Journal of Food Engineering 192 (2017) 36e44

Contents lists available at ScienceDirect

Journal of Food Engineering

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

Analysis of mass transfer equations during solid-liquid extraction and

its application for vanilla extraction kinetics modeling

pez b, G.C. Rodríguez-Jimenes a, J. Carrillo-Ahumada c,
K. Castillo-Santos a, I.I. Ruiz-Lo
M.A. García-Alvarado a, *
a
Chemical and Biochemical Engineering Department, Instituto Tecnolo
gico de Veracruz, Av. Miguel A. de Quevedo 2779, Veracruz, Ver., Mexico
b
Chemical Engineering Faculty, Benem

noma de Puebla, Av. San Claudio y 18 Sur, Puebla, Pue., Mexico
erita Universidad Auto
c
Institute of Applied Chemistry, Universidad del Papaloapan, Circuito Central 200, colonia Parque Industrial, Tuxtepec, Oax., Mexico

a r t i c l e i n f o a b s t r a c t

Article history: Rigorous dimensionless analysis and analytical solution of mass transfer equation (punctual and space
Received 4 March 2016 averaged) governing the solid-liquid extraction dynamic were performed. Analytical solutions result in
Received in revised form continuous and analytical maps between solute mass fractions in extract with respect to time which
6 June 2016
were used both for solute diffusivity estimation and process modeling. In order to validate the models,
Accepted 30 July 2016
Available online 1 August 2016
experimental extraction at equilibrium of 1 cm long vanilla beans pieces were obtained at 0.1, 0.2, 0.3 and
0.4 g vanilla/mL solvent (60% w/w ethanol-water) and 30, 40 and 50
C; and extraction kinetics were
developed at 0.1 g/mL and 30, 40 and 50
C. Apparent equilibrium constant and solute effective diffu-
Keywords:
Solid-liquid extraction
sivity estimation were performed by linear regression on experimental results. Apparent equilibrium
Mass transfer constant was 0.86, and diffusivities were 1.22, 1.98, and 2.43  1011 m2/s at 30, 40 and 50
C respec-
Dimensionless analysis tively. Diffusivities and analytical solutions predicted the experimental results at 30 and 50
C with 0.3,
Analytical solution and 0.4 g/mL respectively.
Diffusivity © 2016 Elsevier Ltd. All rights reserved.
Vanilla extract

1. Introduction during solid-liquid extraction (Simeonov et al., 1999; Amendola

Solid-liquid extraction operation has recently returned to in-
terest due to the extraction of bioactive compounds from natural
sources (Cacace and Mazza, 2003; Pinelo et al., 2006; Spigno and De
Faveri, 2007; Amendola et al., 2010; Linares et al., 2010) and
therefore the mechanistic (mass transfer based) mathematical
modeling of the process has been revised. Mass transfer during
solid-liquid extraction may be conceptualized as solute diffusion in
a media (underflow) contacting a well-stirred solution of finite
volume (extract). The analytical solution to this problem in the
diffusion media was reported by Crank (1975) for negligible inter-
facial resistance and by Mikhailov (1977) for a general case. In both
cases, the solution for concentration in the well-stirred finite vol-
ume solution was not explicitly solved, and the volume of both
phases was assumed as constant during process. Independently of
the reported analytical solutions (Crank, 1975; Mikhailov, 1977), it
has been a common practice the use of empirical models to
describe the concentration kinetic of extractable solutes in extract
* Corresponding author.
E-mail address: miguelg@itver.edu.mx (M.A. García-Alvarado).
et al., 2010; Linares et al., 2010; Rakotondramasy-Rabesiaka et al.,
2010; Chilev et al., 2014); and, the extractable solutes effective
diffusivity has been estimated by different methods: with the
concentration at a punctual given time (Seikova et al., 2004;
Rakotondramasy-Rabesiaka et al., 2010) which is not convenient
because the experimental error is only the punctual error obtained
at one time of the whole kinetic; by regression fit of the analytical
solutions of diffusion equation considering infinite volume of sol-
vent (Pinelo et al., 2006; Franco et al., 2007); by non-linear
regression directly of differential equations (Garcia-Perez et al.,
2010); or by an equation valid at very short times (t < 0.0189)
(Cacace and Mazza, 2003). An alternative approach, for solid-liquid
extraction modeling and/or effective diffusivity estimation, has
been the use of mechanistic models supported in macroscopic
(averaged) mass transfer equations (Veloso et al., 2005; Espinoza-
Pe
rez et al., 2007; Rodríguez-Jimenes et al., 2013). Espinoza-Pe
rez
et al. (2007) deduced an analytical solution of averaged mass
transfer equations, and Rodríguez-Jimenes et al. (2013) applied this
solution during vanilla solid-liquid extraction for the estimation of
aromatic compound effective diffusivity in underflow by non-linear
regression.

