasting
alkalinity resulting in an increase in percolation time. They dem- degree, and temperature – have not been investigated suffi-
onstrated by means of discriminant analysis that alkalinity has a ciently. This work aims to clarify the effect of water on the particle
significant effect on the swelling of coffee particles using dimensions of ground roasted coffee in a statistically significant
particle-size-distribution measurements; however, they did not and reproducible way; it considers the entire range of grinding
present data concerning the quantitative size increase of parti- degrees applied in common brewing methods as well as the influ-
cles. Measurements of particle size and particle size distribution ences of temperature and the degree of roasting. The effect of
have been applied by several other authors during recent water quality, i.e. alkalinity/acidity and mineral content is not con-
decades, but their results differ widely. Sivetz and Desrosier27 sidered in this article. In addition to laser-diffraction measure-
reported a size increase of 7% based on microscopy measure- ments of sieved samples, microscopy is applied to assess the
ments without presenting any information about the applied time evolution of swelling and its relevance for different extrac-
method and the initial particle size. Spiro et al.28 determined an tion techniques.
increase in size of 20 ± 4% for green coffee and an increase of With regard to the influence of temperature on swelling, Spiro
17 ± 5% for roasted coffee with an initial mean diameter of and Chong33 state that swelling is more pronounced at elevated
1.072 and 0.994 mm respectively. Measurements were limited to temperatures but indicate that obtaining quantitative measure-
relatively large particles and visual inspection, which does not ments was challenging. Through microscopy measurements of
enable a conclusion to be drawn for smaller particles applied for single particles at two different temperatures (25 and 80 °C), this
espresso and drip coffee brewing. The swelling of single particles study investigates the effect of temperature on both the rate
was questioned by Hinz et al.29 and Steer30 as they reported a vol- and the degree of swelling.
ume increase for whole coffee beans and pressed tablets of coffee
particles but no increase regarding individual particles under the
microscope. Furthermore, Hinz31 determined the particle size dis- MATERIALS AND METHODS
tribution by laser-diffraction analysis of coffee particles after dif- Roasted coffee
ferent periods of extraction. An increase in the fine fraction of The medium-roast coffee (of a single origin from Marcala, Hondu-
particles < 250 μm was detected along with a slight shift of ras, 100% arabica, organic certified DE-ÖKO-039) was supplied in
coarser particles toward larger particle sizes. In that work, no dis- 250 g packages (of a synthetic polymer with an aluminum barrier
persing unit was mentioned, but Hinz states that the moisture and a one-way degassing valve) by a local roaster (Fausto Kaffeer-
contents of the measured particles were between österei GmbH, Munich, Germany). The light-roast coffee (of mixed
0.047 × 10−3 g kg−1 and 1.847 × 10−3 g kg−1, indicating that origin from Colombia, South America and Tanzania, Africa, 100%
moist particles were measured in an air stream. As the particle size arabica, rainforest alliance certified) was a product called Blonde
differs with varying moisture content, there are advantages to Roast (produced by Tchibo GmbH, Bremen, Germany), which
measuring particle size distribution in an aqueous medium to was packed in 250 g packages (of low-density polyethylene with
eliminate this degree of freedom. Such measurements in a wet aluminum barrier and no degassing valve). Both raw materials
state were applied by Mateus et al.,13 who found an increase of are depicted in Fig. 1. From visual comparison with images pre-
the d4,3 by 20–23% 10 to 15 min after wetting. They measured sented by Wang and Lim,34 the roasting degree of both products
particles with an initial size of 750 and 1050 μm and they also is estimated to be between the first and second crack, i.e. within a
3961

demonstrated (by using scanning electron microscopy) that water conventional range for customary roast coffee.

J Sci Food Agric 2020; 100: 3960–3970 © 2020 The Authors. wileyonlinelibrary.com/jsfa
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
 www.soci.org VB Hargarten, M Kuhn, H Briesen

