Chemical Engineering Science 137 (2015) 216–234

Contents lists available at ScienceDirect

Chemical Engineering Science

journal homepage: www.elsevier.com/locate/ces

Modelling of coffee extraction during brewing using multiscale

methods: An experimentally validated model
K.M. Moroney a,n, W.T. Lee a, S.B.G. O'Brien a, F. Suijver b, J. Marra b
a
MACSI, Department of Mathematics and Statistics, University of Limerick, Ireland
b
Philips Research, Eindhoven, The Netherlands

H I G H L I G H T S G R A P H I C A L A B S T R A C T

We conduct experiments investigating

extraction of solubles from coffee beds.

We develop a model to simulate
extraction of coffee with hot water
( 90 1C).

We model the coffee bed as a satu-
rated, static, doubly porous medium.

Fast and slow release of coffee from
surfaces and kernels of grains are
modelled.

Parameterised model is found to fit the
experimentally measured extraction
curves.

art ic l e i nf o a b s t r a c t

Article history: Accurate and repeatable extraction of solubles from roasted and ground coffee with hot water is vital to
Received 23 January 2015 produce consistently high quality coffee in a variety of brewing techniques. Despite this, there is an
Received in revised form absence in the literature of an experimentally validated model of the physics of coffee extraction. In this
21 May 2015
work, coffee extraction from a coffee bed is modelled using a double porosity model, including the
Accepted 1 June 2015
Available online 12 June 2015
dissolution and transport of coffee. Coffee extraction experiments by hot water at 90 1C were conducted
in two situations: in a well stirred dilute suspension of coffee grains, and in a packed coffee bed.
Keywords: Motivated by experiment, extraction of coffee from the coffee grains is modelled via two mechanisms:
Coffee brewing process an initial rapid extraction from damaged cells on the grain surface, followed by a slower extraction from
Coffee extraction experiments
intact cells in the grain kernel. Using volume averaging techniques, a macroscopic model of coffee
Double porosity model
extraction is developed. This model is parameterised by experimentally measured coffee bed properties.
Static porous medium
Coffee extraction kinetics It is shown that this model can quantitatively reproduce the experimentally measured extraction
Drip filter coffee profiles. The reported model can be easily adapted to describe extraction of coffee in some standard
coffee brewing methods and may be useful to inform the design of future drip filter machines.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Coffee, derived from the seeds (beans) of the coffee plant, is
among the most popular beverages consumed worldwide. Typically,
n
Corresponding author.
E-mail addresses: kevin.moroney@ul.ie (K.M. Moroney),
william.lee@ul.ie (W.T. Lee), stephen.obrien@ul.ie (S.B.G. O'Brien),
freek.suijver@philips.com (F. Suijver), johan.marra@philips.com (J. Marra).
after the beans are roasted and ground, some of their soluble
content is extracted by hot water. The resulting solution of hot
water and coffee solubles is called coffee. Coffee extraction is carried
out on a large number of scales varying from large-scale industrial
extraction to produce instant coffee, right down to one-cup brewing
appliances for domestic use. For the purposes of brewing there are
numerous methods of producing a coffee beverage, which can be
broken into three main categories: decoction methods, infusion
methods and pressure methods. Many of these brewing techniques

http://dx.doi.org/10.1016/j.ces.2015.06.003
0009-2509/& 2015 Elsevier Ltd. All rights reserved.
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234
are described in refs. Petracco (2008) and Pictet et al. (1987). The
intimate contact of water with roasted coffee solids is the cardinal
requirement for producing a coffee beverage (Petracco, 2008). In
line with this, all the coffee brewing methods mentioned rely on
solid–liquid extraction or leaching, which involves the transfer of
solutes from a solid to a fluid. Despite its widespread consumption,
long history and well developed techniques, consistently brewing
high quality coffee remains a difficult task. This difficulty arises from
the dependency of coffee quality on a large number of process
variables. Some of these include brew ratio (dry coffee mass to
water volume used), grind size and distribution, brewing time,
water temperature, agitation, water quality and uniformity of
extraction (Petracco and Technology, 2008; Rao, 2010). According
to Clarke et al. (1987), “The extraction of roast and ground coffee is,
in fact, a highly complex operation, the scientific fundamentals of
which are very difficult to unravel”. This is reflected in the absence
of a satisfactory, experimentally validated mathematical system of
equations to model the extraction process accurately. Such a
description would have obvious benefits in quickly and easily
investigating the influence of various parameters on coffee extrac-
tion and informing the design of the next generation of coffee
brewing equipment. Of course the notion of high quality coffee is a
rather inexact ideal and to some extent a matter of taste perception.
Relating taste to the physical parameters of extracted solubles is in
itself a non-trivial matter, separate to the issue of consistency.
Despite this, certain correlations have been identified between
coffee flavour and extraction yield. The coffee brewing control chart
for example, gives target ranges for brew strength and extraction
yield based on preferences observed in organised taste tests (Rao,
2010). Brew strength is the ratio of mass of dissolved coffee in the
beverage to volume. Extraction yield is the percentage of dry coffee
grind mass that has extracted as solubles into the water.
The chemistry of coffee brewing has received a great deal of
attention in recent times but the physics of the brewing has
received relatively little attention. Very often, as in other food
engineering applications, the importance of the microstructure in
mass-transfer is ignored in extraction models, and solids are
treated as “black boxes” (Aguilera and Stanley, 1999). Some work
has been done modelling the physics of certain brewing systems.
Large scale industrial extractors for the production of instant
coffee have been the subject of detailed investigations. Early work
focused on modelling coffee extraction in large packed columns
called diffusion batteries with a focus on improving the design of
these solid–liquid extractors (Sivetz and Foote, 1963; Spaninks,
1979). Some of these developments are summarised in Clarke et al.
(1987). There has also been some physical modelling of domestic
brewing systems. Experimental investigations have been carried
out into the operation and efficiency of the Moka pot (Navarini et
al., 2009; Gianino, 2007). Fasano et al. have developed general
multiscale models for the extraction of coffee primarily focused on
the espresso coffee machine (Fasano and Talamucci, 2000; Fasano
and Farina, 2010; Fasano et al., 2000; Fasano, 2000). Voilley and
Simatos (1979) conducted a number of extraction experiments on
a well mixed system of coffee grounds and water and investigated
the influence of brewing time, granule size, brew ratio and water
temperature on brew strength. The diffusion equation in a sphere
was also found to be useful to describe the variations in the brew
strength of the coffee during the experiments. There has been very
little investigation into the physics of the drip filter brewing
system. There are a number of aspects in the drip filter brewing
system, where a greater understanding of the physical process
may lead to improved design and increased quality of coffee
produced. Some of these aspects were investigated by a group of
applied mathematicians working on a problem posed by Philips
Research during the ESGI 87 study group with industry in the
University of Limerick. The topics investigated are included in the
217
study group report (Booth et al., 2012). Despite these develop-
ments, there is an absence in the literature of a first principles
model of coffee extraction which is validated by experiment. This
paper aims to address this deficit.
The aim of this study is to formulate a comprehensive experi-
mentally validated model of the physics of coffee extraction. The
model should include the dissolution and transport of coffee
within the coffee bed. It should also take into account the doubly
porous nature of the coffee bed, which consists of pores between
the coffee grains (intergranular), but also smaller pores within the
coffee grains (intragranular). In this paper, flow through a static,
saturated coffee bed, under the influence of a pressure gradient is
modelled using a double porosity model. The parameters in this
macroscopic model are related to the microscopic properties of the
coffee bed by an averaging procedure using Representative Ele-
mentary Volumes. This allows the model to be parameterised from
experimentally measured microscopic quantities. Utilising multi-
scale modelling of extraction from coffee grains we show that we
can quantitatively model extraction from ground coffee in two
situations: in a well stirred dilute suspension of coffee grains and
in a packed coffee bed. In our model, extraction is divided into two
regimes. In the first, a rapid extraction occurs from the surface of
the coffee grains which yields the highest concentrations at any
stage of the brewing process. In the second there is a slow
extraction at lower concentrations from the interior of the grains.
The model can be easily generalised to describe standard coffee
brewing methods such as French press and drip filter coffee. It can
also be extended to include unsaturated flow in the coffee bed.
2. Coffee extraction experiments
A large number of experiments were carried out to investigate
flow and extraction of coffee from coffee beds of various geome-
tries. Two of these experiments will be outlined here and used to
motivate the development of a mathematical model to replicate
their results. The experiments were performed with a number of
different coffee grinds. We focus on two of these coffee grinds. The
first is a relatively fine grind called Jacobs Krönung (JK) standard
drip filter coffee grind. The second is a coarse grind obtained with
a Cimbali burr grinder from Illy coffee beans. The grind used is
obtained by setting the grinder to a coarse setting #20 and will be
referred to as Cimbali #20. The grind size distributions for these
grinds are shown in Fig. 1. It is apparent from the graph that both
distributions are bimodal, having two peaks. A first peak occurs at
a particle size of 25–30 μm while the second peak occurs at a
larger particle size and gradually shifts from left to right on the
graph with the grind coarseness. The first peak accounts for single
cell fragments: the cell size in coffee particles is 25–50 μm. The
second peak accounts for particles comprising intact coffee cells.
The grind size distribution is vitally important in coffee extraction
in that it affects both the fluid flow through the grind and the
grind's extraction kinetics.
2.1. Maximum extractable solubles mass from coffee grind
Extraction from a coffee grain occurs following contact with
water. However not all of the coffee grain mass is soluble.
Experiments conducted show that extractable mass of coffee
grains in water at 90 1C can range from 28% for very fine grinds
to 32% for very coarse grinds. These results were obtained using
fine and coarse grinds from Douwe Egberts (DE) coarse drip filter
coffee. The extraction was carried out in glass beakers by con-
tinuously stirring the grind through the water with a magnetic
stirrer for at least 5 h to ensure maximum extraction. Increasing
the extraction time from 5 h to 10 h did not change the extracted
218 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234
10
8 flow adjusting itself to the pressure. The coffee beverage exiting
Volume fraction
6
4
2 tion, given in milligrams coffee solubles per gram of coffee beverage
0 are shown in Fig. 3. In the case of the JK standard drip filter grind,
0.1 1 10 100 1000
Particle size µm
Fig. 1. Coffee grind size distributions for JK standard drip filter grind (–) and
Cimbali #20 grind (- -). Distributions are expressed in terms of volume fraction
percentages of particles of a given diameter.
40
30
cbrew kg m3
20
10
0
0 100 200 300 400 500
time s
Fig. 2. Coffee solubles concentration profiles for JK standard drip filter grind ð
Þ and
Cimbali #20 grind ð□Þ during batch extraction experiments. In these experiments
60 g of coffee with approximately 4% moisture was mixed with 0.5 l of hot water in
a French press type cylinder.
amount. Reduced extraction from larger coffee particles probably
reflects the phenomenon that some solubles cannot be removed
from closed cells inside the larger particle kernels.
2.2. Coffee extraction kinetics during batch-wise brewing in a fixed
water volume
The extraction kinetics of the two coffee grinds considered here
were investigated by mixing 60 g of coffee with a hot water
volume, V water ¼ 0:5 l, and measuring the concentration cbrew of
extracted species as a function of time. The temperature of the
liquid during extraction is 80–90 1C. This was done by performing
batch-wise extraction in a French-press type cylinder and using a
piston to separate the brew from the grounds, after a certain
extraction time, by pressing the mixture through a sheet of coffee
filter paper on the cylinder bottom. The solubles concentration
cbrew was subsequently determined by measuring the 1Brix with a
pocket refractometer (PAL-3, Atago, Japan). It was found for drip
filter coffee that 1 1Brix corresponds with cbrew ¼ 8:25 g/l. The
latter calibration factor was obtained by evaporating all the water
from the coffee brew and weighing the remaining non-volatile
material. The resulting extraction profiles are shown in Fig. 2.
Extraction profiles for some other coffee grinds are included in
Appendix G.
2.3. Coffee extraction profiles from a cylindrical brewing chamber
A number of coffee extraction experiments were conducted with
both a cylindrical brewing chamber and a conical Melitta filter. We
will focus on the cylindrical brewing chamber here. In this setup,
coffee is placed in a cylindrical flow-through cell and 1 l of water at
90 1C is forced through the coffee bed using a rotary vane pump.
The system can operate in a constant flow mode with the pressure
differential across the bed adjusting itself to the flow. Alternatively
the pressure differential across the coffee bed can be fixed with the
from the flow through cell is collected in a coffee pot. The solubles
concentration of the exiting coffee ðcexit Þ, and the coffee beverage in
the pot ðcbrew Þ is measured throughout the extraction. The extrac-
tion profiles of the brewed and exiting coffee solubles concentra-
here, for both JK standard drip filter grind and Cimbali #20 grind
104
the experiments are repeated for a different coffee bed mass and for
a different absolute pressure of the water. Details of these experi-
ments are included in Appendix G.
We make a number of observations based on the experimental
data presented, which will be used to motivate the development of
a mathematical model of the extraction process. Firstly, it is noted
that the experiments tell us nothing about the filling stage, where
the water initially infiltrates the dry coffee bed. To this end it will
be assumed that when the first coffee brew exits, the coffee bed,
including the intragranular pores, is saturated with water. It is
noted that in the case of the batch extraction experiment, in Fig. 2,
the concentration of the coffee brew increases rapidly from zero to
over half its maximum value at the beginning of extraction, before
a much slower increase over a longer timescale towards its
maximum value. This phenomena can also be seen in the cylind-
rical brewing chamber data in Fig. 3, where the concentration of
600 the initial exiting brew is very high and this is maintained for a
short time, before there is a large drop in the exiting concentration
to a lower level and this gradually declines over a longer period.
This is consistent with the findings in Voilley and Simatos (1979)
where it was noted that extraction yield reached 90% of its final
value within 1 min. Large-sized roasted particles in the coffee
grind feature a kernel comprising undamaged closed cells and a
particle skin formed by damaged open cells. The grind distribution
as seen in Fig. 1 also contains a significant proportion of fines or
damaged cells. Thus the reason for the fast initial extraction may
be due to reduced mass transfer resistances in the damaged
particle skin and in the fines. The slow extraction may then be
due to mass diffusion from intact cells in particle kernel. Thus the
rate limiting step is diffusion from the particle kernel.
3. Mathematical modelling
3.1. Basic modelling assumptions
It is assumed here that the coffee brewing process can be
broken into three stages. Initially in the filling stage, hot water is
poured on the dry coffee grounds and begins to fill the filter, but
does not leave. Next in the steady state stage the bed is saturated,
water is still entering the bed, but also leaving at the same rate. In
the last stage, the draining stage, no more water enters the bed but
it still drains out. In the absence of experimental data to cover the
other stages, only the steady state stage is considered here.
However, the model developed can be easily generalised to
include the unsaturated flow, during the filling and draining
stages. In the steady state stage the coffee bed is considered as a
static, saturated porous medium with the flow driven by a
pressure gradient. This pressure gradient may be mechanically
applied as in an espresso machine or hydrostatic as in a drip filter
machine. The bed is composed of a solid matrix of coffee grains
which are themselves porous. As the filling stage is not modelled
here, initial conditions for the steady state stage will have to be
estimated or inferred from experimental data. Furthermore any
swelling of the coffee grains due to the addition of water will not
be modelled. It is assumed that this swelling would occur during
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 219