http://dx.doi.org/10.1016/j.jfoodeng.2016.07.020
0260-8774/© 2016 Elsevier Ltd. All rights reserved.
 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44 37

Nomenclature f Factor depending of underflow shape Dimensionless

ln Eigenvalues (Eq. (35)) Dimensionless
A Transfer surface m2 r Density kg$m3
D Effective diffusivity m2$s1
kc Mass transfer coefficient m$s1 Subscripts
Keq Apparent distribution constant kg$kg1 e At equilibrium
l Characteristic length m n Number of eigenvalue
m Mass kg i At interface
n Geometric factor Dimensionless b In the underflow
n Unit vector normal to transfer surface Dimensionless g In the extract
t Time s
V Volume m3 Dimensionless groups
x Solute mass fraction kg$kg1 Bi Mass Biot number
〈4〉 Any variable 4 volume averaged a Mass relation
z Rectangular coordinate m J Reduced mass fraction
t Dimensionless time (Fourier number)
Greek symbols x Dimensionless coordinate
d Extraction kinetic slope s1 
V A linear map of gradient operator
ε System porosity m3$m3

Considering the different point of views of solid-liquid extrac-

tion kinetic modeling and diffusivity estimation, it is necessary a xgi ¼ Keq xbi (3)
unified model that represents the process with a theoretical basis. A
Alternatively the phase equilibrium could be represented by a
unified model must be deduced from fundamental principles of
more complex relationship, as the Langmuir relation. However,
mass transfer and thermodynamics, and it must have the property
Nernst's law has been universally assumed (Crank, 1975; Mikhailov,
of predict the phenomenon at different conditions than those used
1977; Seikova et al., 2004; Cacace and Mazza, 2003; Veloso et al.,
to estimate the mass transfer properties. Therefore, in this paper a
2005; Chilev et al., 2014) when the analytical solution of mass
deep analysis of mass transfer equation governing the process dy-
transfer equations are required. Finally, the mass balance of
namic is presented. It was declared as dynamic, because the mass
extractable material in extract is,
transfer equations (local or space-averaged) represent the solid-
liquid extraction kinetics in terms of driven forces. The analysis

was performed for both, local and space-averaged equations and d V g rg xg

¼ kc rg Abg xgi  xg in Vg (4)
included dimensionless analysis, analytical solution and experi- dt
mental validation. Obtained analytical solutions are continuous and
The transfer surface (Abg) in terms of underflow volume (V b) is,
analytical maps that relate extractable solutes mass fraction in both
phases (extract and underflow) with the process time. These maps nV b
were used for the extractable solute effective diffusivity estimation Abg ¼ (5)
l
during vanilla solid-liquid extraction and for the prediction of
experimental extraction kinetics in the same system at different Particular forms of Eq. (5) are: an infinite flat slab with mass
product/solvent ratios. transfer in both sides with l half-thickness and n ¼ 1; an infinite
cylinder with radius l and n ¼ 2; and a sphere with radius l and
n ¼ 3. The solution volume (V g) and medium volume (V b) can be
represented as a fraction of total system volume (V ¼ V g þ V b), and
2. Model development therefore,

2.1. Punctual model V g ¼ εV and V b ¼ ð1  εÞV (6)

Solid-liquid extraction involves the diffusion of extractable The total mass balance of the whole system
material in media (underflow) contacting with a finite volume of a (underflow þ extract) between initial and equilibrium states results
well-stirred solution (extract). The extractable material diffusion in,
within underflow in a general coordinate system is represented by,
mg0 xg0  mge xge

 ¼ 1 (7)
mb0 xb0  mbe xbe
v rb xb

¼ Db V$V rb xb in Vb (1)
vt where underflow and extract masses in terms of their volume are,
The mass transfer at underflow-extract interface in a general
mg ¼ rg V g mb ¼ rb V b (8)
coordinate system is given by,


