The effect of bean origin and temperature on grinding roasted coffee

Abstract

Coffee is prepared by the extraction of a complex array of organic molecules from the roasted bean, which has been ground into fine particulates. The extraction depends on temperature, water chemistry and also the accessible surface area of the coffee. Here we investigate whether variations in the production processes of single origin coffee beans affects the particle size distribution upon grinding. We find that the particle size distribution is independent of the bean origin and processing method. Furthermore, we elucidate the influence of bean temperature on particle size distribution, concluding that grinding cold results in a narrower particle size distribution, and reduced mean particle size. We anticipate these results will influence the production of coffee industrially, as well as contribute to how we store and use coffee daily.

Full text

1
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
www.nature.com/scientificreports
The eect of bean origin and
temperature on grinding roasted
coee
Erol Uman
1
, Maxwell Colonna-Dashwood
2
, Lesley Colonna-Dashwood
2
, Matthew Perger
3
,
Christian Klatt
4
, Stephen Leighton
5
, Brian Miller
1
, Keith T. Butler
6
, Brent C. Melot
7
,
Rory W. Speirs
8
& Christopher H. Hendon
6,9
Coee is prepared by the extraction of a complex array of organic molecules from the roasted bean,
which has been ground into ne particulates. The extraction depends on temperature, water chemistry
and also the accessible surface area of the coee. Here we investigate whether variations in the
production processes of single origin coee beans aects the particle size distribution upon grinding.
We nd that the particle size distribution is independent of the bean origin and processing method.
Furthermore, we elucidate the inuence of bean temperature on particle size distribution, concluding
that grinding cold results in a narrower particle size distribution, and reduced mean particle size. We
anticipate these results will inuence the production of coee industrially, as well as contribute to how
we store and use coee daily.
Second only to oil, coee is the most valuable legally traded commodity. ere are two biologically dissimilar
species of coee grown for consumption; Coea canephora (robusta) and Coea arabica (arabica)
1
. Whilst robusta
is both less chemically complex and less avoursome than arabica, it benets from being feasibly grown at low
altitude and is pest resistant. However, over 60% of the global coee consumption is of arabica. In 2014, Brazil
and Colombia combined to produce over 3.5 million tonnes of green arabica
2
, with Ethiopia and other African
and Central American producers also making signicant contributions. Including countries like Vietnam which
almost exclusively produces robusta, global coee production amounts to 8.5 million tonnes annually.
With the exception of unusual green coee medicinal and dietary preparations, coee is not typically con-
sumed as a solid but rather an extract from the roasted seed
3–9
. Coee beans are imported, roasted, ground
and then brewed (including instant coee) in coee shops and homes. In such a valuable industry, the quality
and yield of the product is paramount. However, there are many variables that inuence the avour, yield and
overall enjoyment of this mass consumed beverage
10
. e challenges associated with ensuring coee quality can
be divided into two categories i) variables associated with the country of origin and ii) variables associated with
consumption.
Besides typical botanical inuences including climate and altitude, there are two general considerations that
aect the coee at the origin: the variety of coee (e.g. Typica, Pacamara, Geisha)
11
and the processing method
(i.e. washed, pulped and natural). e variety denes chemical characteristics of the bean, and also the conditions
in which it may be grown. Ideally, the fruit of the coee bean should not ripen more rapidly than the ovum devel-
ops, otherwise the seed is lacking chemical complexity. Conversely, the fruit should be able to ripen in variable
climate conditions thereby permitting the formation of the seed. Genetic variety hybrids are now ubiquitous and
oen feature the best of both of the parent varieties
12,13
.
Irrespective of the variety, all coee is processed in one of three general methods. e washing (or wet) process
is the most common, and uses water to remove the skin and fruit of the cherry, leaving only the seeds to dry in the
1
Meritics Ltd., 1 Kensworth Gate, Dunstable, LU6 3HS, United Kingdom.
2
Colonna and Smalls, 6 Chapel Row,
Bath, BA1 1HN, United Kingdom.
3
St Ali/Sensory Lab, 12-18 Yarra Pl, South Melbourne, Victoria, 3205, Australia.
4
Mahlkönig GmbH & Co.KG, Tilsiter Str. 142, 22047 Hamburg, Germany.
5
Has Bean Coee Ltd., Unit 16, Ladford
Covert, Staord, ST18 9QL, United Kingdom.
6
Department of Chemistry, University of Bath, BA2 7AY, Bath, United
Kingdom.
7
Department of Chemistry, University of Southern California, Los Angeles, CA 90089, United States of
America.
8
School of Physics, The University of Melbourne, Victoria, 3010, Australia.