200 200

cbrew mg gram

cexit mg gram
150 150

100 100

50 50

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Mbrew grams Mbrew grams

100 100

80 80
cbrew mg gram

cexit mg gram
60 60

40 40

20 20

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Mbrew grams Mbrew grams
Fig. 3. The coffee solubles concentration, measured in mg/gram, is plotted against mass of coffee beverage M brew (grams) for JK drip filter grind with a flow rate of 250 ml/
min and pressure differential of 2.3 bar in (a) the coffee pot and (b) the beverage at filter exit, and for Cimbali #20 grind with a flow rate of 250 ml/min and pressure
differential of 0.65 bar in (c) the coffee pot and (d) the beverage at filter exit. The mass of coffee used in both cases is 60 g including approximately 4% moisture. The brewing
cylinder diameter is 59 mm. The coffee bed heights are 4.05 cm for JK drip filter grind and 5.26 cm for Cimbali #20 grind.

the filling stage and would manifest itself in the steady state stage
possibly as a slight shift to the right of the grind size distribution.
While coffee is composed of over 1800 different chemical com-
pounds (Petracco and Technology, 2008), in this model we will just
consider a single entity and model coffee concentration or brew
strength in line with the experimental data shown. The model can
of course be generalised to model the concentration of any
number of coffee constituents. Modelling a multicomponent
system such as coffee with a single component is a simplification
and requires some justification. Firstly, as mentioned in the
introduction, relating taste to the concentration of the different
coffee components in a beverage is a non-trivial matter and
currently no ideal recipe exists. Apart from professional tasters,
the most widely used measure of coffee quality is the coffee
brewing control chart. This chart gives target ranges for brew
strength and extraction yield. This chart is used by both the
Speciality Coffee Association of Europe (SCAE) and the Speciality
Coffee Association of America (SCAA). Given that the most widely
used measure of coffee quality considers coffee as a single
component, it seems logical to do so as well. Secondly, experi-
ments carried out suggest that many important coffee components
have similar extraction kinetics (see Fig. 8, Appendix F).
3.2. Coffee bed structure
distribution by volume, excluding single cell fragments, will be
approximately an order of magnitude bigger. Finally the size
(depth) of the coffee bed will typically be a few centimetres. In
order to model the transport of coffee and water in the bed
conservation equations can be formed at each of these scales. On
the microscale (cell size scale), conservation equations can be
formed in the h-phase, v-phase and s-phase. At the intermediate
(grain size) scale conservation equations can be formed in the h-
phase and the l-phase. We refer to this as the mesoscale. At this
scale an individual coffee grain is represented by two overlapping
continua representing the void and solid phases within the grain.
At the macroscale (coffee bed scale), the coffee bed is represented
by three overlapping continua representing the h-phase, v-phase
and s-phase. To reconcile the three representations we can use the
methods of homogenisation or volume averaging. Volume aver-
aging will be adopted here. A schematic of the volume averaging
process is shown in Fig. 4. This is very useful since it relates the
averaged macroscopic quantities to the physical parameters at the
microscale. Some macroscopic parameters may be measured by
experiment, while others can be found from their averaged
representation in terms of measurable microscopic quantities.
The influence of microscale properties on the macroscale system
parameters can be easily identified.

The structure of the coffee bed is central to the extraction 3.3. Coffee bed description
process. It is immediately obvious that the bed consists of two
phases. Following Bear and Cheng (2010) the highly permeable The coffee bed is represented by a porous medium domain ΩT
phase consisting of the pores between the coffee grains is called with volume VT. The domain can be split into an intergranular
the h-phase. Similarly the low permeability phase consisting of the pores domain Ωh, with volume Vh and a coffee grain domain Ωl,
coffee grains is called the l-phase. At a microscopic level there are with volume Vl. Ωl is further split into an intragranular pore
two phases within the coffee grains. The pore or void space within domain Ωv, with volume Vv and a solid coffee domain Ωs, with
the grains is called the v-phase, while the solid coffee cellular volume Vs. Clearly the equalities V h þ V l ¼ V T and V v þ V s ¼ V l hold.
matrix is referred to as the s-phase. The coffee bed has three The following volume fractions are now defined as
fundamental length scales. The smallest of these is the size of the
pores within the grains or the size of a coffee cell which may be Vh Vl Vv Vs
ϕh ¼ ; ϕl ¼ ; ϕv ¼ ; ϕs ¼ ; ð1Þ
25–50 μm. The average size of a coffee grain in the grind size VT VT Vl Vl
220 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

3.5.1. v-phase
∂
ðϕ cn Þ ¼  ∇  ðϕv ðcnv vnv þ jv ÞÞ  f v-s :
n n
ð8Þ
∂t v v
∂
ðϕ ρn Þ ¼  ∇  ðϕv ðρnv vnv þ iv ÞÞ:
n
ð9Þ
∂t v v

3.5.2. s-phase
∂
ðϕ cn Þ ¼ ∇  ðϕs js Þ  f s-v :
n n
ð10Þ
∂t s s
n n
The terms f v-s and f s-v are source/sink terms representing
transfer of coffee solubles across the vs-interface and vice-versa.
n
Fig. 4. Macroscale equations are matched to microscale equations using volume The term iv accounts for any mechanical dispersion in the fluid
averaging. At a macroscopic level the system is represented by three overlapping velocity.
continua for the intergranular pores (h-phase), intragranular pores (v-phase) and
solid coffee (s-phase).
3.6. Macroscale point balance equations

which leads to For each phase there are macroscopic point balance equations
ϕh þ ϕl ¼ 1; ϕv þ ϕs ¼ 1: ð2Þ for mass of the coffee and mass of liquid (flow equations). The
solid is assumed stationary. The macroscopic equations take
The concentrations (mass per unit volume) of coffee in the
the form:
respective phases are ch, cv and cs. vh and vv denote the fluid
velocity in the h-phase and v-phase respectively. The velocity of
the solid will be denoted by vs . Further notation will be introduced 3.6.1. h-phase
as required. The formulation of the equations presented here will ∂
ðϕ c~ Þ ¼  ∇  ðϕh ðc~ h v~ h þ j~h ÞÞ  f~ h-l ; ð11Þ
follow (Bear and Cheng, 2010; Gray and Hassanizadeh, 1998). Since ∂t h h
conservation equations will be formed at three different scales the
∂
ðϕ ρ~ Þ ¼  ∇  ðϕh ðρ~ h v~ h þ i~h ÞÞ  f~ h-l :
w
variables at each scale will be denoted as macroscale ð~ Þ, mesoscale ð12Þ
ðn Þ and microscale ðÞ to avoid ambiguity. ∂t h h

3.4. Microscale point balance equations 3.6.2. v-phase

∂
The point balance equations for coffee and liquid within each of ðϕ ϕ c~v Þ ¼  ∇  ðϕl ϕv ðc~ v v~ v þ j~v ÞÞ  f~ v-h  f~ v-s ; ð13Þ
∂t l v
the phases on the microscale are
∂
ðϕ ϕ ρ~ Þ ¼  ∇  ðϕl ϕv ðρ~ v v~ v þ i~ v ÞÞ  f~ l-h :
w
ð14Þ
3.4.1. h-phase ∂t l v v
∂ch
¼  ∇  ðch vh þ jh Þ; ð3Þ 3.6.3. s-phase
∂t
∂ ρh ∂
¼  ∇  ðρh vh Þ: ð4Þ ðϕ ϕ c~s Þ ¼  ∇  ðϕl ϕs j~ s Þ  f~ s-h  f~ s-v : ð15Þ
∂t ∂t l s
f~ α-β is transfer of coffee solubles from the α-phase to the β-phase
across the αβ interface. Similarly f~
w
3.4.2. v-phase α-β is transfer of liquid from the
α-phase to the β-phase across the αβ interface.
∂cv
¼  ∇  ðcv vv þ jv Þ; ð5Þ
∂t 3.7. Upscaling from microscale to macroscale
∂ ρv
¼  ∇  ðρv vv Þ: ð6Þ As mentioned the conservation equations at each of the scales
∂t
can be related by representing the properties of the medium at a
larger scale by averaging the properties at the smaller scale. This is
3.4.3. s-phase useful to find the forms of the mass transfer terms in the
∂cs macroscopic equation. An outline of the general upscaling proce-
¼ 0: ð7Þ dure based on (Bear and Cheng, 2010; Gray and Hassanizadeh,
∂t
1998) is included in Appendix A. The details of upscaling in this
The microscopic balance equations include the terms jh and jv case are included in Appendix B.
which represent molecular diffusion of coffee solubles in the
respective phases. Molecular diffusion in the solid phase is
assumed negligible. 4. Developing macroscale equations