either in a dry state by air pressure (VIBRI and RODOS, Sympatec

GmbH) or in circulated, demineralized water (QUIXEL, Sympatec
GmbH). Data from the detector were collected and evaluated
(based on the Fraunhofer diffraction theory) by a compatible
software (WINDOX Software, Sympatec GmbH). The measuring
ranges of the lenses used for different fractions are shown in
Table 1. Interval classification differs between the measuring
ranges and along the scale of particle sizes with increasing inter-
val sizes for larger particle fractions. Volume distributions were
determined, i.e. the cumulative and density distribution of the
volume of spheres with a diffraction-equivalent diameter. The
volume density describes the difference quotient of the cumula-
Figure 1 Light roast (left) and medium roast (right) coffee beans. tive distribution for a specific particle size interval given by the
measuring range.
For dry measurements, the sieved samples were split (DR 1000,
Grinding and sieving RETSCH GmbH, Haan, Germany; Laborette Type 10.102, FRITSCH
The coffee was ground using a professional grinder (Mahlkönig EK GmbH, Idar-Oberstein, Germany) to achieve homogeneous sam-
43, Hemro Manufacturing Germany GmbH, Hamburg, Germany) ples and weighed to ensure an equal sample weight of 10.00 g
equipped with cast steel grinding disks and a continuous dial (FB6CCE-H, Sartorius AG, Göttingen, Germany). Three samples
for grind adjustment (scaled between 1–11 in increments of per sieving fraction were measured at a conveying speed of
0.1). Good reproducibility of grinding was confirmed in prelimi- 80% relative to maximum speed and a primary pressure of 2.5 bar.
nary experiments. To prevent contamination of the grinder with Wet measurements were conducted after particle-water contact
residual particles, a small amount of coffee was ground and dis- of 20 min, after which the steady state of swelling was assumed.13
carded before collecting the samples. The coffee was ground at Briefly, 10 g of particles were immersed in 300 mL of boiled, demi-
settings of 2.0 (fine), 5.0 (medium), and 8.0 (coarse), which are sub- neralized water at a temperature of 90 °C on a heated magnetic
sequently referred to as grinding degree. Directly after grinding, stirrer and were stirred for 20 min. During this period, the temper-
the coffee was partially sieved in a vibrational sieving tower ature decreased to approximately 60 °C due to heat loss. To inves-
(AS 200, RETSCH GmbH, Haan, Germany). Approximately 50 g of tigate the effect of extraction temperature, stirring was also
ground coffee was sieved continuously for 10 min at an amplitude performed at a temperature of 25 °C, following the same proce-
of 1.2 mm. Using a soft brush, agglomerates were separated and dure. Subsequently, non-sieved samples from the stirred suspen-
fine particles adhering to the sieves were detached. Subse- sion were inserted directly into the dispersion unit of the laser-
quently, the sieving procedure was repeated as described earlier. diffraction instrument with a Pasteur pipette. Before the analysis
In the case of grinding degree 2.0, vibrational sieving was con- commenced, ultrasound was applied for 60 s to separate agglom-
ducted three times for 2 min at 10 s intervals, and between each erates; after a pause of 10 s, laser-diffraction measurements were
sieving step the adhesion and cohesion of the particles was coun- conducted four times for 10 s with 5 s pauses in between. The
teracted using the brush. The sieves used and the corresponding pumping velocity in the system was 30% (relative to maximum
grinding settings and fractions, described as F (fine), S (small), M velocity) for all experiments. In the case of the sieved particle
(medium), and L (large), are listed in Table 1. Air jet screening fractions S, M, and L, the extracted coffee ground was separated
(LPS 200 MC, RHEWUM GmbH, Remscheid, Germany) was applied from water in a 200 μm sieve and washed for 5 min with deminer-
to remove adhering fines from the particles and to further sepa- alized water to remove the fine particles. This step was essential to
rate agglomerates. This was conducted for 5 min at an air flow approximate monomodal particle size distributions, which are
of 70 m3 h−1 and a rotation speed of 50 rpm. necessary for accurate quantification of differences and to
exclude the effects of particle erosion and fluctuations in optical
Laser-diffraction analysis density throughout the measurements. After adding the coffee
Particle-size-distribution measurements were conducted with a particles to the dispersant and waiting 5 min to reach a constant
laser-diffraction system (HELOS/BR, Sympatec GmbH, Clausthal- optical density, measurement commenced. For a period of
Zellerfeld, Germany). Particles were supplied and dispersed 120 s, diffraction was analyzed; the measurement was repeated

Table 1. Specification of grinding and sieving settings and corresponding measuring range of laser diffraction analysis

Sieving Grinding degree Sieving range (μm) Measuring range (μm)

None 2 — 0.5–1750
Fa 2 < 200 0.5–875
Sb 2 200–300 0.5–1750
Mc 5 400–500 0.5–1750
Ld 8 710–800 0.5 – 3500
a
F (fine).
b
S (small).
c
M (medium).
3962

d
L (large).

wileyonlinelibrary.com/jsfa © 2020 The Authors. J Sci Food Agric 2020; 100: 3960–3970
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Swelling properties of roasted coffee particles www.soci.org