 It is important to note that Eq. (7) considers that initial under-
n$Db V rb xbi ¼ kc rg xgi  xg in Abg (2) flow (mb0: feed) mass and initial extract (mg0: solvent) mass may be
different of final underflow (mbe: at equilibrium) mass and final
In which the extractable material must be in phase equilibrium extract (mge: at equilibrium) mass respectively, as expected in any
that may be represented by Nernst's law, solid-liquid extraction system due to the solution retained by inert
38 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44
solids in underflow (McCabe et al., 2004; Castillo-Santos et al.,
2016).
Eqs. (1)e(8) are the unsteady mass transfer equations repre-
senting the extractable material kinetics during solid-liquid
extraction process before reaching equilibrium. A dimensionless
analysis and their analytical solution will be developed in results
section.
2.2. Space-averaged model
The same phenomena may be represented in terms of macro-
scopic mass transfer and average extractable concentration in
R R
underflow 〈x〉b ¼ xb dV = dV (Veloso et al., 2005; Espinoza-
Vb Vb

rez et al., 2007; Rodríguez-Jimenes et al., 2013). Macroscopic
Pe
mass balances of extractable solids in underflow and extract during
unsteady state of solid-liquid extraction process are defined by,
d〈x〉b

rb V b ¼ kcb Abg rb xbi  〈x〉b
dt
dxg

rg V g ¼ kcg Abg rg xgi  xg
dt
In which the mass transfer continuity at interface implies,



kcg Abg rg xgi  xg ¼ kcb Abg rb xbi  〈x〉b
The underflow mass transfer coefficient is a function of effective

rez
diffusivity of extractable solids (Veloso et al., 2005; Espinoza-Pe
et al., 2007; Rodríguez-Jimenes et al., 2013) that can be expressed
as,
fDb
kcb ¼
l
The factor (f) of Eq. (12) depends of the underflow geometry
and the analytical solution of local mass transfer equations
assuming interfacial resistance negligible and solution volume
infinite (Rodríguez-Jimenes et al., 2013). Under this assumption the
geometric factor is: f ¼ p2/4 for flat slab with mass transfer in both
sides; f ¼ 2.40482 for infinite cylinder; and, f ¼ p2 for sphere.
Similarly to Eqs. (1)e(12) are the unsteady mass transfer equations
representing the extractable material kinetics in terms of averaged
concentration in underflow during a solid-liquid extraction process
before reaching equilibrium, and therefore their dimensionless
analysis and analytical solution will be also developed in results
section.
3. Experimental procedure
The analytical solutions of Eqs. (1)e(12) are continuous and
analytical maps describing the mass fraction of extractable material
in extract and underflow as time functions. These continuous maps
represent the solid-liquid extraction dynamic (dynamic because
represent the concentration kinetic in terms of driven forces)
before reaching equilibrium, and may be used to estimate effective
diffusivity of extractable solids and to predict the solid-liquid
extraction kinetic. In order to develop the described application,
experimental data of vanilla solid-liquid extraction were obtained
as follows.
3.1. Vanilla samples
Cured vanilla pods were purchased directly to producers from
Papantla, Veracruz, Mexico. A random sample of ca 3 kg was taken
from three vanilla boxes and the pods were cut into 1 cm pieces.
Different random samples were taken from these cut pieces in or-
der to obtain vanilla characterization, equilibrium relation and
extraction kinetics in such way that the results were representative
of the three vanilla boxes.
3.2. Vanilla characterization
Moisture was determined from three random samples of ca 1 g
taken from the cut pods, and placed in a vacuum oven at 60
C and
6  104 Pa until reach constant weight. It is possible that some of
the volatile compounds were lost during this process, however the
principal volatile compound in vanilla is vanillin and its vapor
pressure is 4 magnitude orders lower that water vapor pressure.
Therefore the loss was considered negligible.
The non-extractable solids were evaluated by an exhaustive
extraction of three random samples of ca 2.5 g taken from the cut
pods in 150 mL of solvent (60% w/w ethanolewater) at 50
C by
25 h. This time was previously reported as the required to reach the
(9)
equilibrium concentration in extract (Rodríguez-Jimenes et al.,
2013). The extracted vanilla was extracted again two times more
at the same conditions. After the third extraction the residue was
(10)
put in vacuum oven (60
C and 6  104 Pa) until constant weight.
The remaining solids were considered the non-extractable. The
extractable solids mass fraction was estimated by difference with
moisture and non-extractable solids. Thickness of vanilla pieces (20
(11)
samples) was measured with a Vernier from 20 samples, and the
characteristic length (l) for solute diffusivity was taken as the half
thickness, since the vanilla pieces were fully immersed in solvent.
3.3. Equilibrium relation
Random samples of ca 10 g, 20 g, 30 g and 40 g of 1 cm pieces
(12) vanilla beans were separately extracted (by duplicate) in 100 mL of
solvent (60% w/w ethanolewater) at 30
C, 40
C and 50
C in a
shaker at 300 rpm (1 cm of orbital radius) with controlled tem-
perature bath, and the process was kept until equilibrium was
reached. The extract and underflow were carefully separated and
weighted in order to avoid loss by evaporation. The extract was
evaporated in a rotary evaporator. The dried extract residue over
the extract total mass is the extractable solids mass fraction in
extract. The difference of total extractable solids with respect to
dried residue over the underflow total mass, results in extractable
solids mass fraction in underflow in equilibrium with extract. The
effective equilibrium constants (Keq) were calculated by linear
regression with equilibrium mass fractions at different vanilla/
solvent ratio. The probability of null hypothesis (Ho): Keq ¼ 0 was
calculated with the statistical procedure detailed in our previous
work Rodríguez-Jimenes et al. (2013), which is a t-student test
considering the square sum of modeling errors and the Jacobean
matrix of the variation of each response with respect to each
parameter (in this case one parameter with respect to one
response).
3.4. Extraction kinetics
Random samples of ca 10 g, taken from cut pods, were sepa-
rately extracted with 100 mL of solvent at different times in order to
obtain experimental extraction kinetics at 0.1 g/mL (vanilla/solvent
ratio). At the end of a given time, the samples were analyzed as it
was described in previous section. The kinetics were developed at
30, 40 and 50
C by duplicate. The extractable solids (called solutes
in the rest of manuscript) mass fractions in extract were recorded
and used for the estimation of their effective diffusivity in under-
flow (Db). The probability of null hypothesis (Ho): Db ¼ 0 was
 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44 39