3.5. Mesoscale point balance equations
The mesoscale balance equations are only required in the
grains (l-phase) since only two scales are needed in the h-phase.
Thus the point balance equations for coffee and liquid in the
v-phase and the s-phase on the mesoscale are
The macroscopic balance equations in Section 3.6 are in a quite
general form. Some assumptions have already been made but in
order to simplify things we make some further assumptions. We
also need to introduce terms to model the transport of fluid and
coffee within the bed. Firstly we assume that the density of the
liquid is constant and does not change with coffee concentration.
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 221

a
This is consistent with Petracco and Technology (2008) where it is For an isotropic porous medium, jh is often modelled by
noted that the material extracted from coffee has little influence D
jh ¼  ∇c~ h ;
a
on liquid density. Thus ρ~h ¼ ρ~v ¼ ρ: It is also assumed that no ð23Þ
τ
transport occurs within the l-phase on the macroscale. This simply
means that liquid or coffee does not transport directly from grain where τ is the tortuosity defined by
to grain within the bed. Any mechanical dispersion in the flow in Le actual path length
the h-phase is not considered. Due to continuity of flux at the
τ¼ ¼ : ð24Þ
L macroscopic path length
interphase boundaries we have f~ α-β ¼  f~ β-α . Thus the five
The tortuosity must be estimated as a function of the porosity.
macroscopic equations reduce to the following:
Various functional relationships are proposed in the literature.
∂
ðϕ c~ Þ ¼ ∇  ðϕh ðc~ h v~ h þ j~h ÞÞ þ f~ v-h þ f~ s-h ;
Some of these are discussed in Pisani (2011). The expression used
ð16Þ  1=3
∂t h h here is τ ¼ ϕh which is adopted from Millington (1959). Thus
we have
∂ ϕh
¼  ρ∇  ðϕh v~ h Þ þ f~ l-h ;
w
ρ ð17Þ D
∂t jh ¼  ∇c~ h ¼  ϕh D∇c~ h :
a 1=3
ð25Þ
τ
∂
ðϕ ϕ c~v Þ ¼  f~ v-h þ f~ s-v ; ð18Þ
∂t l v
4.3. Dispersive flux
∂
ρ ðϕl ϕv Þ ¼  f~ l-h ;
w
ð19Þ Dispersion occurs due to variations in the microscopic velocity
∂t
of the phase with respect to the averaged velocity, and molecular
∂ diffusion (Bear and Cheng, 2010). Thus molecular diffusion con-
ðϕ ϕ c~s Þ ¼  f~ s-h  f~ s-v : ð20Þ
∂t l s tributes to the dispersive flux in addition to the diffusive flux at
the macroscopic level. In general the dispersive flux is given by
It now remains to introduce expressions to model the fluid
D ○ ○ Eh
velocity v~ h , the total macroscopic flux j~ h , the fluid mass transfer jh ¼ ch vh ¼  D~  ∇c~ h ;
b b
ð26Þ
term f~ and the coffee mass transfer terms f~ , f~ and f~
w
l-h v-h s-v s-h
where D
b
~ is both
~ is a rank 2 tensor called the dispersion tensor. D b
in terms of the system variables. The main transfers occurring in
the coffee bed are shown in Fig. 5. positive definite and symmetric. For an isotropic porous medium
the following expression is often used:
 vi vj 
4.1. Fluid velocity Dij ¼ aT δij þ ðaL  aT Þ 2 v: ð27Þ
v
The coefficients aL and aT here are the longitudinal and transverse
Darcy's Law allows us to relate an averaged velocity or
dispersivities of the porous medium. vi ¼ hvi ih is the average
discharge in the pores to the pressure gradient. The relations in
velocity in the i-th direction and v ¼ j vj where v is the average
the h-phase are given by
velocity vector in this instance. δij is the Kronecker delta. Further
k~ h detail on the diffusive and dispersive fluxes is included in
u~ h ¼ ϕh v~ h ; u~ h ¼  ð∇p~ h þ ρgÞ; k~ h ¼ k~ h ðϕh Þ; ð21Þ
μ Appendix C.

where p~ h is the macroscopic pressure gradient in the h-phase, k~ h is

4.4. Coffee mass transfer terms
the permeability and μ is the viscosity of water.
Experimental results in Section 2 suggest that there are two
4.2. Total macroscopic flux fundamental extraction mechanisms from the coffee grains. A
rapid extraction from fines (single cells fragments) and damaged
The total macroscopic flux, j~ h is made up of the macroscopic cells on the surface of larger particles and a slower extraction from
a b
average of molecular diffusion jh and the dispersive flux jh : the kernels of larger particles. Various different models have been
applied to represent such situations, particularly in the area of

h D ○ ○ Eh supercritical fluids. The main models are reviewed in Oliveira et al.
j~h ¼ jh þ ch vh ¼ jh þ jh :
a b
ð22Þ
(2011). These models include linear driving force, bi-linear driving
force, shrinking core, broken plus intact cells in series, broken plus
Water intact cells in parallel and a combined broken plus intact cells with
Reservoir shrinking core model. These models are widely used in modelling
Intragranular extraction with supercritical fluids and have been considered in a
Water to Pores
Water in Grains number of papers, both theoretical and experimental, including
Huang (2012), Sovov (2005), Goto et al. (1996), Machmudah et al.
Coffee Diffusion (2012). In (Huang, 2012) it is noted that the broken plus intact cell
Intergranular From Bulk Coffee Dissolving
model typically features an initial constant extraction period
Pores in grains
dominated by extraction from the broken cells, then a falling
extraction rate period as broken cells on the surface are depleted
Coffee Coffee and finally a diffusion controlled period dominated by extraction
Brew Coffee Solids
Dissolving from the intact cells. These general features are evident in the
Out From
extraction experiments in Section 2. The models used here are
Surface
Coffee quite similar to the broken plus intact cells in parallel. The
Pot extraction term f~ v-h is transferred from the intragranular pores
in the grains to the intergranular pores and is similar to extraction
Fig. 5. Mass transfers occurring in the coffee bed. from intact cells. The term f~ s-h is direct extraction from the solid
222 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

grain matrix surface into the intergranular pores and is similar to
extraction from the broken cells. The term f~ s-v models dissolution
of coffee solubles from the cell walls into the intragranular pores.
The transfer f~ v-h across the interphase boundary is assumed to
occur by diffusion according to Fick's first law. Thus
f~ v-h ¼ αvh ðc~ v  c~ h Þ: ð28Þ
where αvh is the mass transfer coefficient. The form of αvh can be
found from the volume averaging procedure as
Snhl
αvh ¼ ð1  ϕh Þϕv4=3 Dv ; ð29Þ
Δl
where Dv is the diffusion coefficient of coffee in water, Snhl is the
specific surface area of the l-phase and Δl is some length scale
characterising the distance over which diffusion occurs. The
transfers f~ s-v and f~ v-h are solid–fluid transfer rather than fluid–
fluid transfer and so the modelling is slightly different. Under
consideration here is the case where the solid matrix itself is
dissolving. It is assumed that there is a thin layer of liquid next to
the solid which is always saturated with solute. This concentration
is denoted by csat . csat is the concentration in the liquid phase that
would be in equilibrium with the concentration inside the solid c~ s .
The force of extraction from this thin layer to the bulk of the fluid
is assumed to be proportional to the difference in concentration
between the thin layer and the bulk of the fluid. Thus, again using
volume averaging to determine mass transfer coefficient, we have
the only change in porosity occurs within the l-phase. Thus it
seems reasonable to assume that the pressure in the intergranular
pores is always greater than or equal to the pressure in the
intragranular pores, i.e. p~ h Z p~ v . Thus (33) simplifies to
wn kv ðϕv Þ ðp~ h  p~ v Þ
f~ h-l ¼ ð1  ϕh ÞSnhl ρ c~ h : ð34Þ
μ Δl
4.6.1. Specific surface area of l-phase
Approximating a coffee grain of diameter x by a sphere, it has a
surface to volume ratio of 6=x. This can be improved upon if the
roundness or sphericity of the coffee grain is known. However in
the absence of this we make the spherical approximation. We can
use a number of equivalent spherical diameters to represent the
entire size distribution. One such diameter is the diameter of the
spherical particle that has the same specific surface area of that
distribution. This is called the Sauter mean diameter and is defined
by
6
ksv ¼ ; ð35Þ
Sv
where Sv is the surface to volume ratio of the distribution which
can be found from the data. The assumption ϕh is constant means
that the Sauter mean diameter does not change. Thus where
required the specific surface area of the l-phase is given by

S n 6
f~ s-h ¼ ð1  ϕh Þð1  ϕv ÞDh hl ðcsat  c~ h Þ; ð30Þ Snhl ¼ : ð36Þ
Δs ksv

where Δs is some length scale characterising the distance over In fact we will use two separate Sauter mean diameters. In relation to
which diffusion occurs. Similarly flow and extraction from the surface of the grains we need to use the
n
specific surface area of the entire distribution. We denote the
S
f~ s-v ¼ ð1  ϕh Þð1  ϕv ÞDv sv ðcsat  c~ v Þ; ð31Þ corresponding Sauter mean diameter ksv1. However when dealing
Δs with extraction from the grain kernel, we should ignore the particles
where Snsv is the specific surface area of the s-phase. The transfer which are just broken cell fragments and do not have a kernel of
terms are considered in more detail in Appendix D. intact cells. In this case we introduce a second Sauter mean diameter
ksv2 which is representative of the specific surface area of particles
4.5. Liquid transfer term above a certain diameter, chosen here to be 50 μm. This Sauter mean
diameter is used for extraction from the grain kernel.
It is also possible to have a transfer of liquid from the
intergranular pores to the intragranular pores (or vice versa). This 4.6.2. Specific surface area of s-phase
could occur for example due to a pressure imbalance in the phases The specific surface area of the s-phase is more difficult to
due to the dissolution of the solid matrix within the grains. Using estimate. Assuming that the coffee grain is made up of solid
Darcy's Law and volume averaging we find that spherical cells with the same diameter m we could approximate
the specific surface area by
kv ðϕv Þ ðp~ h  p~ v Þ
f~ l-h ¼  ð1  ϕh ÞSnlh ρ
w
; ð32Þ
μ Δl Snsv ¼
6
: ð37Þ
m
where kv ðϕv Þ is the coffee grain permeability and p~ h and p~ v are the
However we do not have any information about the specific
water pressures in the h-phase and the v-phase. Since there is a
surface area of the s-phase from the grind size distribution and
difference in concentration in the h and v-phases the transfer of
in practice this will form part of a lumped mass transfer coefficient
fluid in either direction would be expected to also result in the
which will have to be fitted to the experimental data.
transfer of solute from one phase to another. This can be included
as an extra coffee mass transfer term as
4.6.3. Permeability kh
8 The permeability can be estimated using the Kozeny–Carman
<  ð1  ϕh ÞSnlh ρkv ðμϕv Þ ðp h Δ p v Þc~ h
~ ~
if p~ h Z p~ v
wn
f l-h ¼
l
ð33Þ equation for spheres (Holdich, 2002):
: ð1  ϕh ÞSlh ρ n kv ðϕv Þ ðp~ h  p~ v Þ
μ Δl c~ v if p~ h o p~ v
ϕ3h
kh ¼ : ð38Þ
κ ð1  ϕh Þ2 Snhl 2
4.6. Coffee bed properties
Here again Snhl is the specific surface area, while κ is a factor which
In order to proceed the quantities such as permeability, surface accounts for the shape and tortuosity. Utilising the derived form
area, specific surface area and distances over which diffusion for Snhl gives
occurs must be expressed in terms of measurable properties of ksv1 ϕh
2 3
ksv1 ϕh
2 3
the coffee bed. We can estimate these quantities from the particle kh ¼ ¼ : ð39Þ
36κ ð1  ϕh Þ 2
36κ ð1  ϕh Þ2
size distribution. In order to simplify things we assume that the
intergranular porosity ϕh is constant. We note that this means that The shape factor κ is usually taken to be in the range 2–6.
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 223