four times with 5 s pauses. No ultrasound was used to avoid an was pumped through at temperatures of approximately 80 and
undesired degradation of particles. Four to five samples were ana- 25 °C at a flow rate of 50 to 80 mL min−1 respectively using a peri-
lyzed per sieved fraction, and three to five samples were analyzed staltic pump (TU/200, medorex e. K., Nörten-Hardenberg, Ger-
from three ground batches with reference to the non-sieved sam- many). Prior to each experiment, a beaker on a magnetic stirrer
ples. Weighted mean values and variances, outliers, and 95% con- (serving as a water reservoir for the pump) was filled with fresh
fidence intervals (assumption of t-distribution) were calculated in demineralized water; in the case of the measurements at 80 °C
Microsoft Excel 2016. Statistical analysis was conducted in it was preheated with boiling demineralized water. The flow cell,
MATLAB R2018a software using the functions ttest2 and ranksum hoses, and beaker were isolated with rubber foam (adhesive rub-
to identify significant differences between particle size classes. ber tape 5 m x 50 mm x 3 mm, HORNBACH Baumarkt AG, Born-
Additionally, the functions fitlm combined with anova were used heim, Germany; insulating tube and insulating mat: 9 mm
to check for linear correlation. The relative diameter increase insulating thickness, 6 mm tube diameter, Koste/Weitzel GbR,
was calculated as: Wiesloch, Germany) to reduce heat loss. The temperature inside
the flow cell (after passing through the beaker, pump, and hose)
ðd f −d0 Þ was 80 °C (temperature sensor: HI 98509, Hanna Instruments
Δdrel =  100%: ð1Þ
d0 Inc., Woonsocket, RI, USA). A characteristic extraction temperature
of 90 °C was not achieved due to heat loss despite insulation mea-
where df defines the final diameter of the swollen particle and d0 sures and preheating. Shortly before the water boundary surface
the initial diameter. The relative diameter increase was deter- had reached the particles, video recording commenced and then
mined for the diameters d10,3, d16,3, d50,3, d84,3, d90,3, and d99,3 stopped after 150 s. Images were taken from this video for analy-
which correspond to a specific quantile of the volume distribu- sis after the beginning of wetting at 10, 20, 30, 40, 50, 60, 90, and
tion, where the first index represents the respective quantile 120 s, after which single images were regularly taken at defined
expressed in a percentage. This description must be distinguished time intervals. The water temperature in the beaker reached
from the De Brouckere mean diameter d4,3 which defines the vol- approximately 60 °C at the end of the experiment due to inevita-
ume moment mean of the particle size distribution. Based on the ble heat loss, yielding a temperature profile close to the extraction
calculations using Eqn (1) for the previously mentioned quantiles, conditions described for laser-diffraction analysis (see earlier).
the mean diameter increase and the respective standard devia- All photographs were transformed into binary images and the
tion was calculated for different sized particles. Outliers, (occur- particles' projection area Aproj as well as the maximum and mini-
ring mostly for fine particles where the standard deviation was mum Feret diameter dF, max and dF, min were determined. Any bub-
large) were excluded, but through comparison with the result bles not removed by flow were separated from the particle outline
for a complete data set, it was revealed that this did not change using image processing. Based on the collected data, the
the mean value (see Results section). projection-area-equivalent diameter of a circle dproj was calcu-
lated as
Microscopy
rffiffiffiffiffiffiffiffiffiffiffiffiffi
A light microscope (BX51, OLYMPUS Corporation, Tokyo, Japan) 4
was used for single-particle measurements. The microscope was dproj = Aproj ð2Þ
π
equipped with a motorized revolving nosepiece (BX-REMCB & U-
HS, OLYMPUS Corporation), a scanning stage (SCAN, 130 × 85 and the aspect ratio AR was computed as the ratio of the mini-
travel range, 4 mm ball screw pitch, Märzhäuser Wetzlar GmbH & mum to the maximum Feret diameter, as follows:
Co. KG, Wetzlar, Germany), a motorized focus drive (MFD-2, März-
häuser Wetzlar GmbH & Co. KG), and a stepper motor controller d F,min
connected to a joystick (TANGO 3 Desktop & 3-Axes Joystick, AR= : ð3Þ
d F,max
Märzhäuser Wetzlar GmbH & Co. KG). This enabled the fine adjust-
ment of focus and table position. Images and videos were The relative diameter increase was determined according to
recorded using a high-resolution camera (XC50, OLYMPUS Corpo- Eqn (1) using dproj at time t as diameter of the swollen particle
ration) and analyzed in a compatible software (analySIS, OLYM- and dproj in the dry state (t = 0) as initial diameter. The progress
PUS Corporation). of swelling was determined as
To remove gas bubbles from the surface of coffee particles that
appeared with degassing, a sealed flow cell was constructed from ðd ðtÞ−d0 Þ
acrylic glass, which enabled the analysis of particles surrounded Progress of swelling=  100%, ð4Þ
ðd f −d 0 Þ
by water flow. Without removal, these gas bubbles would have
disturbed the measurement of the projection area as they where d(t) is dproj at time t and df is the final projection area equiv-
obscure the outline of particles. The coffee particles were affixed alent diameter 20 min after the beginning of wetting.
on a cover glass using a water-resistant, transparent, two-
component adhesive (epoxy resins and amines, UHU GmbH &
Co. KG, Bühl, Germany) and fixed inside the flow cell with trans-
parent tape (tesafilm transparent, tesa SE, Norderstedt, Germany). RESULTS
The adhesive layer was spread as thin as possible to prevent Impact of swelling on the particle size distribution
immobilization of the particles in the direction of the projection As described, non-sieved medium-roast coffee grounds were ana-
plane. All particles analyzed were within the size range of 400 to lyzed at two different extraction temperatures: 25 and 90 °C. The
1000 μm, which reflects the coarse fraction of coffee grounds volume distributions of these samples are presented in Fig. 2. The
applied for drip and infusion brewing. Smaller particles were not presented error bars represent 95% confidence intervals, which
3963

measurable due to handling difficulties. Demineralized water applies to all particle size distributions subsequently presented.

J Sci Food Agric 2020; 100: 3960–3970 © 2020 The Authors. wileyonlinelibrary.com/jsfa
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
 www.soci.org VB Hargarten, M Kuhn, H Briesen