9
Department of Chemistry,
Massachusetts Institute of Technology, Cambridge, MA 02139, United States of America. Correspondence and
requests for materials should be addressed to C.H.H. (email: hendon@mit.edu)
Received: 11 December 2015
Accepted: 16 March 2016
Published: 18 April 2016
OPEN

www.nature.com/scientificreports/
2
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
sun. e pulped (pulped natural) processing method removes the skin from the cherry, but does not fully remove
the mucilage. is then forms a sun-hardened sugar-rich shell around the parchment (the thin protective layer
for the seed). e natural process is simply the sun-drying of the coee cherries with both seed and fruit intact.
Whilst the processing method used has a profound impact on avour, the chemical mechanisms which dic-
tate these dierences are not well-understood. Regardless of the cherry processing method, aer drying the
beans are hulled, which exposes the bean by removing all the dry parchment, mucilage, or skin. e green coee
beans are then transported to roasteries, where the roaster develops a roast prole with the aim of producing the
most avoursome cup to their palate. e roast prole is a two variable problem of temperature and time, but
due to limitations of roasting equipment and the inhomogeneity of heat transfer into green coee
14
, the devel-
opment of a roast prole is more artistic than scientic, although there is certainly room for improvement in
this area.
e roast prole presented in Fig.1 shows the measured roaster temperature as the roasting progresses for the
particular Tanzanian coee listed in Table1. e chemical constituents of roasted coee depend on the tempera-
tures of green coee molecular decomposition. e generation and concentration control of these compounds is
achieved through ne tuning of the roast prole
15–17
. Whilst most compounds in roasted coee are likely Maillard
products (an example of which is not shown in Fig.1)
18
, we present various pathways that permit the formation
of acids, phenolic compounds, and also the cleavage of cellulose into sugar-related products like levoglucosan.
e le-most process in Fig.1 shows an example of decomposition of a chlorogenic acid (a group of molecules
contributing to 66% of the acidity in green coee) through low temperature hydrolysis, in which the formation of
products depend on the water content within the seed
19,20
.
Undoubtedly the extent and quality of extraction is dictated by the accessibility of the organic molecules con-
tained within roasted coee. Many factors inuence the total amount, and relative proportions of the dierent
organic molecules extracted, including temperature of brew, water chemistry and water-to-coee ratio
21–24
. Here,
however, we are specically concerned with physical method of increasing accessible surface area; i.e. the eect
of the grinder.
Figure 1. e roast prole for the Tanzanian Burka (Has Bean). In this case, 10 kg of the Burka coee
was roasted in a 12 kg Probat Roaster. e temperature was monitored with a probe in the headspace of the
oven, and hence the hot air rapidly cools due to thermal energy transfer to the green coee. e temperature
trajectory throughout the roasting process determines the decomposition of organic materials in coee. ree
illustrative decomposition reactions are shown that are representative processes throughout the heating process.
At lower temperature a chlorogenic acid (le) may decompose through either hydrolysis or pyrolysis into quinic
acid, acetic acid and the phenolic compound 3,4-dihydroxybenzyl alcohol
40
, or quinic acid, carbon dioxide and
3,4-dihydroxystyrene
41,42
. Oxalic acid (centre) may decarboxylate to either CO
2
or in the case of incomplete
combustion CO
2
and formic acid
19
. At higher temperatures cellulose can undergo hydrolysis to smaller sugar
derivatives including glucose and levoclucosan
43–45
. Both the temperature and time determine the chemical
composition of the roasted coee: In this case, the coee was removed from the oven aer 9 m 54 s as this time
was determined to result in a soluble, sweet and favourably acidic product.
Farm/Estate Origin Varie ty Roaster Roast Agtron colour
Las Ilusiones (W) Guatemala Caturra and Bourbon Round Hill Espresso 62
Santa Petrona (W) El Salvador Pacamara Has Bean Espresso 59
Burka (W) Tanzania Red Bourbon Has Bean Espresso 59
Sasaba (N) Ethiopia Mixed Heirloom James Gourmet Filter 68
Table 1. Details on the four coees (Coea arabica) that were ground in this experiment: two African and
two South American. (W) and (N) indicate washed and natural processing methods, respectively. ‘Roast’ is used
as an indication for whether the coee was roasted for lter (lighter) or espresso (darker) style coee, and can be
quantied by the ‘Agtron colour’ as determined by the Agtron spectrophotometeric measurement
46
. All coees
examined here would be considered light/medium roasted relative to typical commodity grade coee.