 
Experiments performed measuring the pressure drop in an airflow ∂ψ v 12Dv ϕc0 csat  c~ v
¼ rv ψ v ; ð44Þ
through compacted coffee beds estimate it at κ ¼ 3:1. ∂t m2 c~ s
where r s ¼ 1=ϕs;s0 and r v ¼ 1=ϕs;b0 .
4.6.4. Permeability kv
Similarly the permeability of the grain is estimated using the 4.7. Macroscale equations
Kozeny–Carman equations for the spherical cells so that
Models have now been introduced for the various processes
m2 ϕ v
3
kv ¼ : ð40Þ occurring in the coffee bed. The description of the process has
180ð1  ϕv Þ2 been extended to seven coupled partial differential equations.
These equations are presented in full in (45)–(51). Different
In the absence of experimental data we have chosen κ ¼ 5 which is
presentations of the equations are possible and one equation can
often adopted (Aubertin and Chapuis, 2003).
be reduced to an algebraic one. Boundary conditions will depend
on the geometry of the problem. Initial conditions will have to be
4.6.5. Average diffusion distances determined or inferred from the experiment following the filling
Expressions are required for Δl and Δs. The distance over which stage.
diffusion from the grains to the large pores occurs, is assumed to ksv1 ϕh
3
∂c~h
2

be equal to the mean radius of the grains weighted by volume ϕh ¼ ∇  ðc~ h ð∇p~ h þ ρgÞÞ
∂t 36κμð1  ϕh Þ2
which we denote Δl ¼ ll . The distance over which diffusion occurs 4

from the surface of the solid to the h-phase is assumed to be equal ~  ∇2 c~

þ ϕ3h Dh ∇2 c~ h þ ϕh D
b
h
to the mean radius of the coffee cells. Thus Δs ¼ ls . 6
 ð1  ϕh Þϕv Dv
4=3
ðc~  c~ v Þ
ksv2 ll h
12Dh ϕc0
4.6.6. Coffee extraction limits þ ð1  ϕh Þ ðcsat  c~ h Þψ s
ksv1 m
The current forms of the coffee mass transfer terms are
6ð1  ϕh Þm2 ϕv 3
incomplete as they do not account for the fact that for a given  ðp~  p~ v Þc~ h ; ð45Þ
grind at a given temperature there is a maximum amount of coffee 180μksv2 ll ð1  ϕv Þ2 h
that can be extracted. Looking at our equations, in order to track
ksv1 ϕh
2 3
the amount of coffee extracted we can either allow the grain
porosity ϕv to change or allow the solid coffee concentration c~ s to 0¼  ∇  ð∇p~ h þ ρgÞ
36κμð1  ϕh Þ2
change, or both. The decision is made here to allow ϕv to change
6ð1  ϕh Þm2 ϕv 3
as coffee is extracted and consider the solid concentration c~ s fixed. þ ðp~  p~ v Þ; ð46Þ
180μksv2 ll ð1  ϕv Þ2 h
Thus ϕv is allowed to increase as coffee is dissolved until it reaches
a point when ϕs consists of insoluble material. This maximum
∂ 6
ðϕ c~v Þ ¼  ϕv Dv
4=3
value may depend on the grind size and temperature of extracting ðc~  c~ v Þ
water and should be found from the experimentally determined ∂t v ksv2 ll h
12ϕc0 Dv
maximum extractable solubles mass. It is also unrealistic for the þ ðcsat  c~ v Þψ v
mass transfer coefficients from the solid matrix surface to remain m2
approximately constant even when there is very little soluble 6ð1  ϕh Þm2 ϕ3v
þ ðp~  p~ v Þc~ h ; ð47Þ
coffee left in the solid. This issue is sometimes solved through use 180μksv2 ll ð1  ϕv Þ2 h
of a partition coefficient. Here we make the simple assumption
∂ ϕv ϕv 3
that the extraction term be proportional to the amount of coffee 6 m2
on the surface. Let ϕc be the volume fraction of coffee in the grains. ¼ ðp~  p~ v Þ; ð48Þ
∂t 180μksv2 ll ð1  ϕv Þ2 h
We divide this into coffee in the surface of the grains and fines ϕs;s ,
and coffee in the grain kernels ϕs;b , so that ϕc ¼ ϕs;s þ ϕs;b . The ∂ ϕv 1 ∂ψ s 1 ∂ψ v
initial volume fractions of coffee in dry grains are ϕc0, ϕs;s0 and ¼  ; ð49Þ
∂t r s ∂t r v ∂t
ϕs;b0 . Assuming that the soluble coffee is uniformly distributed  
within the grains the initial volume fraction everywhere is ϕc0. ∂ψ s 12Dh ϕc0 csat  c~ h
¼ rs ψ s ; ð50Þ
Thus at a given time the volume fractions of soluble coffee on the ∂t ksv1 m c~ s
surface and in the grain kernels are given by  
∂ψ v 12Dv ϕc0 csat  c~ v
ϕc0 ¼ rv ψ v ; ð51Þ
ϕ ðx; tÞ ¼ ϕc0 ψ s ðx; tÞ; ð41Þ ∂t m2 c~ s
ϕs;s0 s;s
While these equations give estimates for the different coefficients
ϕc0 in terms of process parameters, these parameters may not always
ϕ ðx; tÞ ¼ ϕc0 ψ v ðx; tÞ; ð42Þ be easy to determine accurately. Other processes which were not
ϕs;b0 s;b
considered may also affect the transport of coffee and water. Thus
where x is the position within the coffee bed and ψs and ψv are the when comparing to experiment it is necessary to introduce fitting
fractions of the original amount of coffee left on the grain surfaces parameters particularly to the terms controlling extraction from
and in the grain kernels respectively. We can now substitute the grain surfaces and diffusion of coffee from the grain kernel.
ϕs ¼ ϕs;i þ ϕs;s þ ϕs;b . ϕs;i represents the volume fraction of insolu-
ble solid in the grains. Using the expressions we have developed so
far this leads to two further partial differential equations for ψs 5. Numerical simulations
and ψv:
  Numerical simulations of the experiments in Sections 2.2 and
∂ψ s 12Dh ϕc0 csat  c~ h 2.3 are conducted using the model equations. The equations can
¼ rs ψ s ; ð43Þ
∂t ksv1 m c~ s be reduced in both of these cases.
224 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

5.1. Batch-wise brewing with initial conditions

c~ h ð0Þ ¼ 0; c~ v ð0Þ ¼ cv0 ; ϕv ð0Þ ¼ ϕv0 ; ψ s ð0Þ ¼ 1: ð56Þ
In the French press type brewing apparatus there is no pressure
n
induced flow and we assume the solution in the h-phase is well The parameters αn and β are used as fitting parameters to fit the
mixed since only the average concentration in this phase is experimental results. The mass transfer coefficients depend on the
measured anyway. Thus all spatial derivatives in the equations volume fraction of the interstitial water in the coffee bed in
drop out leaving a system of ordinary differential equations. In question, so the fitting parameters are just intended to correct
order to solve the system we require some initial conditions. Here for errors in the other parameters in the mass transfer coefficient.
Thus we look for values of αn and β which fit both experiments
n
we need to make some assumptions since we do not model the
initial infiltration of the water into the grains when the coffee and equally well. The initial volume fractions of coffee ϕs;s0 and ϕs;b0
water are mixed. Firstly, we assume that the grains are saturated can, to some extent, be estimated from the grind size distribution
with water initially. We further assume that the initial concentra- but here we will use them to allow for any differences between the
tion in the h-phase is zero, so that none of the coffee has dissolved batch extraction and coffee flow-through-cell experiments. These
from the surface of the grains. It is also necessary to give an initial differences may arise, for example, due to the fact that we do not
condition for the concentration in the v-phase. It is assumed model the initial water infiltration into the coffee grains. The
initially that all the soluble coffee in the grains has dissolved into parameters used in the simulations are listed in Table 1. The other
the v-phase. This is of course unlikely to be the case. However, parameters are all measured experimentally, estimated from
since coffee diffusion from the v-phase to the h-phase seems to be experiments or sourced in the literature. The value for the coffee
the rate limiting process, occurring much slower than dissolution solubility csat is estimated from the highest observed concentra-
of coffee, this assumption is unlikely to have much impact on the tion across the four experiments. The diffusion coefficient is for
simulation results. We also assume that, only a change in porosity caffeine in water at 80 1C (Jaganyi and Madlala, 2000). This would
in the intact cells in the grain kernel will result in a change in be slightly higher at 90 1C. However this is only an estimate of the
pressure in the v-phase, since any pressure imbalance in the effective diffusion coefficient of coffee in water. The fitting para-
damaged cells on the grain surface would be almost instanta- meters are used to correct any errors in these parameters. The
neously corrected. Thus, as all coffee in kernel is assumed comparison between the numerical solution and the experimental
dissolved initially, no pressure difference between phases will results is shown in Fig. 6.
occur. Note the value of intragranular porosity ϕv ¼ 0:56 is
adopted here for dry coffee in air. Allowing for these assumptions 5.2. Cylindrical brewing chamber
the following system of equations are solved numerically in
MATLABs . The cylindrical brewing chamber geometry also allows us to
make some simplifications to the general model. Assuming that
dc~ 4 6
ϕh h ¼  αn ð1  ϕh Þϕ3v Dv ðc~  c~ v Þ the coffee bed properties are homogeneous in any cross-section,
dt ksv2 ll h
the equations can be reduced to one spatial dimension, parallel to
12Dh ϕc0 the flow direction. Thus the bed depth is labelled by the z-
þ β ð1  ϕh Þ
n
ðcsat  c~ h Þψ s ; ð52Þ
ksv1 m coordinate. The height of the coffee bed is L with the bottom
(filter exit) at z ¼0 and the top (filter entrance) at z ¼L. For the
d 6
ðϕ c~v Þ ¼  αn ϕv Dv experiments here it is shown in Appendix E that advection
4=3
ðc~  c~ v Þ; ð53Þ
dt v ksv2 ll h dominates over diffusion and mechanical dispersion, so these
two processes are neglected. This means we just require one
dϕv 1 ∂ψ s concentration boundary condition for c~ h . We use the condition
¼ ; ð54Þ
dt r s ∂t that the water entering at the top z¼L has zero concentration. The
  pressure boundary conditions are p~ h ¼ Δp at z ¼L and p~ h ¼ 0 at
dψ s n 12Dh ϕc0 csat  c ~h
¼ β rs ψ s ; ð55Þ z¼0. Δp is the pressure difference across the bed. In these
dt ksv1 m c~ s
experiments there is a large pressure difference across the bed
which can lead to bed compaction and a reduction in the porosity.
Table 1 Thus the porosity cannot be measured a priori. It assumed that ϕh
adjusts to the pressure while ϕv stays constant. We then choose ϕh
Parameters for simulation of the batch extraction experiments.

Parameter JK drip filter Cimbali #20 by matching the volume flow from Darcy's Law and the Kozeny–
Carman equation to the experimental volume flow. Once again the
ϕv0 0.6444 0.6120 initial conditions need to be inferred. It is assumed that once
ϕh 0.8272 0.8272
brewed coffee starts to flow from the bed that the bed is fully
ksv1 27:35 μm 38:77 μm
ksv2 322:49 μm 569:45 μm
saturated with liquid. Again it is assumed that all the coffee in the
ll 282 μm 463 μm grain kernels has dissolved into the intragranular pores (v-phase)
Dh ¼ Dv 2:2  10  9 m2 s  1 2:2  10  9 m2 s  1 during filling and this is uniformly distributed in the bed. It is also
ρ 965:3 kg m  3 965:3 kg m  3 necessary to estimate the initial concentration c~ h as a function of z.
μ 0:315  10  3 Pas 0:315  10  3 Pas Based on the initial exiting concentrations we assume for the fine
m 30 μm 30 μm grind (JK) that initially c~ h is at the coffee solubility throughout the
csat 212:4 kg m  3 212:4 kg m  3
h-phase. For the coarser grain, (Cimbali # 20), we assume a linear
c~ s 1400 kg m  3 1400 kg m  3
concentration profile in the h-phase, rising from zero at the top to
κ 3.1 3.1
ϕc0 0.143435 0.122 the initial exiting concentration at the bottom of the bed. Thus it is
ϕs;s0 0.059 0.07 assumed that some extraction from the fines and broken surface
ϕs;b0 0.084435 0.052 cells occurs during filling. To this end the amount of surface coffee
αn 0.1833 0.0881 is uniformly reduced by the corresponding amount of coffee
βn 0.0447 0.0086
initially present in the h-phase. In reality we would expect more
rs 16.94 14.28
cv0 183:43 kg m  3 118:95 kg m  3
extraction would have occurred at the top of the bed during filling
than the bottom, but, in the absence of experimental guidance, we
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 225

40 35

35 30

30
25
25
C (kg/m3)

Ch (kg/m )
3
20
20
15
h

15
10
10

5 5

0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600

t (s) t (s)

Fig. 6. Comparison between numerical solution (–) with parameters from Table 1 and experiment (n) for the batch extraction experiments for (a) JK drip filter grind and
(b) Cimbali #20 grind.