treated with caution as separation by means of sieving is not

achieved with sufficient precision.
The relative diameter increase is plotted against the initial parti-
cle diameter in Fig. 4. Data from all particle fractions F, S, M, and L
after the exclusion of three outliers (d10,3 and d50,3 of fraction F
and d99,3 of fraction S) is shown. The presented values result in a
mean relative diameter increase of 15 ± 4%. Including the com-
plete dataset in the calculation yields a relative diameter increase
of 15 ± 7%. A linear regression coupled with an analysis of vari-
ance for the linear model's slope reveals no significant correlation
between relative diameter increase and initial particle diameter
(see Table A2, Appendix). Thus, it is concluded that the diameter
increase is independent of the initial particle size when referring
to the size range presented in Fig. 4.
To investigate possible effects from the type of roasting, mea-
Figure 2 Density ( ) and cumulative ( ) distributions of non-sieved surements were repeated with a light-roast coffee, as described
medium-roast coffee particles in dry state (white), wetted at 25 °C (gray) earlier. The dry and wet particle size distributions for sieving
and wetted at 90 °C (black).
fractions F, S, and M are shown in Fig. 5. The same shift of the
modes and cumulative distribution curves toward larger particle
For better visibility of the fine fraction, the scaling of the abscissa sizes is visible as for medium-roast coffee (see Fig. 3) with respect
is logarithmic. For clarity, it should be noted that the area of the to sieving fractions S and M. Sieving fraction F does not exhibit a
density distribution does not represent the actual volume propor- shift of the mode located at around 40 μm, but the cumulative
tion due to its logarithmic transformation, i.e. the fine fraction is distribution is shifted to the right for particles > 100 μm. It is
distinctly smaller than it appears from this plot. Each data point assumed that the very fine particles located at the distribution's
represents the middle of the respective particle size interval. For left end (which are not measured in a dry state due to their adher-
the dry coffee grounds, a bimodal distribution typical for coffee ence on the larger particle's surface) are responsible for this differ-
is recognizable with the first mode between 20 and 30 μm ence compared to larger fractions where separation is feasible.
and the second at approximately 400 μm. Comparing the cof- The separation of agglomerates in a wet state might also be
fee grounds' distribution in dry and wet state, it is evident that influential.
the volume density for particles < 40 μm is significantly higher It is evident from Fig. 6 that the mean relative diameter increase
after wetting. For the coarse fraction, the volume density for for the sieved light-roast coffee ground is identical to that for
particles > 250 μm is reduced for wet particles. This result is medium-roast coffee at 15 ± 3% when the three extreme outliers
confirmed by a two-sample t-test and a Mann–Whitney U-test d10,3 and d16,3 of fraction F and d99,3 of fraction M are excluded.
at the 1% significance level, summarized in Table A1 in the Including outliers slightly shifts the mean value to 13 ± 10%.
Appendix. In a wet state, both modes are shifted to smaller par- Additionally, no significant correlation is found between relative
ticle sizes; the first one is located at 12 μm and the second at size increase and initial particle diameter, as shown in Table A2
330 μm. Additionally, it is evident that very fine particles in the Appendix. These results imply that the roasting degree
appear with the wet samples, which are small enough to does not significantly influence the degree of swelling if the water
exceed the lower threshold of the measuring range quality and temperature is kept constant.
(< 10 μm). A further increase in the fine particle fraction can
be seen with increased extraction temperature. With reference Time dependency
to the distribution of hot, extracted coffee grounds, the volume Time courses of the relative diameter increase Δdrel for individual
density of particles < 50 μm is significantly increased com- particles, which were determined by microscopy according to
pared to the wet measurements at an extraction temperature Eqn (1), are presented in Fig. 7(a) for the medium-roast coffee
of 25 °C and is simultaneously decreased in the range and in Fig. 7(b) for the light-roast coffee. As the final relative diam-
90–500 μm (see Table A1, Appendix). eter increase differs strongly for particles measured at the same
Figure 3 shows the particle size distributions of the sieving frac- conditions and exhibiting similar initial particle sizes, image anal-
tions F, S, M, and L (for specifications see Table 1) derived from ysis by microscopy is found to be an inappropriate method for the
medium-roast coffee grounds. The lines between values enable quantification of total swelling. In Table A3 (Appendix), statistical
accurate readability of the individual distributions. A shift of the results are listed for the comparison of total swelling for different
entire density and cumulative distribution toward larger particle temperatures and roasts. No significant difference in the total size
sizes after wetting can be seen in the coarse fractions S, M, and increase is found for different roasting degrees due to the strong
L. This shift indicates a size increase of the particles. A small peak deviation of measurements confirming the result from laser-
at the density distribution's left end is still visible for the wet dis- diffraction analysis. Comparing the results determined at different
persion. However, the proportion of these fine particles is small water temperatures (80 versus 25°C), no significant difference
due to the washing action described earlier; therefore, their influ- between the medium-roast and the light-roast coffee is evident,
ence on the cumulative distribution can be neglected. This is not which is also based on the large scattering within both series of
the case for the fine fraction F, where a strong increase in the vol- experiments. At 25 °C single, markedly higher diameter increases
ume density is still visible at the distribution's left end. A signifi- are measured for both roasts. However, this is possibly caused by
cant shift of the cumulative distribution on the right side of the the growth and adhesion of gas bubbles that could not be
mode is detected, and the mode itself shifts slightly from around completely removed by water flow at lower temperatures and
3964

28 μm to 34 μm. Nonetheless, results in this size range must be that had to be distinguished and subtracted from the particles

wileyonlinelibrary.com/jsfa © 2020 The Authors. J Sci Food Agric 2020; 100: 3960–3970
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Swelling properties of roasted coffee particles www.soci.org

Figure 3 Density ( ) and cumulative ( ) distributions of sieved medium- Figure 5 Density ( ) and cumulative ( ) distributions of sieved light-
roast coffee particles in dry state (white) and wetted at 90 °C (black). (a) roast coffee particles in dry state (white) and wetted at 90 °C (black). (a)
Sieving fraction F (fine). (b) From left to right: sieving fractions S (small), Sieving fraction F (fine). (b) From left to right: sieving fractions S (small),
M (medium), and L (large). M (medium), and L (large).