www.nature.com/scientificreports/
3
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
Whilst routine in the pharmaceutical industry, it is challenging to both design and execute a grind to a homo-
geneous particle size in a coee shop. is, however, is of critical importance in coee brewing because variable
accessible surface area causes the small particles to extract more rapidly relative to larger ones. As a result, brew-
ing coee is challenging with variable particle size, especially in espresso-style pressurised brews, where packing
eects become important
25,26
. Given the importance of particle size, we assess if bean origin, cherry processing
method, and roast prole have any signicant eect on the particle size distribution of the ground coee.
Additionally, it was suspected that the temperature of the beans could also inuence the bean fracturing
dynamics, and therefore the nal size distribution. Whilst ideally the beans and burrs would both be brought to
the desired temperature, controlled active heating or cooling of the burrs is not presently feasible. To investigate
the temperature eects we pursued the controlled cooling of the coee itself. Given that many people store coee
in the refrigerator or freezer (if devoid of water vapour this is a chemically reasonable method of storage), we
examine if varying bean temperature results in an observable modulation of grind distribution.
Methods
For this study, it was assumed that the most important property of the ground coee which can vary in the grind-
ing process is the distribution of particle sizes. Whilst it is possible that particle shape may have an eect on the
nal extracted brew, it is dicult to see how this can be reliably controlled on the micrometer scale, and it is likely
that most ground coee has a similar spread of particle shapes.
e rst set of experiments explored if the origin, type, or processing method of the bean had any eect on the
particle size distribution, when ground under identical conditions. e second set of measurements explored if
bean temperature at the time of grinding had any eect on produced particle size distribution.
To probe these eects we employed laser diraction particle size analysis of roasted coees ground on a
Mahlkönig EK 43 coee grinder.
Laser diraction particle size analysis. e laser diraction particle size analysis was performed on the
multiple wavelength Beckman Coulter LS13 320 MW. e instrument has a built in dark eld reticule which is
used to ensure correct optical alignment. An alignment check was carried out prior to every run to ensure the
optimum accuracy of the particle size distribution. Particular care was taken to ensure correct optical alignment
because ground coee contains particles sizes spanning 3 orders of magnitude, including components larger
than 100 m, which can be challenging to measure with diraction based techniques since they rely on distance
measurements in reciprocal space.
Grinding. e Mahlkönig EK 43 grinder, shown in Fig.2, was selected for this study because it is designed to
have minimum retention time between placing the coee in the hopper and subsequent grinding. Like all grind-
ers, the EK 43 burrs are replaceable and are susceptible to becoming misaligned (where the two grinding discs
are not perfectly parallel). We had access to three separate EK 43 s on the day of testing; two which were tted
with so-called coee burrs and one with Turkish burrs (Fig.2). Burr alignment can initially be assessed audibly
by closing the burr aperture with the grinder turned on, causing them to ‘chirp’. e pitch of the chirp provides
insight into the alignment, with deeper chirps indicating more contact between the burrs and therefore better
alignment. Assessing the smoothness and spread of ground particle distribution can also give information on
burr alignment, though it is dicult and slow to reliably adjust alignment based on this information. We have
Figure 2. e EK 43 grinder, (a) consists of two burrs; one stationary and one mobile. e hopper-to-shoot
path is linear resulting in minimal retention of ground coee in the burrs and shoot. ere are two types of
burrs: Turkish, (b) and Coee, (c). e primary dierences between the two burrs are emphasised in blue in
(b,c), respectively. e at triangular ends are intended to polish the particulates. For this study we employed
the Turkish burr set. Photographs taken by Spencer Webb.

www.nature.com/scientificreports/
4
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
provided one example of particle size distributions from a burr misalignment in Figure S1. Ultimately, we elected
to use the grinder that was producing subjectively marketable espresso shots that day, as determined by a resident
qualied Q-grader and shop owner (Maxwell Colonna-Dashwood)
27
. Experiments herein were performed with
the Mahlkönig EK 43 grinder spinning at 1480 rpm and grinding with Turkish burrs.
Coee origin and processing. To determine if bean origin has an eect on particle size distribution aer
grinding, beans were tested from four countries: Guatemala, El Salvador, Tanzania, and Ethiopia. e beans had
been roasted by roasteries listed in Table1, between seven and sixteen days prior to the grinding test, and so had
sucient time for CO
2
degassing, but were still considered ‘freshly roasted’. Further details of the four coees
considered in this study are presented in Table1. All beans were allowed to equilibriate to room temperature (at
the time, 20 °C and 79% relative humidity), densities of the roasted coee beans were not measured. e grinder
burr aperture was kept constant for all coees throughout the experiment, xed at 2.7 (arbitrary units) on the
stock EK 43 dial. For each measurement, 20 grams of coee was ground, and the grinder was allowed to cool for
10 minutes aer each grind (returning to room temperature).