Table 2 and boundary conditions

Parameters for simulation of the cylindrical brewing chamber extraction
c~ h ðL; tÞ ¼ 0; p~ h ð0; tÞ ¼ 0; p~ h ðL; tÞ ¼ Δp: ð65Þ
experiments.
The initial concentration profile in the JK drip filter grind is
Parameter JK drip filter Cimbali #20
given by ch0 ðzÞ ¼ csat . In the Cimbali #20 grind the initial concen-
ϕv0 0.6231 0.6218 tration is given by ch0 ðzÞ ¼ ðcmax =LÞðL  zÞ. Some of the parameters
ϕh 0.2 0.25 are the same as those in the batch extraction case. Any new
cmax – 82:63 kg m  3 parameters or parameters that have changed are included in
cs 1400 kg m  3 1400 kg m  3 Table 2. The numerical solution is found using finite differences
ϕs;s0 0.11 0.07
in the spatial direction and the method of lines. It should be noted
ϕs;b0 0.033435 0.052
αn 0.1833 0.0881
that any initial discontinuities in the initial and boundary condi-
βn 0.0447 0.0086 tions will be smoothed out by numerical diffusion. If accuracy is
rs 9.09 14.28 required in the initial stages a large number of steps should be
cv0 78:88 kg m  3 78:88 kg m  3 used in the spatial partition. The comparison between the numer-
Δp 230000 Pa 65000 Pa ical solution and the experimental results is shown in Fig. 7.
L 0.0405 m 0.0526 m

make the simplest assumption. As with the batch experiments the 6. Conclusion
assumptions mean that p~ h ¼ p~ v . Based on these assumptions the
reduced set of equations to model extraction in the cylindrical The coffee extraction process can be described very effectively
brewing chamber is given by using mathematical models. In this paper a general model is
   introduced to describe coffee extraction by hot water from a bed
ksv1 ϕh
3
∂c~
2
∂ ∂p~ h
ϕh h ¼ ~h
c þ ρg of coffee grains. The coffee bed is modelled as a saturated porous
∂t 36κμð1  ϕh Þ2 ∂z ∂z medium using a double porosity model. The bed consists of two
6 kinds of pores: pores between the grains (intergranular) and pores
 αn ð1  ϕh Þϕv Dv
4=3
ðc~  c~ v Þ
ksv2 ll h within the grains (intragranular). Flow of liquid within the coffee
12Dh ϕc0 bed is modelled using Darcy's Law and the Kozeny–Carman
þ β ð1  ϕh Þ
n
ðcsat  c~ h Þψ s ; ð57Þ
ksv1 m equation. Motivated by experiment, extraction of coffee from the
coffee grains is modelled using two mechanisms. Coffee on the
∂2 p~ h damaged grain surface and in coffee cell fragments or fines
¼ 0; ð58Þ
∂z2 extracts quickly into the intergranular pores due to a relatively
low mass transfer resistance. Coffee in intact cells in the grain
∂ 6
ðϕ c~v Þ ¼  αn ϕv Dv
4=3
ðc~  c~ v Þ; ð59Þ kernels, first extracts into the intragranular pores and then slowly
∂t v ksv2 ll h
diffuses through the grain into the intergranular pores.
The model is parameterised using experimental data. Numer-
∂ ϕv 1 ∂ψ s
¼ ; ð60Þ ical simulations are performed and fitted to data. It is shown that
∂t r s ∂t
the developed models can quantitatively describe extraction from
 
∂ψ s n 12Dh ϕc0 csat  c ~h ground coffee in two situations: in a well stirred dilute suspension
¼ β rs ψ s ; ð61Þ of coffee grains, and in a packed coffee bed. The extraction curves
∂t ksv1 m c~ s
fit the data for both a fine and a coarse grind as the parameters in
for the model vary with the surface area and the mean grain radius of
t 4 0; 0 o z oL; ð62Þ the grind size distribution. Provided the grind size distribution is
known, and the physics of extraction is the same, the model
with initial conditions
should work well for an even wider range of grind sizes.
c~ h ðz; 0Þ ¼ ch0 ðzÞ; c~ v ðz; 0Þ ¼ cv0 ; ð63Þ The model described can be easily generalised to describe
standard coffee brewing techniques. It can also be extended to
ϕv ðz; 0Þ ¼ ϕv0 ; ψ s ðz; 0Þ ¼ ψ s0 ; ð64Þ include unsaturated flow during water infiltration into the coffee
226 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

250 100

90

200 80

70
Cexit (kg/m )

Cexit (kg/m )
3

3
150 60

50

100 40

30

50 20

10

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Mbrew (kg) Mbrew (kg)

Fig. 7. Comparison between numerical solution (–) with parameters from Table 2 and experiment (n) for the cylindrical brewing chamber extraction experiments for (a) JK
drip filter grind and (b) Cimbali #20 grind.

bed and drainage of water from the coffee bed. The model is also pore length scale and L is the full problem length scale then
adaptable to different bed geometries. The model is presented for d⪡l⪡L: ð68Þ
isothermal conditions but may also be extended in future work to
include heat transfers within the coffee bed. The following notation is used

The volume of an REV is δV.

The portion of the volume of an REV occupied by the α-phase is
Acknowledgements δV α .

The union of the interfacial regions within the REV between the
The authors acknowledge the support of MACSI, the Mathe- α-phase and a β-phase is denoted Sαβ .
matics Applications Consortium for Science and Industry (www.
The unit vector normal to this surface oriented outward from
macsi.ul.ie ) funded by the Science Foundation Ireland Investigator the α-phase is nα .
Award 12/IA/1683.
To average the microscopic balance equations two theorems that
transform the average of a derivative to the derivative of an
average are needed. The time averaging theorem is given by
Appendix A. General upscaling procedure Z Z XZ
∂F ∂
dV ¼ F dV  nα  wb F j α dS; ð69Þ
δV ∂t ∂t δV β a α Sαβ
The following procedure is mainly adapted from Gray and
Hassanizadeh (1998) and Bear and Cheng (2010). In order to where F is some scalar field property of the microscale and F j α just
compare microscopic and macroscopic equations a procedure is indicates that the microscale property F in the α-phase is being
needed for upscaling or averaging equations from a smaller scale integrated over the αβ -interface. wb is the velocity of Sαβ . The
to a larger scale. A general upscaling procedure from a microscopic P
summation β a α just denotes a summation over all phases
scale to a macroscopic scale is outlined here. except the α-phase.
Let the general form of a point balance equation for the The divergence averaging theorem is given by
concentration of a γ-species per unit volume of the α-phase be Z Z X Z
∇  B dV ¼ ∇  B dV þ nα  Bj α dS; ð70Þ
∂cα δV δV
¼  ∇  ðcα vα þ jα Þ þ Gα ; ð66Þ βaα Sαβ
∂t
where B is some vector field property of the microscale and Bj α
where cα is the concentration of the γ-species, vα is the fluid just indicates that B is being integrated over the αβ-interface. The
velocity in the α-phase, jα is the diffusive flux of the γ-species in P
summation β a α just denotes a summation over all phases
the α-phase, and Gα is a source or sink of the γ-species in the α- except the α-phase.
phase. Next some phase averages are defined
Let β represent all other phases in the porous medium. The α The Intrinsic Phase Average of a quantity F over the α-phase is
subscript will be dropped where convenient. Eq. (66) is the defined by
microscopic balance. The corresponding macroscopic balance Z
α 1
equation is hF iα ¼ F ¼ α F dV: ð71Þ
δ V δV α
∂
ðϕ cn Þ ¼  ∇  ðϕα cn vn þ ϕα j Þ þ ϕα Gn ; α-phase is defined by
n
ð67Þ The Phase Average of a quantity F over the
∂t α Z
1 α
where ϕα is the volume fraction of the α-phase and the n hF i ¼ F ¼ F dV ¼ ϕα F : ð72Þ
δV δV α
variables are macroscopic variables.
To relate (66) and (67) the averaged macroscale properties are The Mass Weighted Average of a quantity F over the α-phase is
defined by averaging (or integrating) (66) over an appropriate REV defined by
or Representative Elementary Volume. The length scale of an REV Z
α

0 α α 1 ρF
is much greater than the pore scale but much less than the full F ¼ F 0 ¼
α α ρ F dV ¼
α : ð73Þ
ρ δ V δV α ρ
scale of the system. So if l is the length scale of an REV, d is the
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 227

It will also be necessary to be able to write the average of a Thus it can be seen that the microscale convection and diffusion
product of the form hcviα in terms of the individual averages hciα processes at the interfaces are source terms for the macroscopic
and hviα . To do this consider the following: equation. It can also be seen that the macroscopic diffusive flux is
the sum of the averaged microscopic diffusive flux and the

Let v○ ¼ v  hviα be the deviation of the true velocity from the dispersive flux.
mean velocity.

Let c○ ¼ c  hciα be the deviation of the true concentration from
the mean concentration. Appendix B. Upscaling from microscale to macroscale
equations
Then
D ○ Eα
The upscaling process basically involves choosing an REV
α
v ¼ v  hviα ¼ hviα  hviα ¼ 0; ð74Þ (Representative Elementary Volume) around every point on the
larger scale and representing the properties of the medium by the
D ○ Eα
α averaged properties of the smaller scale over the REV. For these
c ¼ c  hciα ¼ hciα  hciα ¼ 0: ð75Þ
purposes suitable averages need to be defined including a phase
Thus average and an intrinsic phase average. Also needed are a time
D E averaging theorem and a divergence averaging theorem. An out-
○ ○ α
hcviα ¼ ðhciα þcÞðhviα þ vÞ line of the upscaling procedure used here is given in Appendix A.
D ○ ○ Eα D ○ Eα D ○ Eα Before continuing some notes on the REVs being used are
¼ hciα hviα þ cv þ c hviα þ hciα v necessary. Two different REVs will be used. One will have a scale
D ○ ○ Eα between that of a coffee cell and a coffee grain. The second will
¼ hciα hviα þ cv : ð76Þ have a scale between that of a coffee grain and the coffee bed. For
D○ ○ Eα the REVs the following notation is used.
The term cv is called the dispersive flux.
Now the averaging process can be performed. Firstly (66)
integrated over δV α yields V1 : Volume of the smaller REV.
Z Z Z V 1s : Volume of solid in the smaller REV.
∂c V1v : Volume of void space in smaller REV.
dV ¼  ∇  ðcv þ jÞ dV þ G dV: ð77Þ
δV α ∂t δV α δV α V0 : Volume of larger REV.
Now applying theorems (69) and (70) yields V 0l : Volume of grains (l-phase) in larger REV.
Z Z  V 0h : Volume of void space (h-phase) in larger REV.
∂
c dV ¼  ∇  ðcv þ jÞ dV Also due to the properties of an REV
∂t δV α δV α
XZ V 0h V h V 0l V l
 nα  ðcðv wb Þ þjÞj α dS ϕh ¼ ¼ ; ϕl ¼ ¼ ; ð85Þ
V0 VT V0 VT
β a α Sαβ
Z
þ ð78Þ V 1v V v V 1s V s
G dV: ϕv ¼ ¼ ; ϕs ¼ ¼ : ð86Þ
δV α V1 Vl V1 Vl
This can be written in terms of averaged quantities by dividing
α
across by δV ¼ δϕV to give
α
∂
α B.1. Equations for upscaling
ðϕ hciα Þ ¼  ∇  ϕα ðhcviα þ j Þ
∂t α Z
1 X For convenience the equations at each of the coffee bed length
 nα  ðcðv  wb Þ þ jÞj α dS
δV β a α Sαβ scales from the paper are reproduced here.
þ ϕα hGiα : ð79Þ
B.1.1. Microscale point balance equations
Utilising the formula for the average of a product it can be seen The point balance equations for coffee and liquid within each of
that the phases on the microscale are
∂ D○ ○ Eα

α
ðϕα hciα Þ ¼  ∇  ϕα ðhciα hviα þ cv þ j Þ
∂t Z B.1.1.1. h-phase.
1 X
 nα  ðcðv  wb Þ þ jÞj α dS ∂ch
δV β a α Sαβ ¼ ∇  ðch vh þ jh Þ; ð87Þ
∂t
α
þ ϕα hGi : ð80Þ
∂ ρh
¼  ∇  ðρh vh Þ: ð88Þ
Comparing this to the macroscopic point balance Eq. (67) gives the ∂t
following relations between the macroscopic and microscopic
quantities
B.1.1.2. v-phase.
cn ¼ hciα ; ð81Þ
∂cv
¼  ∇  ðcv vv þ jv Þ; ð89Þ
vn ¼ hviα ; ð82Þ ∂t

D○ ○ Eα
 ∂ ρv
n α ¼  ∇  ðρv vv Þ: ð90Þ
j ¼ cv þ j ; ð83Þ ∂t

ϕα Gn ¼ ϕα hGiα B.1.1.3. s-phase.