evaluated as described in the above section.

Finally, two additional kinetics were performed by duplicate in 1 f
Jbi ¼  J þ 〈J〉b (22)
order to show the dynamic model capacity to predict extraction a þ fa=Bim g Bim þ f
kinetics at different conditions. Random samples of ca 30 g were
separately extracted with 100 mL of solvent at different times and
dJg d〈J〉b
30
C; and other random samples of ca 40 g were separately ¼ (23)
extracted with 100 mL of solvent at different times and 50
C. dt dt
Mass balance (7) and equilibrium distribution (3) may combined
4. Results for equilibrium mass fraction calculation,

mg0 m
4.1. Dimensionless analysis mbe xg0 þ mbb0e xb0
xbe ¼ ; xge ¼ Keq xbe (24)
1þa
In order to take into account the differences between initial and
equilibrium mass phases explained in Eq. (7), the following The application of Eq. (24) for equilibrium concentrations pre-
dimensionless variables were introduced, diction requires the prediction of underflow (or extract) mass at
equilibrium which depend of retained solution by inert solids in
mbe xb  mbe xbe m x  mge xge underflow (McCabe et al., 2004; Castillo-Santos et al., 2016). In this
Jb ¼ Jg ¼ ge g
mb0 xb0  mbe xbe mg0 xg0  mge xge work the equilibrium data were calculated directly from experi-
(13) mental results.
mbe 〈x〉b  mbe xbe
〈J〉b ¼
mb0 xb0  mbe xbe

where mb0 and mbe are the underflow initial (feed) and equilibrium
4.2. Analytical solution
masses, mg0 and mge are the extract initial (solvent) and equilib-
rium masses. The Biot mass number was defined like suggested by

rdova-Quiroz et al. (1996) in the case of a solute with equilibrium Eqs. (17) and (18) written in terms of a 1D rectangular coordi-
Co
nate (infinite slab) with mass transfer by both sides and l as the
distribution between phases (in this problem Nernst's law),
half-thickness are,
kcg l rg Keq
Bi ¼ (14) vJb v2 Jb
Db rb
¼ for 0  x  1 and t > 0 (25)
vt vx2