Coee temperature. For temperature studies, we selected the Guatemalan coee because this particular
Guatemalan crop is representative of contemporary speciality grade coee (i.e. it has a favourable balance of acid-
ity, oral complexity and overall taste). e four temperatures were achieved using the following method: 20 g of
whole roasted coee beans were placed into a paper cup, covered, and placed into either liquid nitrogen, a tub of
dry ice, the freezer and on the counter top. No visible condensation of atmospheric water was observed on any
of the samples cooled below 0 °C. e beans were equilibriated at each temperature for 2 hours prior to grinding.
e grinder was switched on 5 seconds before grinding and the beans were taken directly from their climates
and fed into the hopper. e EK 43 is rated to grind 1200–1500 g/min, suggesting each 20 g dose of coee was
exposed to ambient conditions for no longer than 1 second. To prevent condensation of atmospheric water onto
the surface of the ground coee, the ground particulates were immediately placed into sample vials for laser dif-
fraction particle size analysis. Absorption of atmospheric water proved not to be a problem, as duplicate samples
which were exposed to atmospheric moisture as they equilibrated to room temperature, showed no dierence to
those that where sealed immediately upon grinding.
Each data set was obtained in triplicate, and each temperature was obtained in duplicate thereby generating
6 data sets per temperature. ANOVA was employed for determination of similarities in particle number distri-
butions with consideration of the bean origin, processing method, roast and roaster included. e output of this
statistical analysis is included in the supporting information.
Do Dierences in the Green Bean Aect the Final Grind?
e physical structure of roasted coee beans is a complex composite of materials, containing high molecular
weight brous molecules interspersed with amorphous and partially crystalline domains of a vast array of smaller
organics. e extremely complex structure of both the roasted beans and grinding apparatus makes accurate
rst principles modeling a daunting prospect, and so fracturing is best studied experimentally (in line with pre-
vious studies of grinding other amorphous materials)
28–32
. at said, it could well be expected that the specic
mix of chemicals that give dierent coees their distinctive avour may change the way in which the bean is
fragmentised.
To investigate this, we elected to sample four speciality grade coees. e selection spans the variables of ori-
gin, variety, processing method and roast prole, and is a representative cross section of contemporary speciality
coee. e four coees described in Table1 were ground at ambient conditions using the stipulated methods.
Here we are concerned with the deviations in grind prole as a function of coee origin, although before
embarking on these experiments it was unclear what the grind prole looked like. e EK 43 produces particles
ranging from 0.1 m to 1000 m, and whilst we have elected to present most of the data on a logarithmic scale,
the linear scale is shown for the Tanzanian coee in the upper panel of Fig.3. All grind proles appear as a
skewed-Gaussian shape. In this case, we present the particle number distribution in the shaded blue region, and
the integral in grey. We can arbitrarily dene the ne particulate cuto, graphically represented as a purple dashed
line = n where:
∫
.=099cumulativenumber
(1)
n
0
Here, n is a diameter in m. From the upper panel of Fig.3, the Tanzanian n = 70 m (mode = 13.0 m, where the
mode is the most frequent size occurrence). Given the skewed nature of the distribution, the mode is helpful in
assigning key features of the distribution. However, it is not only the number of particles that contributes to the
extraction of coee, but also the available surface area obtained from these particles.
e grind proles for the four coees examined here are shown in the middle and lower panels of Fig.3. ey
are presented on a logarithmic scale to accommodate the surface area contribution from the large particles. e
surface area is estimated using a spherical approximation for the particles
33
, and is shown by the dotted line. Here,
the data appears distinctly bimodal because the ne particulates contribute to the majority of the accessible sur-
face area (modes ii and v), but large particulates (one/two orders of magnitude larger in diameter, iii and vi) are
also present. ese have an inuence even at low concentrations.
ere are minor dierences in the grind proles: e proles shown in black and purple share similar par-
ticle number modes (i), and have a ne particulate cuto of 76.4 ± 3.5 m. e proles shown in red and blue
produced a slightly ner particle distribution with a number mode (iv) 1.3 ± 0.7 m) more ne than the black/
purple coees, and a ne particulates cuto of 69.6 ± 3.1 m. In summary, the coees appear to produce a very
similar grind distribution irrespective of the variables associated with bean production. Full ANOVA details are

www.nature.com/scientificreports/
5
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
presented in Table S1. It should be noted that all of the beans considered here are roasted relatively ‘light’ com-
pared to typical consumer grade coee (although on the ‘Agtron Gourmet Scale’, these coees all are catagorised
as light-medium roast). We can only speculate how heavily decomposed beans (e.g. ‘dark’ or French roast) may
deviate from these results; further experimentation is required to elucidate that eect.