Z
1 X
þ nα  ðcðv  wb Þ þ jÞj α dS: ð84Þ ∂cs
δV β a α Sαβ ¼ 0: ð91Þ
∂t
228 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

The microscopic balance equations include the terms jh and jv which Comparing this averaged form with the mesoscale form in (92) it
represent molecular diffusion of coffee solubles in the respective can be seen that
phases. Molecular diffusion in the solid phase is assumed negligible.
cnv ¼ hcv iv ; ð101Þ

B.1.2. Mesoscale point balance equations vnv ¼ hvv iv ; ð102Þ

The mesoscale balance equations are only required in the
D ○ ○ Ev

grains (l-phase) since only two scales are needed in the h-phase. n v
jv ¼ cv vv þ jv ; ð103Þ
Thus the point balance equations for coffee and liquid in the v-
phase and the s-phase on the mesoscale are Z
n 1
f v-s ¼ nv  jv dS: ð104Þ
V1 Svs
B.1.2.1. v-phase.
∂
ðϕ cn Þ ¼  ∇  ðϕv ðcnv vnv þ jv ÞÞ  f v-s :
n n
ð92Þ B.2.2. Conservation of liquid in v-phase
∂t v v
The averaged form of the microscopic point balance Eq. (90)
∂ over the smaller REV is
ðϕ ρn Þ ¼  ∇  ðϕv ðρnv vnv þ iv ÞÞ:
n
ð93Þ
∂t v v  D ○ ○ Ev 
∂
v
v
ðϕv ρv Þ ¼  ∇  ϕv ρv hvv iv þ ρv vv
∂t
B.1.2.2. s-phase. Z
1
 nv  ðρv ðvv  wvs ÞÞ dS: ð105Þ
∂ V 1 Svs
ðϕ cn Þ ¼  ∇  ðϕs js Þ  f s-v :
n n
ð94Þ
∂t s s
Now the surface Svs is a material surface so vv  wvs ¼ 0. Compar-
ing this averaged form with the mesoscale form in (93) it can be
n n
The terms f v-s and f s-v are source/sink terms representing seen that
transfer of coffee solubles across the vs-interface and vice-versa.

n
The term iv accounts for any mechanical dispersion in the fluid ρnv ¼ ρv v ; ð106Þ
velocity.
vnv ¼ hvv iv ; ð107Þ
B.1.3. Macroscale point balance equations D ○ ○
Ev
iv ¼ ρv vv
n
For each phase there are macroscopic point balance equations for : ð108Þ
mass of the coffee and mass of liquid (flow equations). The solid is
Here in fact ρ will be taken to be constant so iv ¼ 0 in this case.
n
assumed stationary. The macroscopic equations take the form

B.2.3. Conservation of coffee solid in s-phase

B.1.3.1. h-phase.
The averaged form of the microscopic point balance Eq. (91)
∂
ðϕ c~ Þ ¼ ∇  ðϕh ðc~ h v~ h þ j~h ÞÞ  f~ h-l ; ð95Þ over the smaller REV is
∂t h h Z
∂ 1
ðϕs hcs is Þ ¼  ns  ð  cs wsv Þ dS: ð109Þ
∂ ∂t V 1 Ssv
ðϕ ρ~ Þ ¼  ∇  ðϕh ðρ~ h v~ h þ i~ h ÞÞ  f~ h-l :
w
ð96Þ
∂t h h
Comparing this averaged form with the mesoscale form in (94) it
can be seen that
B.1.3.2. v-phase.
cns ¼ hcs is ; ð110Þ
∂
ðϕ ϕ c~v Þ ¼  ∇  ðϕl ϕv ðc~ v v~ v þ j~ v ÞÞ  f~ v-h  f~ v-s ; ð97Þ
∂t l v vns ¼ 0; ð111Þ
∂ Z
ðϕ ϕ ρ~ Þ ¼ ∇  ðϕl ϕv ðρ~ v v~ v þ i~v ÞÞ  f~ l-h :
w
ð98Þ n 1
∂t l v v f s-v ¼ ns  ð  cs wsv Þ dS: ð112Þ
V1 Ssv

B.1.3.3. s-phase.
B.3. Upscaling to macroscale
∂
ðϕ ϕ c~s Þ ¼  ∇  ðϕl ϕs j~s Þ  f~ s-h  f~ s-v : ð99Þ
∂t l s B.3.1. Conservation of coffee in h-phase
The averaged form of the microscopic point balance Eq. (87)
f~ α-β is transfer of coffee solubles from α-phase to β-phase over the larger REV is
across αβ interface. Similarly f~ α-β is transfer of liquid from α-
w
 
∂
h
h
h D ○ ○ Eh
h
phase to β-phase across αβ interface. ðϕh c h Þ ¼  ∇  ϕh c h vh þ ch vh þ jh
∂t
Z
1
B.2. Upscaling in l-phase: microscale to mesoscale  nh  ðch ðvh  whl Þ þ jh Þ dS: ð113Þ
V 0 Shl

B.2.1. Conservation of coffee in v-phase Comparing this averaged form with the macroscale form in (95) it
The averaged form of the microscopic point balance Eq. (89) can be seen that
over the smaller REV is
h
D ○ ○ Ev
 c~ h ¼ ch ; ð114Þ
∂
ðϕv hcv iv Þ ¼  ∇  ϕv ðhcv iv hvv iv þ cv vv þ jv Þ
v
∂t
h
Z v~ h ¼ vh ; ð115Þ
1 
 nv  cv ðvv  wvs Þ þ jv dS: ð100Þ
V 1 Svs D ○ ○ Eh

j~ h ¼ ch vh þ jh ;
h
ð116Þ
Now the surface Svs is a material surface so vv  wvs ¼ 0.
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 229

Z Z
1 ϕv
f~ h-l ¼ nh  ðch ðvh  whl Þ þ jh Þ dS: ð117Þ  nl  ðρnv ðvnv wlh Þ þ iv Þ dS:
n
ð129Þ
V0 Shl V0 Slh

Here Shl is the effective surface between the h-phase and Comparing this averaged form with the macroscale form in (98) it
the l-phase and includes boundaries between the h-phase can be seen that
and the v-phase and the h-phase and the s-phase.

 D
 El
ρ~ h ¼ ρnv l ¼ ρv v ; ð130Þ
B.3.2. Conservation of liquid in h-phase
The averaged form of the microscopic point balance Eq. (88)
l
l
v~ v ¼ vnv ¼ hvv iv ; ð131Þ
over the larger REV is
 
○ ○ l
∂
h
h
h D ○ ○ Eh
n l
ðϕh ρh Þ ¼  ∇  ϕh ρh v h þ ρh v h i~ v ¼ iv þ ρnv vnv
∂t
Z
1 D
 El
D ○ ○ Ev l
○ ○ l
 nh  ðρh ðvh  whl ÞÞ dS: ð118Þ
V 0 Shl ¼ iv
v
þ ρv v v þ ρnv vnv ; ð132Þ

Comparing this averaged form with the macroscale form in (96) it Z

ϕ
v
v
f~ l-h ¼ v
can be seen that w
nl  ð ρv ðhvv iv  wlh Þ þ iv Þ dS: ð133Þ

 V0 Slh
ρ~ h ¼ ρh h ; ð119Þ

h
v~ h ¼ vh ; ð120Þ B.3.5. Conservation of coffee in s-phase
The averaged form of the microscopic point balance Eq. (94)
D ○ ○ Eh over the larger REV is
i~h ¼ ρh vh ; ð121Þ
∂
l
n l
Z ðϕ ϕ c n Þ ¼  ∇  ϕl ϕs ð j s Þ
1 ∂t l s s
f~ h-l ¼
w
nh  ðρh ðvh  whl ÞÞ dS: ð122Þ Z
V0 ϕ n
Shl  s nl  ðcns ð  wlh Þ þ js Þ dS
V 0 Slh

n l
 ϕl f s-v : ð134Þ
B.3.3. Conservation of coffee in v-phase
The averaged form of the microscopic point balance Eq. (92) Comparing this averaged form with the macroscale form in (99) it
over the larger REV is can be seen that

○ ○ l !
∂
l
l
l
n l
l
l
ðϕl ϕv cnv Þ ¼  ∇  ϕl ϕv cnv vnv þ cnv vnv þ jv c~ s ¼ cns ¼ hcs is ; ð135Þ
∂t
Z
ϕ n
v~ s ¼ 0; ð136Þ
 v nl  ðcnv ðvnv  wlh Þ þ jv Þ dS
V 0 Slh

n l
n l D
s El
 ϕl f v-s : ð123Þ j~s ¼ js ¼ js ¼ 0; ð137Þ

Comparing this averaged form with the macroscale form in (97) it Z

ϕ
can be seen that f~ s-h ¼ s nl  ðhcs is ð  wlh ÞÞ dS; ð138Þ
V0 Slh

l
l
c~ h ¼ cnv ¼ hcv iv ; ð124Þ

n l

l
l f~ s-v ¼ ϕl f s-v
v~ v ¼ vnv ¼ hvv iv ; ð125Þ Z  Z 
1 1
¼ ns  ð  cs wlh ÞdS dV 0l : ð139Þ

○ ○ l V 0 V 0l V 1 Ssv

n l
j~v ¼ jv þ cnv vnv

D
 El
D ○ ○ Ev l
○ ○ l Appendix C. Macroscopic diffusion and dispersion fluxes
v
¼ jv þ cv v v þ cnv vnv ; ð126Þ
The total macroscopic flux, j~h is made up of the macroscopic
Z
ϕv
v average of molecular diffusion and the dispersive flux:
f~ v-h ¼ nl  ðhcv iv ðhvv iv  wlh Þ þ jv Þ dS; ð127Þ
V0 Slh
h D ○ ○ Eh
j~ h ¼ jh þ ch vh : ð140Þ
Z  Z 
n
n l 1 1
f~ v-s ¼ ϕl f v-s ¼ nv  jv dS dV 0l : ð128Þ The microscopic diffusive flux can be represented by Fick's Law:
V0 V0 l
V1 Svs

It will be later assumed that v~ v ¼ 0 and j~v ¼ 0 but of course it will jh ¼  D∇ch ; ð141Þ
n
still be possible to have vnv a 0 and jv a 0. where D is the diffusion coefficient of the species in water. The
macroscopic equivalent is obtained by averaging this expression
B.3.4. Conservation of liquid in v-phase and will generally depend on the structure of the porous medium.
The averaged form of the microscopic point balance Eq. (93) This average is represented by
over the larger REV is
jh ¼  DT~ ðϕÞ  ∇c~ h ¼  D
~ ðϕÞ  ∇c~ :
a a a
! h ð142Þ