Dimensionless gradient operator ( V ) and Fourier mass number
(t) were defined as,
vJb Bi
 ¼ BiJbi þ Jg for x ¼ 1 and t > 0 (26)
 Db t vx a
V ¼ Vl and t¼ (15)
l2
where x ¼ z/l. Eqs. (25) and (26) are symmetric with respect to x and
Crank (1975) and Mikhailov (1977) used the dimensionless therefore vJb/vx ¼ 0 at x ¼ 0.
solvent/product volume ratio. However as Eqs. (1) and (2) are Eqs. (25) and (26) assume,
expressed in mass fractions terms, instead of dimensionless volume
ratio, it is better a dimensionless extract/underflow mass relation at 1 Constant solute effective diffusivity.
equilibrium, 2 Instantaneous mass transfer of retained solution. That is, the
underflow extract mass are equal to mb0 and mg0 only at t ¼ 0,
εrg Keq mge Keq
a¼ ¼ (16) and equal to mbe and mge at t > 0.
ð1  εÞrb mbe 3 Constant underflow particle dimensions.
4 Nernst's law with constant Keq.
Introducing the dimensionless variables and relations into Eqs.
(1)e(12), the following systems were obtained,Local equations,
Taking initial homogeneous mass fraction (Jb ¼ 1 and Jg ¼ 1)
vJb   at t ¼ 0, applying the Laplace transform with respect to t and
¼ V $V Jb in Vb (17) solving, Eq. (25) gives,
vt
pffiffi  pffiffi  1

n$V Jbi ¼ BiJbi þ
Bi
Jg in Abg (18)
Jb ðs; xÞ ¼ C1 cosh s x þ C2 sinh s x þ (27)
s
a
where Jb(s,x) is the Laplace transform of Jb(t,x). By symmetry of
1 dJg Bi Eq. (25) C2 ¼ 0 From Eqs. (18) and (19) it is evident that,
 ¼ BiJbi þ Jg in Vg (19)
n dt a
vJg vJb
Space-averaged equations,  ¼ in x¼1 (28)
vt vx
d〈J〉b

¼ nf Jbi  〈J〉b (20) And therefore the Laplace transform of Eq. (19) is,
dt
pffiffi 
  1 C1 sinh s
dJg Jg Jg ðsÞ ¼ þ pffiffi (29)
¼ nBim Jbi þ (21) s s
dt a
Applying Eqs. (29) and (28) in (27),
40 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44

aþ1 1 X
∞ 2
ð1 þ aÞsinðln Þeln t
C1 ¼ (30) Jg ðtÞ ¼ 2 (40)
a suðsÞ
n¼1 au0 ðln Þl3n
where, Additionally Eq. (39) demonstrates that 〈J〉b(t) defined in
pffiffi pffiffi  pffiffi  dimensionless variables (13) and used in macroscopic Eqs.
ssinh s pffiffi  sinh s
uðsÞ ¼   cosh s  pffiffi (31) (20)e(23) is the same. This last result was not obvious because the
Bi a s averaged equations were deduced fully independently of punctual
equations. Eq. (40) is a continuous and analytical map that relates
Then, by the residual theorem of Cauchy, mechanistically (or dynamically) the dimensionless solute mass
pffiffi  fraction in extract Jg with time, the representation of this relation
sð1 þ aÞcosh s x est X ∞
Jb ðt; xÞ ¼ 1 þ lim þ (using the first 64 eigenvalues calculated with Eq. (35)) is plotted in
s/0 asuðsÞ n¼1 Fig. 1 at two mass Biot numbers and four values of a. It is obvious
pffiffi  that at lower values of a (lower extract mass with respect to
ðs  sn Þ ð1 þ aÞcosh s x est
 lim (32) underflow) the mass transfer is faster; however this tendency is
s/sn uðsÞ as
limited to the feasibility of minimal extract to underflow ratio.
where sn are the complex roots of u(sn) ¼ 0, for n ¼ 1,2, …,∞. With respect of the averaged mass transfer equations, taking
^pital's rule, Eq.
Considering that lim coshðxÞ/1 , and applying l’Ho Laplace transform of Eq. (20), solving for 〈J〉b(s) (with Eqs.
x/0 (21)e(23)) and taking the Laplace inverse, resulting in,
(32) can be written as,
 