For espresso, the coee grinds can be thought of as a granular material, where the increase in pressure during
tamping jams the materials
34–36
. e variability in particle size plays a signicant role in the accessible surface
area, but also in the vacuous space in which the water may ow through. From the work of Herman
37
, it is appar-
ent that large particles install signicant order of neighbouring small particles, which increases local density and
therefore can result in inhomogeneous water ow through the espresso puck. However, given the subjectivity of
coee avour and the preferences of practitioners working in the industry, it is not clear if there is an ideal particle
size distribution: We only hope to shed light on the surprising consistencies between coees.
Do Dierences in the Roasted Bean Grind Temperature Aect the Final Grind?
Temperature changes in amorphous materials can lead to well dened glass transitions, where the material
changes from rubbery and exible to being hard and brittle
38
. Some solids can also undergo shattering transi-
tions, where there is an increased fragmentation rate as particle size decreases, resulting in production of greater
numbers of ne particles
39
. is property is instigated by both temperature and crack velocity. It is understood
that crystalline materials progress towards this shatter transition point with decreased temperature, because the
strain on the lattice becomes proportionally larger with decreased lattice kinetics. However, roasted coee is a
complex material and glass or shattering transition points are unlikely to be constant across macroscopic regions
of the bean, if present at all. erefore, while it is reasonable to expect that a change in temperature will aect the
grinding result, describing how and why this occurred is problematic. Experiment provides the simplest and most
reliable route to assessing how temperature inuences ground coee particle size.
e lower the original bean temperature, the colder the produced particles will be at every stage of grinding.
However colder bean fragments will absorb heat from their surroundings more quickly due to the larger tem-
perature gradient, eectively reducing the indicated temperature dierence between the samples. erefore, the
observed change in grind prole should be considered a lower limit on the eects of grinding at reduced tempera-
tures. Given the inhomogeneous nature of the beans, it is likely that cooling the burrs (and hence further reducing
the temperature of the particles as they are ground) would smoothly continue the trend observed in Fig.4.
Some fraction of particles are produced in their nal size from the initial fracturing of the whole bean (or large
portion thereof), and so are truly produced at the stated temperature. However, experiments using a single impact
event (i.e. hitting a cold bean with a mallet), show that only a small amount of small particles are produced on
initial bean fracturing, so most particles do have some time for thermalisation before further fracturing occurs.
0
1
0
1
110 100 1000
Particle Diameter (µm)
Relative particle number frequency
Surface Area
Counts
0
1
05
01
00
Particle Diameter (µm)
70
number
iiiiii
iv vvi
1%
0
1
0
1
0
1
Relative surface area contribution
Figure 3. Upper panel: e particle size distribution as a function of number (cumulative) of physical
particles (shown in blue) and the integral of this data (shown in grey), yields 99% of the particles with a
diameter of 70 μm or less. e ne particular cuto is depicted as a purple dashed line. Middle and lower panel:
e grind proles of the four coees examined here. e cumulative number and surface area contribution are
shown in solid and dashed lines, respectively. e Tanzanian, Ethoipian, El Salvadorian and Guatemalan proles
are shown in black, purple, red and blue, respectively. Data modes i-vi are included for visual aid: i - 14.3, ii -
27.4, iii - 282.1, iv - 13.0, v - 27.4 and vi - 256.9 m.

www.nature.com/scientificreports/
6
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
Even with some particle thermalisation due to room temperature burrs, the initial bean temperature has a
signicant eect on the modal particle size distribution (Fig.4a) reducing the mode by 31% as the beans are
cooled from room temperature to − 200 °C, as shown in Fig.4b. Additionally, the distribution generally becomes
narrower as the beans are cooled (Fig.4c) with the biggest change occurring between room temperature and
− 19 °C beans. e room temperature grind prole is also distinctly less Gaussian-like, with the development of
a hip at approximately 9.5 m. is detail could indicate that some components of the bean undergo a shattering
transition between 20 °C and − 19 °C, and studies are ongoing into the origin of this feature.
To probe the reversibility of this transition, we performed the same room temperature experiments with coee
beans that had been cooled to liquid nitrogen temperatures and then allowed to reheat to room temperature. It
appears that if there is a transition, it is reversible as there were no notable dierences between the two samples.
is is not surprising given the very low water concentration in roasted coee: e thermal contraction and
re-expansion of coee did not play a signicant role in the grind prole obtained from either test set.