○ ○ l
∂
l
l
l
n l
where the ~ here on T~ means this is a tensor of rank two which
a
ðϕl ϕv ρnv Þ ¼  ∇  ϕl ϕv ρnv vnv þ ρnv vnv þ iv
∂t
represents the tortuosity of the porous medium. For an isotropic
230 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

porous medium this can be adjusted to source or sink on the interphase boundary, i.e., there is no jump in
D the normal flux of the considered species across the boundary so
jh ¼  ∇c~ h ;
a
ð143Þ that f α-β ¼  f β-α . In this instance it is assumed that a chemical
τ
species can reach the microscopic interphase boundary by two
where τ is the tortuosity defined by modes of transport, namely advection and diffusion. Hence, as we
Le actual path length have already seen from the averaging procedure in Appendices
τ¼ ¼ : ð144Þ
L macroscopic path length Appendix A and Appendix B the strength of a source of a
considered species in the α-phase is given by
The tortuosity must be estimated in terms of the porosity. Various
Z
estimated are used in the literature. Some of these include 1
f α-β ¼ nα  ðcα ðvα  wαβ Þ þ jα Þ dS; ð151Þ
1 U 0 Sαβ
τ ¼ ϕ  1=3 ; τ ¼ ϕ  1=2 ; τ¼ : ð145Þ
1  αð1  ϕÞ where U0 is the volume of the REV, Sαβ is the boundary between
In the final expression α ¼ ðr σ Þ=V is a shape factor with r being the the α-phase and all other phases, nα is the unit outward normal
object radius, σ being the object cross sectional area and V being vector on this surface and wαβ is the velocity of points on the
the object volume. Thus for spheres for example α ¼ 0:75. In this interphase boundary. If Sαβ is a material surface as will generally
case the first expression from Millington (1959) will be adopted. be the case here than vα  wαβ ¼ 0 and
The other expressions and tortuosity in general are discussed in Z
1
Pisani (2011). Thus macroscopic diffusion is approximated by f α-β ¼ nα  jα dS: ð152Þ
U 0 Sαβ
D
jh ¼  ∇c~ h ¼  ϕh D∇c~ h :
a 1=3
ð146Þ In this case the chemical species crosses the interphase boundary
τ
by diffusion only. In the coffee bed model developed here non-
Dispersion occurs due to variations in the microscopic velocity equilibrium fluid–fluid and mass–fluid transfers will be consid-
of the phase with respect to the averaged velocity, and molecular ered. The first mechanism considered that drives transfer (in an
diffusion (Bear and Cheng, 2010). Molecular diffusion contributes effort to bring the system closer to equilibrium) is the difference in
to the dispersive flux in addition to the diffusive flux at the concentrations (or more rigorously the difference in chemical
macroscopic level. In general the dispersive flux is given by potentials) at the interface, visualised as a thin film. Therefore
D ○ ○ Eh the rate of transfer, f α-β , of the mass of the considered species
b
jh ¼ ch vh ¼  D ~ b  ∇c~ h ; ð147Þ from an α-phase to an adjacent β-phase across an interface Sαβ is
often assumed to be proportional to the difference in concentra-
where D~ is a rank 2 tensor called the dispersion tensor. D
~ b is both
b

tion between the phases. Thus

positive definite and symmetric. One commonly used expression is
vk vl f α-β ¼ ααβ ðcβ  cα Þ: ð153Þ
Dij ¼ aijkl ; ð148Þ
v To estimate the mass transfer coefficient ααβ the following form is
where aijkl is a fourth order tensor and vi ¼ hvi ih is the average used:
velocity in the i-th direction and v ¼ j vj where v is the average
Dα ðcβ  cα Þ
velocity vector in this instance. For an isotropic porous medium jα  nα ¼  : ð154Þ
this expression reduces to
Δα
 vi vj  Here Dα is the coefficient of diffusion of the considered species in
Dij ¼ aT δij þ ðaL aT Þ 2 v: ð149Þ the α-phase and Δα is the length characterising the mean size of
v
the phase or the length over which diffusion occurs. For example
The coefficients aL and aT here are the longitudinal and transverse
one possible definition of Δα is Δα ¼ U 0α =Sαβ , the volume to
dispersivities of the porous medium. For a phase that completely
surface ratio of the α-phase within the REV. Then
fills a pore space, aL is a length that should be of the same order of Z
the pore size. δij is the Kronecker delta. Also it is required that f α-β ¼
1
nα  jα dS
U 0 Sαβ
aL Z 0 aT Z 0: ð150Þ Z
1 Dα ðcβ  cα Þ
Laboratory experiments have found that aT is 8–24 times smaller ¼ dS
U 0 Sαβ Δα
than aL (Bear and Cheng, 2010).  
Sαβ Dα ðcβ  cα Þ
¼
U0 Δα
Appendix D. Method of estimating mass transfer terms Snαβ
¼ ϕα Dα ðcβ  cα Þ; ð155Þ
Δα
It is necessary to estimate the mass transfer terms f α-β which
govern the transfer of solute from the solid phase to the liquid where Snαβ ¼ Sαβ =U 0α is the specific surface area of the α-phase or
phase both within the grains and from the surface of the grains. It the surface area per unit volume. Thus it can be seen that the mass
is also necessary to estimate the mass transfer of solute from liquid transfer coefficient is given by
within the grains to liquid in the pores between the grains. This
Snαβ
subject is dealt with from a food processing and engineering ααβ ¼  ϕα Dα : ð156Þ
standpoint in Aguilera and Stanley (1999). More general and
Δα
technical developments are found in Bear and Cheng (2010) and The above derivation is for a fluid–fluid transfer. For a solid–fluid
Cussler (1997). The transfer term f α-β may be due to a number of transfer a slightly different approach is used.
processes. Some typical examples are adsorption (from the liquid Under consideration here is the case where the solid matrix
phase to the solid), evaporation or volatilization (i.e. a liquid–gas itself is dissolving (i.e. the soluble part of the coffee grains). It is
transfer), dissolution (i.e. solid–liquid transfer), and liquid–liquid again assumed that the porous medium is saturated and a single
transfer. It is possible that a number of these transfers occur constituent is considered. It is then assumed that there is a thin
simultaneously so that the transfer term comprises a number of layer of liquid next to the solid which is always saturated with the
different transfer processes. It is assumed here that there is no solute under consideration. This concentration is denoted by csat .
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 231

csat is the concentration in the liquid phase that would be in Snhl

¼ ϕl ϕv Dv ðc~ h  c~ v Þ
equilibrium with the concentration inside the solid cs. This Δl τ
assumes that the dissolution process occurs faster than the 4 Snhl
transfer from this thin layer to the bulk of the fluid. It is now ¼ ϕl ϕ3v Dv ðc~ h  c~ v Þ; ð165Þ
Δl
assumed that the force of extraction from this thin layer to the
 1=3
bulk of the fluid is given by where τ ¼ ϕv has been used. Next transfer of solute from the
solid on the surface of the grains, f~ s-h is considered. Recalling
f s-f ¼ αsf ðcsat  cf Þ: ð157Þ
(138) and (157) and that diffusion of solute into the bulk of the
Proceeding as above it can be shown that the mass transfer fluid is the rate limiting step it follows that
coefficient from a solid phase s to a fluid phase f is given by
f~ s-h ¼ αsh ðcsat  c~ h Þ
Snsf
αsf ¼  ϕs Df : ð158Þ Snhl
Δs ¼ ϕl ϕs Dh ðcsat  c~ h Þ
Δs
The transfers looked at so far have been transfer of constituent Snhl
or solute due to diffusion. It is also possible to have a transfer of ¼ ϕl ð1  ϕv ÞDh ðcsat  c~ h Þ: ð166Þ
Δs
liquid from the large pores between the grains to the small pores wn
within the grains (or vice versa). This will occur due to a pressure As outlined above the third transfer term, f~ l-h , arises due to
imbalance between the phases due to the dissolution of the solid solute being carried in the fluid that transfers between phases due
matrix within the grains. As was seen from the averaging proce- to pressure differences between pores this is given as above by
dure this term has the form 8
<  ϕl Snlh ρkv ðμϕv Þ ðp h Δ p v Þc~ h if p~ h Z p~ v
~ ~
Z wn
ϕ 

v  f~ l-h ¼
l
ð167Þ
f~ l-h ¼ v
w
nl  ρ ðhvv iv  wlh Þ þ iv
v
dS: ð159Þ : ϕl Snlh ρkv ðϕv Þ ðp~ h  p~ v Þc~ v if p~ h o p~ v
V 0 Slh μ Δ l

v
Now it is assumed that iv ¼ 0 and Darcy's Law gives The macroscopic transfer term from the s-phase to the v-phase
can be arrived at using (139) and (157) to get
kv ðϕv Þ
hvv iv  wlh ¼  ∇p ð160Þ
ϕv μ f~ s-v ¼ αsv ðcsat  c~ v Þ
On Slh this means that Snsv
¼ ϕl ϕs Dv ðcsat  c~ h Þ: ð168Þ
Δs
v kv ðϕv Þ ðph  pv Þ
hvv i  wlh ¼  : ð161Þ
ϕv μ Δl
Thus
Z   Appendix E. Dominance of advection over mechanical
ϕ kv ðϕv Þ ðp~ h  p~ v Þ
f~ l-h ¼ v
w
nl  ρ dS dispersion and diffusion
V0 Slh ϕv μ Δl
S kv ðϕv Þ ðp~ h  p~ v Þ Considering the dimensional equations from the paper we can
¼  lh ρ
V0 μ Δl consider the relative importance of the coffee transport processes
k ð ϕ Þ ð ~ h  p~ v Þ
p in the intergranular pores by comparing their magnitudes. This
¼  ϕl Snlh ρ
v v
: ð162Þ
μ Δl should give us an idea of the dominant transport mechanism in
the bed, although of course there may be narrow regions where
Again here Slh is the specific surface area or surface area per unit other balances hold. Firstly we compare advection and dispersion.
volume of the l-phase. Since there is a difference in concentration The ratio of the magnitudes of the terms is
in the h and v-phases the transfer of fluid in either direction will  
c~ h v~ h 
also result in the transfer of solute from one phase to another. This  b : ð169Þ
~ 
transfer will be of the following form: D h ∇c~ h 

8 Recall that
<  ϕl Snlh ρkv ðμϕv Þ ðp h Δ p v Þc~ h
~ ~ !
if p~ h Z p~ v
wn
f l-h ¼
l
ð163Þ vi vj   vi vj
ðD~ h Þij ¼
b
: ϕl S ρ n kv ðϕv Þ ðp~ h  p~ v Þ
c~ v if p~ h o p~ v aT δij þ ðaL  aT Þ 2 vh  ¼ aijkl  : ð170Þ
lh μ Δl vh  vh

We now use some characteristic   scales. Note aL  ll r500 μm.

D.1. Form of individual mass transfers Thus take aijkl  aL , c~ h  C, v~ h   vc , vi  vc , and z  L. Thus (169)
becomes
It is now possible to describe the source and sink terms in the
Cvc L
macroscopic equations arising from mass transfers in terms of the
ll vc C
¼  102 : ð171Þ
ll
microscopic quantities of the system. Now the total mass transfer L
from the l-phase to the h-phase can be written as
Thus advection is approximately one hundred times larger than
f~ l-h ¼ f~ v-h þ f~ s-h þ f l-h :
wn dispersion and so dominates. More generally advection dominates
ð164Þ
over dispersion when L⪢ll unless there are very large concentra-
f~ v-h represents transfer of solute between the pores within the tion gradients in the bed.
grains and the pores between the grains due to diffusion. Now The ratio of advection to diffusion is given by
considering (127) and (153) it follows that  
c~ h v~ h 
f~ v-h ¼ αvh ðc~ h  c~ v Þ  : ð172Þ
 4=3 
ϕh Dh ∇c~ h 
Snhl
¼ ϕl ϕv Dnv ðc~ h  c~ v Þ
Δl Adopting similar approximations as in the dispersion case we find
232 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

that grinds were obtained by grinding Illy coffee beans using a Cimbali
Cvc Lvc burr grinder. One very coarse grind was obtained using the #20
¼ : ð173Þ setting on the grinder. A second extremely coarse grind was
ϕ4=3
h
Dh C ϕ4=3
h Dh obtained using the #30 setting on the grinder. The grind size
L

Now from the experiments typical approximate values for these distribution of the Cimbali #30 grind was too coarse to be
quantities are L  0:05 m, vc  0:007 ms  1 , ϕh  0:2 and analysed by the optical particle size analyzer used (Mastersizer
Dh  2:2e 9 m2 s  1 . Thus 2000; Malvern Instruments Ltd, UK) and so is not included in
Fig. 9.
Lvc
 106 : ð174Þ
ϕh4=3 Dh
G.2. Coffee extraction kinetics during batch-wise brewing in a fixed
These estimates show that unless there are extremely large water volume
concentration gradients somewhere within the bed that advection
dominates over diffusion. The extraction kinetics of the five coffee grinds were investi-
gated by mixing 60 g of coffee grounds with a hot water volume,
Appendix F. Extraction kinetics of coffee components V water ¼ 0:5 l, and measuring the concentration cbrew of extracted
species as a function of time. The temperature of the liquid during
extraction is 80–90 1C. The experimental procedure is identical to
that outlined in the paper but the results for the other grinds are
shown here in Fig. 10. This experiment clearly illustrates the key
influence that the grind size distribution has on extraction.