Fig. 8. Large and small gas bubbles as well as cell fragments adher-
ing to the particle surfaces are visible. Additionally, a detachment
of small particles and color pigments carried away with the fluid
flow was observed, complying with the increase in fines detected
by laser diffraction analysis.
In addition to the particles' projection area, the minimal and
maximal Feret diameter was determined and used for calculating
the aspect ratio as a characteristic descriptor of particle shape
(Eqn (3)). Figure 9 shows the aspect ratio of the same particles
as depicted in Fig. 7, plotted over time. Most particles exhibit an
aspect ratio between 0.6 and 0.8, visible for both roasts. This con-
firms that coffee particles within the measured size range are not
spherical in shape but instead are flat and elongated. It is revealed
by applying linear regression and analysis of variance (ANOVA)
that no significant change occurs during swelling, as presented
in Table A4 in the Appendix.
Figure 4 Relative diameter increase for medium-roast coffee versus initial Figure 7 reveals that the final degree of swelling is reached
particle diameters of sieving fractions F (fine) ( ), S (small) ( ), M (medium) within the first couple of minutes. Small decreases and fluctua-
( ), and L (large) ( ). tions of the relative diameter increase are visible, but these are
based on the experimental error from detaching fragments and
gas bubbles. To investigate the time dependency of swelling,
during image processing. However, this could not guarantee the the progress of swelling described by Eqn (4) is shown for the first
complete exclusion of undetected bubbles from the measured 5 min in Fig. 10. No clear distinction in terms of time dependency
projection area. Two exemplary images showing light-roast parti- is possible when comparing different roasts and water tempera-
3965

cles before and after a wetting period of 20 min are presented in tures due to the large differences between individual particles.

J Sci Food Agric 2020; 100: 3960–3970 © 2020 The Authors. wileyonlinelibrary.com/jsfa
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
 www.soci.org VB Hargarten, M Kuhn, H Briesen

Figure 6 Relative diameter increase for light-roast coffee versus initial

particle diameters of sieving fractions F (fine) ( ), S (small) ( ), and M
(medium) ( ).

The results for light-roast coffee show a broader scattering of the

values for the first 100 s, which can be attributed to the slightly
larger sample size. However, estimations about the status of swell-
ing in the scope of characteristic brewing times are possible.
Table 2 lists the average values for two conventional brewing
times: 30 s as a typical residence time for espresso and 4 min as
a usual contact time for infusion methods. After 30 s, around
83% of the final diameter is reached (on average) both at
80 and 25 °C for medium-roast coffee; the average for light-roast
coffee is 71% and 59%, respectively. After 4 min of wetting, a
steady state is reached. Values > 100% arise from relating the
diameter increase to the diameter increase after 20 min (see
Eqn (4)), which is subject to experimental error; therefore, it is par-
tially lower than previous measures.
Figure 7 Relative diameter increase after different times of wetting at 80 °
C ( ) and at 25 °C ( ). (a) Medium-roast coffee particles with initial diame-
ters (in μm) of 608 ( ), 519 ( ), 675 ( ), 585 ( ), 543 ( ), and 914 ( ). (b)
DISCUSSION Light-roast coffee particles with initial diameters (in μm) of 652 ( ), 626
Particle erosion ( ), 462 ( ), 665 ( ), 653 ( ), 547 ( ), 472 ( ), 589 ( ), 520 ( ), 568 ( ),
Using laser-diffraction analysis and microscopy, it can be con- and 592 ( ).
firmed that during dispersion in water, fine particles < 40 μm
adhering to the surface of coffee particles are detached by the
surrounding water flow. Regarding the total particle size distribu- and water ingress and probably enables more fragments to be
tion, a significant increase of the fine fraction is discernible. This washed from pores and capillaries. Accordingly, the effect of tem-
effect of particle erosion was previously observed by Corro- perature on particle erosion must be taken into account – espe-
chano32 and stated to be relevant for extraction modeling by cially for filtration methods where fine particles have a strong
Ellero and Navarini.35 A contribution by lipid droplets to the fine influence on the specific cake filtration resistance.35,38
fraction can be neglected as the lipid content in coffee extracts
is low compared to the volume of fine particles.36 Moreover, this Degree of swelling
study revealed a remarkable increase in the fine fraction with an Particle size distribution analysis yielded a total increase of the
increase in extraction temperature. This positive correlation of particle diameter by approximately 15%. This value is close to
the fine fraction with temperature refutes Corrochano's second the results obtained by Spiro et al.28 but is higher than the results
hypothesis that the detected particles were lumps of solutes, from laser-diffraction analysis subsequently produced by other
which are soluble at higher temperatures. Instead, this research authors.13,31,32 This difference can be explained by the effect of
proposes that these fines are caused by particle erosion. It is particle erosion. The increase of the volume density for fine parti-
assumed that the increase of solubility with increasing cle size classes connected with the simultaneous decrease for
temperature,3 (e.g. valid for bitter components such as caffeine), larger particle size classes leads to a change in the bimodal distri-
leads to a higher pore accessibility and thus enables the release bution, thereby affecting the value of the d4,3. Thus, the d4,3
of more fine fragments into the liquid phase. Moreover, a reduc- (including all particle size classes) is underestimated along with
tion of the liquid's surface tension with extraction and tempera- the total increase calculated from the difference to the dry state.
ture (according to the work of Navarini et al.37) as well as the By previous sieving, air jet screening, and washing, the effect of
3966

drop in water viscosity with rising temperature facilitates wetting particle erosion can be excluded – except for fine particles

wileyonlinelibrary.com/jsfa © 2020 The Authors. J Sci Food Agric 2020; 100: 3960–3970
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Swelling properties of roasted coffee particles www.soci.org