Applications and Concluding Remarks
In busy coee shops, it is common practice to reduce burr grinding aperture as the day progresses in order to pro-
duce a consistent cup of coee. From work presented here, we propose that this phenomenon is a direct product
of the grinding burrs (and potentially beans sitting in the hopper in grinders other than the EK 43) becoming
increasingly warm as the grinder is used. e particle warming at the interface between the coee bean and warm
burr - which can certainly be much higher in temperature than explored in this study - shis both the mode and
spread of the particle size distribution. us, as the grinder gets warm a ner grind setting may be required to
obtain the same eective surface area as the same coee ground on cooler burrs. However, we also observe a dif-
ference in the shape of the distribution with temperature, which indicates that simply grinding ner with warm
burrs will not produce the same result as grinding coarsely with cold burrs. e impact on taste and preference is
not the focus of this study, but is certainly an interesting avenue to explore in the future.
e distinct lack of dependence on origin and processing method is comforting for coee shops that serve cof-
fees from multiple origins, and also for roasters who develop and market blends (mixtures of origins). One grand
challenge with blended coee is to produce a product where each desired component is equally soluble, such that
the cup of coee tastes appropriately extracted. Consider the traditional blend of Brazilian and Ethiopian coees:
e two are combined to obtain the body and nuttiness from the Brazilian, and the fruit and complexity from the
Ethiopian. But such results are only obtained if both beans reach terminal extraction at similar rates. Here, we
1
0
Particle Diameter (µm)
-79 ºC -19 ºC
-196 ºC
20 ºC
110100 1000
Relative particle
number frequency
Surface Area
Number
11
13
15
-250 -150 -5
05
0
Bean Temperature (ºC)
Number
mode (µm)
9
Skewness
0.8
1.2
17
15
14
16
1
0
Relative surface
area contribution
Number
mean (µm)
a
d
b
c
Figure 4. e temperature dependence on the grind prole of the El Salvadorian coee, (a). e temperatures
were achieved by grinding liquid nitrogen, dry ice, freezer and room temperature coee, respectively. e
ne particulate cuto is schematically shown, with exact values corresponding to; − 196 °C = 61 ± 3 m,
− 79 °C = 63 ± 3 m, − 19 °C = 73 ± 3 m and 20 °C = 70 ± 3 m. e mode of the number distribution, (b)
shows a clear and non-linear trend of increasing mode with increasing temperature. e distribution skewness
is inversely proportional to temperature. From a avour perspective this is a favourable feature because the
surface area to volume ratio becomes increasingly signicant for the smaller particles. e mean particle size,
(d) is discontinuous with temperature likely indicating a transition between freezer and room temperature.

www.nature.com/scientificreports/
7
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
have minimised one variable by showing that at least the accessible surface area is kept constant whilst grinding,
thereby placing the chemical problems associated with blending solely on the roast prole.
From a physical chemistry perspective, the temperature dependence presents many interesting questions.
Given the minimal dierence between liquid nitrogen and dry ice temperatures and the reversibility of the cool-
ing, we question whether it is possible in the future to cryogenically store roasted coee at these temperatures.
Indeed, water content in the roasted bean is of paramount importance at these temperatures, as water expansion
may lead to be fracturing. Also, prolonged exposure to water can result in the solvation of avoursome molecules,
thereby decreasing the lifetime of the frozen product. But if these variables were managed, there are a host of
subsequent implications for the storage and relative quality assessment allowing for access to direct year-to-year
comparison of crop quality. From a consumption perspective, cooling of coee beans signicantly decreases the
rate of mass loss through volatile sublimation/evaporation. us, coee that is ground and brewed cold could
potentially demonstrate increased aroma and or avour in the eventual brewed cup.
From an industrial perspective, the yield of extraction is paramount. Grinding colder coee beans produces
a more uniform particle distribution, with a decreased particle size. While the decreased particle size will tend
to speed up extraction due to the larger surface area, the increased uniformity should minimise the amount of
wasted bean, which is discarded without being extracted to completion. Whilst active cooling of either the coee
beans or burrs is energy consuming, the benet of cold coee grinding may oset this cost with more ecient
extraction from the smaller particles.
References
1. Denoeud, F. et al. e Coee Genome Provides Insight into the Convergent Evolution of Caeine Biosynthesis. Science 345, 1181
(2014).
2. Http://www.ico.org/, International Coee Organisation, Date Accessed: 26/02/2016 (2016).
3. Farah, A., Monteiro, M., Donangelo, C. M. & Lafay, S. Chlorogenic acids from green coee extract are highly bioavailable in humans.