G.3. Coffee extraction profiles from a cylindrical brewing chamber

under different conditions

In this paper we have presented coffee extraction profiles from

Fig. 8. Extraction experiments suggest that a large number of compounds found in
coffee extract with similar kinetics (Booth et al., 2012).
Appendix G. Supplementary coffee extraction experiments
The experimental results used in this paper are drawn from a
much larger collection of coffee extraction experiments. To com-
plement and support these results and satisfy the interested
reader some other relevant experiments are included in this
section.
G.1. Coffee grinds used
In the experiments presented here we make use of five
different coffee grinds ranging from a fine drip filter grind to a
very coarse grind. The grind size distributions of these grinds are
shown in Fig. 9. The first grind is a relatively fine grind, called
Jacobs Krönung (JK) standard drip filter coffee grind. The next
grinds used are the Douwe Egberts (DE) standard drip filter grind
a cylindrical brewing chamber for one fine grind (JK standard drip
filter grind) and one coarse grind (Cimbali #20 grind). Here we
present some ancillary experiments for JK standard drip filter
grind for a different coffee bed mass and for a different value of
absolute pressure in the coffee bed. The experimental apparatus is
the same as that outlined in the paper. To compare results for
different masses (and hence different bed lengths) the extraction
is performed for coffee bed masses of 12.5 g and 60 g. These
masses correspond to bed depths of 1.12 cm and 4.05 cm respec-
tively. The flow rate to the coffee bed is 250 ml/min in both cases.
The pressure difference across the bed is measured in both cases.
The solubles concentrations are measured in the coffee pot and at
the filter exit. The results are shown in Fig. 11.
The influence of absolute pressure on extraction was also
investigated by repeating experiments in the coffee brewing
cylinder at different values of absolute pressure but maintaining
the same coffee bed mass and flow rate. In the case of JK standard
drip filter grind the absolute pressure in the brewing cylinder is
increased from 2.3 bar to 9 bar. The resulting solubles concentra-
tions profiles are plotted in Fig. 12. The results are seen to be
virtually identical which indicates that the extraction kinetics are
substantially independent of the absolute pressure (at least for the
range of values considered) and that the results are reproducible
40

and the Douwe Egberts coarse drip filter grind. Finally two further
30
10
cbrew kg m3

JK standard drip filter grind

8 20 JK standard drip filter grind
Illy, Cimbali 20 grind
Illy, Cimbali 20 grind
Volume fraction

DE standard drip filter grind

6 Illy, Cimbali 30 grind
DE coarse drip filter grind 10
DE standard drip filter grind

4 DE coarse drip filter grind

0
2 0 100 200 300 400 500 600
time s
0
0.1 1 10 100 1000 104 Fig. 10. Coffee solubles concentration profiles for different coffees and grind size
distributions during batch extraction experiments. In these experiments 60 g of
Particle size µm
coffee with approximately 4% moisture was mixed with 0.5 l of hot water in a
Fig. 9. Grind size distributions of the coffee grinds used in experiments. French press type cylinder.
 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234 233

200 JK, 60g, 2.3bar, 250ml min 200 JK, 60g, 2.3bar, 250ml min

cbrew mg gram

cexit mg gram
JK, 12.5g, 0.5bar, 250ml min JK, 12.5g, 0.5bar, 250ml min
150 150

100 100

50 50

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Mbrew grams Mbrew grams
Fig. 11. The coffee solubles concentration, measured in mg/gram, is plotted against mass of coffee beverage M brew (grams) for JK drip filter grind with a flow rate of 250 ml/
min in (a) the coffee pot and (b) the beverage at filter exit for different coffee bed masses.

200 JK, 60g, 2.3bar, 250ml min 200 JK, 60g, 2.3bar, 250ml min
cbrew mg gram

cexit mg gram
150 JK, 60g, 9.0bar, 250ml min 150 JK, 60g, 9.0bar, 250ml min

100 100

50 50

0 0
0 200 400 600 800 1000 0 200 400 600 800 1000
Mbrew grams Mbrew grams
Fig. 12. The coffee solubles concentration, measured in mg/gram, is plotted against mass of coffee beverage M brew (grams) for JK drip filter grind with a flow rate of 250 ml/
min in (a) the coffee pot and (b) the beverage at filter exit for different values of absolute pressure in the brewing cylinder.

to within a few percent. It also suggests that a possibly faster
particle penetration with water at higher pressures does not
substantially affect the observed extraction results.
References
Aguilera, J., Stanley, D., 1999. Microstructural Principles of Food Processing and
Engineering, A Chapman & Hall Food Science Book. Springer, USA, URL 〈http://
books.google.ie/books?id=nIeJiL_dLeQC〉.
Aubertin, M., Chapuis, R., 2003. École polytechnique de Montréal, Predicting the
Coefficient of Permeability of Soils Using the Kozeny-Carman Equation, Rapport
Technique. 〈http://books.google.ie/books?id=yk8nNAEACAAJ〉.
Bear, J., Cheng, A.H.-D., 2010. Modeling groundwater flow and contaminant
transport, theory and applications of transport in porous media. Springer,
Heidelberg.
Booth, C., Cummins, C., Dalwadi, M., Dellar, P., Devereux, M., Dewynne, J., Donohue,
J., Duncan, A., Fitzmaurice, F., Gordon, A., Hennessy, M., Hinch, J., Hickey, C.,
Hjorth, P., Kyrke-Smith, T., Leahy, D., Lee, W., Lynch, E., Mercier, O., Miklavcic, S.,
Russell, S., Schwartz, L., Shozi, B., Swierczynski, P., Timoney, C., Tomczyk, J.,
Warneford, E, 2012. In: Technical Report from MACSI's 2012 Problem-Solving
Workshop with Industry. 〈http://www.macsi.ul.ie/esgi87/ReportWeb.pdf〉.
Clarke, R., 1987. Extraction. In: Clarke, R., Macrae, R. (Eds.), Coffee, Springer, The
Netherlands, pp. 109–145. http://dx.doi.org/10.1007/978-94-009-3417-7_5.
Cussler, E., 1997. Diffusion: Mass Transfer in Fluid Systems, Cambridge Series in
Chemical Engineering. Cambridge University Press, Cambridge UK 〈http://
books.google.ie/books?id=TGRmfTrsPTQC〉.
Fasano, A., Farina, A., 2010. Modelling complex flows in porous media by means of
upscaling procedures. Rend. Istit. Mat. Univ. Trieste 42, 65–102.
Fasano, A., 2000. Filtration problems in various industrial processes. In: Fasano, A.
(Ed.), Filtration in 906 Porous Media and Industrial Application, vol. 1734 of
Lecture Notes in Mathematics, Springer 907 Berlin, Heidelberg pp. 79–126.
http://dx.doi.org/10.1007/BFb0103976.
Fasano, A., Talamucci, F., 2000. A comprehensive mathematical model for a multi-
species flow through ground coffee. SIAM J. Math. Anal. 31 (2), 251–273. http:
//dx.doi.org/10.1137/S0036141098336698.
Fasano, A., Talamucci, F., Petracco, M., 2000. The espresso coffee problem. In:
Fasano, A. (Ed.), Complex Flows in Industrial Processes, Modeling and Simula-
tion in Science, Engineering and Technology. Birkhuser, Boston,
pp. 241–280. http://dx.doi.org/10.1007/978-1-4612-1348-2_8.
Gianino, C., 2007. Experimental analysis of the Italian coffee pot Moka. Am. J. Phys.
75 (1), 43–47. http://dx.doi.org/10.1119/1.2358157.
Goto, M., Roy, B.C., Hirose, T., 1996. Shrinking-core leaching model for supercritical-
fluid extraction. J. Supercrit. Fluids 9 (2), 128–133. http://dx.doi.org/10.1016/
S0896-8446(96)90009-1, URL 〈http://www.sciencedirect.com/science/article/
pii/S0896844696900091〉.
Gray, W., Hassanizadeh, S., 1998. Macroscale continuum mechanics for multiphase
porous-media flow including phases, interfaces, common lines and common
points. Adv. Water Resour. 21 (4), 261–281.
Holdich, R., 2002. Fundamentals of Particle Technology, Midland Information
Technology and Publishing, Leistershire, UK. URL 〈http://books.google.ie/
books?id=BiTKAAAACAAJ〉.
Huang, Z., Shi, X. han, Jiang, W. juan, 2012. Theoretical models for supercritical fluid
extraction. J. Chromatogr. A 1250 (0), 2–26. http://dx.doi.org/10.1016/j.
chroma.2012.04.032, supercritical Fluid Extraction and Chromatography URL
〈http://www.sciencedirect.com/science/article/pii/S0021967312005894〉.
Jaganyi, D., Madlala, S.P., 2000. Kinetics of coffee infusion: a comparative study on
the extraction kinetics of mineral ions and caffeine from several types of
medium roasted coffees. J. Sci. Food Agric. 80 (1), 85–90.
Machmudah, S., Martin, A., Sasaki, M., Goto, M., 2012. Mathematical modeling for
simultaneous extraction and fractionation process of coffee beans with super-
critical CO2 and water. J. Supercrit. Fluids 66 (0), 111–119. http://dx.doi.org/
10.1016/j.supflu.2011.11.011, special Edition on the Occasion of Gerd Brunner's
70th Birthday. URL 〈http://www.sciencedirect.com/science/article/pii/
S0896844611004815〉.
Millington, R.J., 1959. Gas diffusion in porous media. Science 130 (3367), 100–102.
http://dx.doi.org/10.1126/science.130.3367.100-a, URL 〈http://www.sciencemag.
org/content/130/3367/100.2.abstract〉arXiv:130/3367/100.2.full.pdf.
Navarini, L., Nobile, E., Pinto, F., Scheri, A., Suggi-Liverani, F., 2009. Experimental
investigation of steam pressure coffee extraction in a stove-top coffee maker.
Appl. Therm. Eng. 29 (56), 998–1004. http://dx.doi.org/10.1016/j.appltherma-
leng.2008.05.014, URL 〈http://www.sciencedirect.com/science/article/pii/
S1359431108002299〉.
Oliveira, E.L., Silvestre, A.J., Silva, C.M., 2011. Review of kinetic models for super-
critical fluid extraction. Chem. Eng. Res. Des. 89 (7), 1104–1117. http://dx.doi.
org/10.1016/j.cherd.2010.10.025, URL 〈http://www.sciencedirect.com/science/
article/pii/S0263876210003278〉.
Petracco, M., 2008. Technology IV Beverage Preparation: Brewing Trends for the
New Millennium. Blackwell Science Ltd., Oxford UK, pp. 140–164.
Pictet, G., 1987. Home and catering brewing of coffee. In: Clarke, R.,
Macrae, R. (Eds.), Coffee. Springer, The Netherlands, pp. 221–256
http://dx.doi.org/10.1007/978-94-009-3417-7_8.
Pisani, L., 2011. Simple expression for the tortuosity of porous media. Transp.
Porous Media 88 (2), 193–203. http://dx.doi.org/10.1007/s11242-011-9734-9.
Rao, S., 2010. Everything But Expresso, Independent Publisher Scott Rao, Available
from www.scottrao.com.
234 K.M. Moroney et al. / Chemical Engineering Science 137 (2015) 216–234

Sivetz, M., Foote, H., 1963. Coffee processing technology, no. v. 2 in Coffee
Processing Technology, Avi Pub. Co. URL 〈http://books.google.ie/books?
id=nbBTAAAAMAAJ〉.
Sovov, H., 2005. Mathematical model for supercritical fluid extraction of natural
products and extraction curve evaluation. J. Supercrit. Fluids 33 (1), 35–52.
http://dx.doi.org/10.1016/j.supflu.2004.03.005, URL 〈http://www.sciencedirect.
com/science/article/pii/S0896844604000932.
Spaninks, J.A.M., 1979. Design procedures for solid–liquid extractors and the effect
of hydrodynamic instabilities on extractor performance (Ph.D. thesis), Agricul-
tural University of Washington.
Voilley, A., Simatos, D., 1979. Modeling the solubilization process during coffee
brewing. J. Food Process Eng. 3 (4), 185–198.