Figure 9 Aspect ratio after different times of wetting at 80 °C ( ) and at

25 °C ( ). (a) Medium-roast coffee particles with initial diameters (in μm)
Figure 8 Light-roast particles in a dry and swollen state at 25 °C; gas bub- of 608 ( ), 519 ( ), 675 ( ), 585 ( ), 543 ( ), and 914 ( ). (b) Light-roast
bles and loosening fragments are visible at the surface of particles. coffee particles with initial diameters (in μm) of 652 ( ), 626 ( ), 462 ( ),
665 ( ), 653 ( ), 547 ( ), 472 ( ), 589 ( ), 520 ( ), 568 ( ), and 592 ( ).
(a) Before wetting. (b) After 20 min of wetting.

< 100 μm where separation efficiency was insufficient. This Time dependency of swelling
explains why this study's results for the total increase of the parti- Image analysis revealed that the rate and final degree of swelling
cle diameter are higher. Moreover, the analysis of cumulative dis- varies widely when comparing individual particles, which makes a
tributions instead of the d4,3 allowed an investigation of individual verification of laser-diffraction measurements difficult. Addition-
regions of the distributions and to exclude outliers at the distribu- ally, the wide scattering of values and the appearance of gas bub-
tions' left end, (which occurred due to residual fines and experi- bles complicated the investigation of influencing factors on
mental error). swelling kinetics such as temperature and roasting degree. The
large differences between single particles can be attributed to
Influence of initial particle diameter and roasting degree their individual physical structure, e.g. of pores and fibers, as well
Neither by analysis of the particle size distribution nor by investi- as their composition (depending on where the respective cells
gating individual particles under the microscope could a differ- were located in the bean before grinding) and on the whole
ence in swelling be seen with different initial particle diameters. growth and processing history of the beans. To enable statistically
An impact of the initial diameter (as might be assumed from Cor- significant results, vastly larger sample sizes would be needed,
rochano's32 results) does not seem to exist, which implies that the along with an automated method of analysis. However, it can be
relative degree of swelling is not affected by grinding variations. stated that swelling is quick and not negligible for the regular
The same applies to the roasting degree, as no significant differ- time of espresso brewing, as approximately 60–80% of the final
ence was observed for light-roast coffee compared to medium diameter is reached after only 30 s. Regarding espresso extrac-
roast. With regard to the content of insoluble polysaccharides, tion, however, it has to be considered that the effect of an ele-
only the cellulose content remains virtually constant throughout vated pressure on the swelling dynamics is not covered by this
roasting.18 Thus, it is hypothesized that cellulose has the highest study and potential effects from this parameter require further
influence on swelling according to the fact that no difference in investigations. With a typical duration of infusion brewing of
swelling was observed for different roasting degrees at identical around 4 min, it can be expected that swelling is completed
3967

water quality and temperature. according to the measurements in this research. Therefore, the

J Sci Food Agric 2020; 100: 3960–3970 © 2020 The Authors. wileyonlinelibrary.com/jsfa
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
 www.soci.org VB Hargarten, M Kuhn, H Briesen

Table 2. Average progress of swelling after conventional brewing

times

Progress of swelling

Water temperature Medium roast Light roast

30 s of wetting
80 °C 83% 71%
25 °C 82% 59%
4 min of wetting
80 °C 109% 101%
25 °C 111% 108%

grounds used for different brewing techniques. Two different

conventional roasts have been compared with regard to their
total degree of swelling. Both roasts exhibit the same size
increase, leading to the conclusion that swelling is not influenced
significantly by the roasting degree.
We furthermore reveal by measurements at different times after
wetting that swelling is homogeneous in all directions and that it
is completed within the first 4 min. A remarkable size increase
already appears within the first 30 s. These findings emphasize
the requirement of further research on potential effects from
swelling on the extraction kinetics at specific extraction
conditions.