J. Nutri. 138, 2309 (2008).
4. Stelmach, E., Pohl, P. & Szymczycha-Madeja, A. e content of ca, cu, fe, mg and mn and antioxidant activity of green coee brews.
Food Chem. 182, 302 (2015).
5. Daglia, M., Papetti, A., Gregotti, C., Bertè, F. & Gazzani, G. In vitro antioxidant and ex vivo protective activities of green and roasted
coee. J. Agric. Food Chem. 48, 1449 (2000).
6. Cliord, M. N. in Chemical and physical aspects of green coee and coee products, 305 (Springer, 1985).
7. Charurin, P., Ames, J. M. & del Castillo, M. D. Antioxidant activity of coee model systems. J. Agric. Food. Chem. 50, 3751 (2002).
8. Shimoda, H., Sei, E. & Aitani, M. Inhibitory eect of green coee bean extract on fat accumulation and body weight gain in mice.
BMC Comp. Alt. Med. 6, 9 (2006).
9. Watan abe, T. et al. e blood pressure-lowering eect and safety of chlorogenic acid from green coee bean extract in essential
hypertension. Clin. Exp. Hyperten. 28, 439 (2006).
10. van Doorn, G., Colonna-Dashwood, M., Hudd-Baillie, . & Spence, C. Latté art inuences both the expected and rated value of
mil-based coee drins. J. Sens. Stud. 30, 305 (2015).
11. Https://www.stumptowncoee.com/varieties/, Stumptown Coee oasters, Date Accessed: 26/02/2016 (2016).
12. Carneiro, M. F. Coee biotechnology and its application in genetic transformation. Euphytica 96, 167 (1997).
13. Sera, T. Coee genetic breeding at iapar. Crop Breed. Appl. Biotechnol. 1, 179 (2001).
14. Fabbri, A., Cevoli, C., Alessandrini, L. & omani, S. Numerical modeling of heat and mass transfer during coee roasting process.
J. Food Eng. 105, 264 (2011).
15. Johnston, W. . & Frey, C. N. e volatile constituents of roasted coee. J. Am. Chem. Soc. 60, 1624 (1938).
16. Gianturco, M. A., Giammarino, A. S. & Friedel, P. Volatile Constituents of Coee-V. Nature 210, 1358 (1966).
17. Franca, A. S., Oliveira, L. S., Mendona, J. C. F. & Silva, X. A. Physical and chemical attributes of defective crude and roasted coee
beans. Food Chem. 90, 89 (2005).
18. Somoza, V. & Fogliano, V. 100 years of the maillard reaction: why our food turns brown. J. Agric. Food Chem. 61, 10197 (2013).
19. Flament, I. & Bessière-omas, Y. in Coee avor chemistry. (John Wiley & Sons, 2002).
20. Moon, J. ., Hyui Yoo, S. U. N. & Shibamoto, T. ole of roasting conditions in the level of chlorogenic acid content in coee beans:
Correlation with coee acidity. J. Agric. Food Chem. 57, 5365 (2009).
21. Hendon, C. H., Colonna-Dashwood, L. & Colonna-Dashwood, M. e role of dissolved cations in coee extraction. J. Agric. Food
Chem. 62, 4947 (2014).
22. Shimizu, S. Caeine dimerization: eects of sugar, salts, and water structure. Food Funct. 6, 3228 (2015).
23. Colonna-Dashwood, M & Hendon, C. H. in Water For Coee. (Self-Published, 2015).
24. Navarini, L. & ivetti, D. Water quality for Espresso coee. Food Chem. 122, 424 (2010).
25. Möbius, M. E., Lauderdale, B. E., Nagel, S. . & Jaeger, H. M. Brazil-nut eect: Size separation of granular particles. Nature 414, 270
(2001).
26. Hong, D. C., Quinn, P. V. & Luding, S. everse brazil nut problem: competition between percolation and condensation. Phys. ev.
Lett. 86, 3423 (2001).
27. Http://www.coeeinstitute.org/, Coee Quality Institute, Q-Coee System, Date Accessed: 26/02/2016 (2016).
28. itajima, ., Cai, G. Q., urnagai, N., Tanaa, Y. & Zheng, H. W. Study on Mechanism of Ceramics Grinding. CIP Ann. - Manuf.
Technol. 41, 367 (1992).
29. Hamdi, H., Zahouani, H. & Bergheau, J. M. esidual stresses computation in a grinding process. J. Mater. Process. Technol. 147, 277
(2004).