ACKNOWLEDGEMENTS
Thanks to Fausto Kaffeerösterei GmbH, Munich, Germany, for sup-
plying freshly roasted coffee beans. Special thanks go to the work-
shop of the Chair of Process Systems Engineering of Technical
University of Munich for manufacturing the flow cell that has been
Figure 10 Progress of swelling after different times of wetting at 80 °C ( ) used for microscopy and to Johann Landauer for providing valu-
and 25 °C ( ). (a) Medium-roast coffee particles with initial diameters (in able advice on particle analysis.
μm) of 608 ( ), 519 ( ), 675 ( ), 585 ( ), 543 ( ), and 914 ( ). (b) Light-roast
coffee particles with initial diameters (in μm) of 652 ( ), 626 ( ), 462 ( ), 665
( ), 653 ( ), 547 ( ), 472 ( ), 589 ( ), 520 ( ), 568 ( ), and 592 ( ).
REFERENCES
1 Voilley A, Sauvageot F, Simatos D and Wojcik G, Influence of some pro-
cessing conditions on the quality of coffee brew. J Food Process Pres
swelling duration is expected to be distinctly shorter than that 5:135–143 (1981).
2 Andueza S, Maeztu L, Dean B, de Peña MP, Bello J and Cid C, Influence
stated by Mateus et al.13 of water pressure on the final quality of arabica espresso coffee.
Furthermore, it can be concluded from the unchanged aspect Application of multivariate analysis. J Agric Food Chem 50:
ratio of particles and from visual observation (see Fig. 8) that cof- 7426–7431 (2002).
fee particles swell rather homogeneously, i.e. volume increase is 3 Andueza S, Maeztu L, Pascual L, Ibáñez C, de Peña MP and Cid C, Influ-
isotropic and particle shape is barely influenced by swelling. This ence of extraction temperature on the final quality of espresso cof-
fee. J Sci Food Agric 83:240–248 (2003).
perception is consistent with the swelling behavior of cellulose 4 Andueza S, Vila MA, Paz de Peña M and Cid C, Influence of coffee/water
being homogeneous in aqueous media.25 ratio on the final quality of espresso coffee. J Sci Food Agric 87:
586–592 (2007).
5 Gloess AN, Schönbächler B, Klopprogge B, D`Ambrosio L, Chatelain K,
Bongartz A et al., Comparison of nine common coffee extraction
CONCLUSIONS methods: instrumental and sensory analysis. Eur Food Res Technol
Using laser diffraction analysis and microscopy, we have demon- 236:607–627 (2013).
strated that single particles of ground coffee swell distinctly as a 6 Navarini L and Rivetti D, Water quality for espresso coffee. Food Chem
consequence of wetting. The total size increase amounts to 122:424–428 (2010).
7 Voilley A and Simatos D, Modeling the solubilization process during
approximately 15%. The strong discrepancies in values measured coffee brewing. J Food Process Eng 3:185–198 (1979).
by different authors in the past is traced back to the phenomenon 8 Spiro M and Selwood RM, The kinetics and mechanism of caffeine infu-
of particle erosion which is increased with a rise in extraction tem- sion from coffee: the effect of particle size. J Sci Food Agric 35:
perature from 25 to 90 °C. A sieving method is proposed in this 915–924 (1984).
9 Bear J, Modeling Phenomena of Flow and Transport in Porous Media.
article to circumvent this distortion of the volume distribution. Springer International Publishing, Cham (2018).
The final degree of swelling is shown to be independent of the
3968

10 Kuhn M, Lang S, Bezold F, Minceva M and Briesen H, Time-resolved

initial particle size. Our results are hence transferable to different extraction of caffeine and trigonelline from finely-ground espresso

wileyonlinelibrary.com/jsfa © 2020 The Authors. J Sci Food Agric 2020; 100: 3960–3970
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Swelling properties of roasted coffee particles www.soci.org

coffee with varying particle sizes and tamping pressures. J Food Eng 25 Cuissinat C and Navard P, Swelling and dissolution of cellulose, part III:
206:37–47 (2017). plant fibres in aqueous systems. Cellulose 15:67–74 (2008).
11 Illy A and Viani R, Espresso Coffee. The Science of Quality, 2nd edn. Else- 26 Rivetti D, Navarini L, Cappuccio R, Abatangelo A, Petracco M and
vier Academic, Amsterdam (2005). Suggi-Liverani F, Effect of water composition and water treatment
12 MacDougal DT and Spoehr HA, Growth and imbibition. Proc Am Philos on espresso coffee percolation. In ASIC Proceedings, 19th Interna-
Soc 56:289–352 (1917). tional Scientific Colloquium on Coffee, Trieste, Italy (2001).
13 Mateus M-L, Rouvet M, Gumy J-C and Liardon R, Interactions of water 27 Sivetz M and Desrosier NW, Coffee Technology. AVI Publishing Com-
with roasted and ground coffee in the wetting process investigated pany, Westport, CT (1979).
by a combination of physical determinations. J Agric Food Chem 55: 28 Spiro M, Toumi R and Kandiah M, The kinetics and mechanism of caf-
2979–2984 (2007). feine infusion from coffee: the hindrance factor in intra-bean diffu-
14 Wolfrom ML and Patin DL, Carbohydrates of the coffee bean. IV. An sion. J Sci Food Agric 46:349–356 (1989).
arabinogalactan 1. J Org Chem 30:4060–4063 (1965). 29 Hinz T, Steer A, Waldmann C, Cammenga HK and Eggers R, Röstkaffee-
15 Wolfrom ML, Laver ML and Patin DL, Carbohydrates of the coffee bean. Extraktion: Einflußparameter und Modellierung. Chem Ing Tech 69:
II. Isolation and characterization of a mannan 1. J Org Chem 26: 685–690 (1997).
4533–4535 (1961). 30 Steer AG, Physikalisch-chemische Parameter des Kaffeegetränkes und
16 Wolfrom ML and Patin DL, Coffee constituents, isolation and character- Untersuchungen zur Röstkaffee-Extraktion, 1st edn. Cuvillier Verlag,
ization of cellulose in coffee bean. J Agric Food Chem 12:376–377 Göttingen (2003).
(1964). 31 Hinz, T. Strukturbeschreibung disperser Naturstoffe als Grundlage für
17 Bradbury AG and Halliday DJ, Chemical structures of green coffee bean Modellentwicklungen. Doctoral Thesis, Technical University Ham-
polysaccharides. J Agric Food Chem 38:389–392 (1990). burg, Hamburg (1997).
18 Thaler H and Arneth W, Untersuchungen an Kaffee und Kaffee-Ersatz. 32 Corrochano, B. R. Advancing the Engineering Understanding of Coffee
XIII Mitteilung Verhalten der Polysaccharid-Komplexe des rohen Extraction. Doctoral Thesis, The University of Birmingham, Birming-
Arabica-Kaffees beim Rösten. Zeitschrift für Lebensmittel- ham (2015).
Untersuchung und Forschung 140:101–109 (1969). 33 Spiro M and Chong YY, The kinetics and mechanism of caffeine infu-
19 Asante M and Thaler H, Untersuchungen an Kaffee und Kaffee-Ersatz. sion from coffee: the temperature variation of the hindrance factor.
XVII Verhalten der Polysaccharid-Komplexe von Robusta-Kaffee J Sci Food Agric 74:416–420 (1997).
beim Rösten. Zeitschrift für Lebensmittel-Untersuchung und For- 34 Wang X and Lim L-T, A kinetics and modeling study of coffee roast-
schung 159:93–96 (1975). ing under isothermal conditions. Food Bioproc Tech 7:621–632
20 Sachslehner A, Foidl G, Foidl N, Gübitz G and Haltrich D, Hydrolysis of (2014).
isolated coffee mannan and coffee extract by mannanases of Sclero- 35 Ellero M and Navarini L, Mesoscopic modelling and simulation of
tium rolfsii. J Biotechnol 80:127–134 (2000). espresso coffee extraction. J Food Eng 263:181–194 (2019).
21 Grignon J and Scallan AM, Effect of pH and neutral salts upon the swell- 36 Ratnayake WMN, Hollywood R, O'Grady E and Stavric B, Lipid content
ing of cellulose gels. J Appl Polym Sci 25:2829–2843 (1980). and composition of coffee brews prepared by different methods.
22 Carlsson G, Kolseth P and Lindström T, Polyelectrolyte swelling behav- Food Chem Toxicol 31:263–269 (1993).
ior of chlorite delignified spruce wood fibers. Wood Sci Technol 17: 37 Navarini L, Ferrari M, Liverani FS, Liggieri L and Ravera F, Dynamic ten-
69–73 (1983). siometric characterization of espresso coffee beverage. Food Hydro-
23 Eriksson I, Haglind I, Lidbrandt O and Sahnén L, Fiber swelling favoured coll 18:387–393 (2004).
by lignin softening. Wood Sci Technol 25:135–144 (1991). 38 Hwang K-J and Lin I-L, Effects of mixing ratio of binary fine particles on
24 Luukko K and Maloney TC, Swelling of mechanical pulp fines. Cellulose the packing density and filtration characteristics. KONA Powder Part J
6:123–135 (1999). 33:296–303 (2016).

3969

J Sci Food Agric 2020; 100: 3960–3970 © 2020 The Authors. wileyonlinelibrary.com/jsfa
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
 www.soci.org VB Hargarten, M Kuhn, H Briesen

APPENDIX
STATISTICAL ANALYSIS RESULTS
In this appendix, the results from statistical tests are presented
supporting our statements in the Results section. The P-values of
the two-sample t-tests, Mann–Whitney U-tests, and ANOVA are
listed as well as the slopes from linear regressions.

Table A1 Statistical comparison of volume density for dry versus wet measurements and 25 versus 90 °C

P-Value P-Value
a
Compared parameters Range of size classes (μm) Difference of means t-Test Range of size classes (μm) Mann–Whitney

Dry versus wet 0.5–31 <0 < 0.01 0.5–22 < 0.01
31–60 >0 > 0.01b 22–75 > 0.01
60–250 <0 < 0.01c 75–250 < 0.01
250–730 >0 < 0.01 250–730 < 0.01
730–1750 ≤0 > 0.01d 730–1750 > 0.01
25 °C versus 90 °C 0.5–50 <0 < 0.01 0.5–50 < 0.01
50–90 < 0 or > 0 > 0.01 90–510 > 0.01
90–510 >0 < 0.01 90–510 < 0.01
510–1750 ≤ 0 or > 0 > 0.01 510–1750 > 0.01
a
Difference is calculated as mean of first parameter minus mean of second parameter, i.e. dry minus wet and 90–25 °C.
b
Except for 31–37 μm: P < 0.01.
c
Except for 210–250 μm: P > 0.01.
d
Except for 730–870 μm: P < 0.01.

Table A2 Statistical results for the correlation analysis of relative Table A4 Statistical results for the correlation analysis of aspect
diameter increase and initial particle diameter ratio and time for single particles

Slope × 10−3 (% μm−1) P-Value 80 °C 25 °C

Medium roast −2.95 0.291 Slope Slope

Light roast −2.54 0.498 d0 ×10−5 P- d0 ×10−5 P-
(μm) (s−1) Value (μm) (s−1) Value

Medium 608 −2.11 0.100 585 −1.63 0.095

roast
Table A3 Statistical comparison of total swelling for different tem- 519 1.62 0.171 543 0.660 0.559
peratures and roasts based on microscopy measurements 675 −1.24 0.250 914 2.84 4.35 × 10−6
Light 652 1.26 0.088 472 2.51 0.025
Compared Identical roast
parameters parameters P-Value 626 −0.723 0.392 589 2.78 0.013
t-Test Mann–Whitney 462 3.37 0.008 520 2.97 0.051
25 °C Medium roast 0.372 0.400 665 −0.292 0.788 568 −0.291 0.706
versus Light roast 0.065 0.017 653 0.368 0.206 592 −0.344 0.667
80 °C 547 −0.621 0.075
Medium roast versus 80 °C 0.517 0.905
light roast 25 °C 0.173 0.143
3970

wileyonlinelibrary.com/jsfa © 2020 The Authors. J Sci Food Agric 2020; 100: 3960–3970
Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.