30. Inasai, I. Grinding of Hard and Brittle Materials. CIP Ann. - Manuf. Technol. 36, 463 (1987).
31. onig, W. A Numerical Method to Describe the inematics of Grinding. CIP Ann. - Manuf. Technol. 31, 201 (1982).
32. Li, X. & Bhushan, B. Measurement of fracture toughness of ultra-thin amorphous carbon lms. in Solid Films 315, 214 (1998).
33. Corrochano, B. ., Melrose, J. ., Bentley, A. C., Fryer, P. J. & Baalis, S. A new methodology to estimate the steady-state permeability
of roast and ground coee in paced beds. J. Food Chem. 150, 106 (2015).
34. Iwashita, . & Oda, M. in Mechanics of granular materials: an introduction. (CC Press Inc., 1999).
35. Li, X. & Li, X.-S. Micro-Macro Quantication of the Internal Structure of Granular Materials. J. Eng. Mech. 135, 641 (2009).
36. Corwin, E. I., Jaeger, H. M. & Nagel, S. . Structural signature of jamming in granular media. Nature 435, 1075 (2005).
37. Herman, A. Shear-jamming in two-dimensional granular materials with power-law grain-size distribution. Entropy 15, 4802 (2013).
38. Lubliner, J. in Plasticity eory, Ch. 2, e Physics of Plasticity Section. (Dover Publications, 2008).
39. Mcgrady, E. D. & Zi, . M. “Shattering” Transition in Fragmentation. Phys. ev. Lett. 58, 892 (1987).
40. Sugumaran, M., Semensi, V., Dali, H. & Nellaiappan, . Oxidation of 3,4-dihydroxybenzyl alcohol: A sclerotizing precursor for
cocroach ootheca. Arch. Insect. Biochem. Physiol. 16, 31 (1991).

www.nature.com/scientificreports/
8
Scientific RepoRts | 6:24483 | DOI: 10.1038/srep24483
41. Fran, O., Zehentbauer, G. & Hofmann, T. Bioresponse-guided decomposition of roast coee beverage and identication of ey
bitter taste compounds. Eur. Food es. Technol. 222, 492 (2006).
42. Lentner, C. & Deatherage, F. E. Organic acids in coee in relation to the degree of roast. J. Food Sci. 24, 483 (1959).
43. Minamisawa, M., Yoshida, S. & Taai, N. Determination of biologically active substances in roasted coees using a diode-array hplc
system. Anal. Sci. 20, 325 (2004).
44. Minowa, T., Fang, Z., Ogi, T. & Várhegyi, G. Decomposition of cellulose and glucose in hot-compressed water under catalyst-free
conditions. J. Chem. Eng. Jap. 31, 131 (1998).
45. aufman, P. B., Csee, L. J., Warber, S., Due, J. A. & Brielmann, H. L. in Natural products from plants. (CC Press Inc., 1998).
46. e Agtron value for roasted cofee is an spectrophotometeric measurement of reected near-I light. Higher values indicate lighter
roasted coee. For reference, an ‘Italian’ or dar-roasted bean typically has an Agtron value of ca. 30.
Acknowledgements
e authors would like to acknowledge Beckman Coulter and Meritics for their generous donation of the LS13
320 MW Laser Diffraction Particle Size Analyzer. Their equipment and expertise enabled this project. We
appreciate the contributions from Mahlkönig mbH, who provided insight into their grinder constructions.
All authors are also humbled by the generosity of Colonna and Small’s and their patient customers who were
deprived of their daily coee whilst these experiments were performed in store. We would like to acknowledge
the continued support and interest in our work from the global specialty coee industry. Finally the authors
are thankful for S. Webb’s photography of the EK 43 burrs, and the lyrical insights from A. omas Murray and
C. Derek Molloy.
Author Contributions
E.U., B.M., M.P., C.K., L.C.-D. and M.C.-D. contributed to the experimental design and execution. S.L. roasted
the Has Bean coee and provided the roast prole. K.T.B., B.C.M. and R.W.S. aided in both experimental design
and data interpretation. C.H.H. conceived the study and wrote the rst dra of the paper. All authors contributed
equally to subsequent revisions of the paper.
Additional Information
Supplementary information accompanies this paper at http://www.nature.com/srep
Competing nancial interests: e authors declare no competing nancial interests.
How to cite this article: Uman, E. et al. e eect of bean origin and temperature on grinding roasted coee.
Sci. Rep. 6, 24483; doi: 10.1038/srep24483 (2016).
is work is licensed under a Creative Commons Attribution 4.0 International License. e images
or other third party material in this article are included in the article’s Creative Commons license,
unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material. To view a copy of this
license, visit http://creativecommons.org/licenses/by/4.0/