X
∞ pffiffiffiffiffi fn 1þaþfa1=Bim Bimfþf t
ð1 þ aÞcoshð sn xÞesn t
Jb ðt; xÞ ¼ (33) 〈J〉b ðtÞ ¼ Jg ðtÞ ¼ e (41)
n¼1
u01 ðsn Þasn
Eq. (41) is analog to the analytical solution reported by
Definings ¼ il and considering that sinh(iz) ¼ isin(z),
rez et al. (2007) and Rodríguez-Jimenes et al. (2013)
Espinoza-Pe
cosh(iz) ¼ cos(z), Eq. (33) are expressed in real dominion as, but in dimensionless form and reduced to its minimal expression.

X
∞ 2
ð1 þ aÞcosðln xÞeln t
Jb ðt; xÞ ¼ 2 (34) 4.3. Application to vanilla extraction kinetic
n¼1 u0 ðln Þal2n
In order to apply the mechanistic (dynamics) models obtained
where ln are the eigenvalues generated by the roots of, in above section to experimental extraction kinetics, first it is
necessary the estimation of an apparent distribution constant of
ln sinðln Þ sinðln Þ
 cosðln Þ  ¼0 (35) solutes between phases (Keq). This distribution constant was esti-
Bi aln mated from experimental vanilla extraction in 60% w/w ethanol-
water solvent until equilibrium at different vanilla/solvent ratios
And
and temperatures. These results are summarized in Fig. 2. The
cosðln Þ sinðln Þ sinðln Þ cosðln Þ sinðln Þ dispersion observed in Fig. 2 is due to the fact that while solutes
u0 ðln Þ ¼ þ þ  þ (36) mass fraction in solvent was evaluated by direct evaporation of
Bi Biln ln al2 n al3 n extract, the solutes mass fraction in underflow was calculated by
Eq. (34) is a continuous and analytical map that relates the difference, assuming that the initial solute concentration in
dimensionless solutes mass fraction (Jb) with a given position in different vanilla samples was equal. Since vanilla initial solutes
underflow (in 1D rectangular coordinate) and time, and represent
the analytical solution of Eqs. (25) and (26). The average solution in
1D rectangular coordinate is,

Z1 ,Z1
〈J〉b ðtÞ ¼ Jb ðx; tÞdx dx (37)
0 0

Therefore,

X
∞ 2
ð1 þ aÞsinðln Þeln t
〈J〉b ðtÞ ¼ 2 (38)
n¼1 au0 ðln Þl3n

Applying Eq. (37) to Eq. (27)

Z 1 pffiffi  1

C1 cosh s þ dx pffiffi 
s 1 C1 sinh s
〈J〉b ðsÞ ¼ 0
Z 1 ¼ þ p ffiffi (39)
s s
dx
0

,
Which is exactly equal to Eq. (29) and therefore Jg(s) ¼ 〈J〉b(s), Fig. 1. Eq. (38) or Eq. (40) behavior at two values of mass Biot number and 4 values of
or Jg(t) ¼ 〈J〉b(t), or, a. J may be 〈J〉b or Jg.
 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44 41

calculated by Rodríguez-Jimenes et al. (2013) for vanillin in a

similar system at similar conditions (ethanol-water solvent and
300 rpm with 1 cm of orbital radius). The values were
kcg ¼ 1  104 m s1 and Dg ¼ 1.45  109 m2 s1. Applying these
values in Eq. (15), a value of Bi ¼ 162 was obtained, and considering
that Db could be a magnitude order lower than Dg, the mass Biot
number can be considered as Bi / ∞ for the present system.
Therefore, the mass transfer during solid-liquid extraction of 1 cm
vanilla pieces with characteristic length of 0.15 cm (Table 1) is
diffusion controlled. The a values were calculated with Eq. (16) and
first part of Table 1 data. The results were listed in the same Table 1.
Biot number has effect on analytical solution (Eq. (40)) through
Eqs. (35) and (36). In order to take into account the fact that solid-
liquid extraction of vanilla pieces is diffusion controlled, Eqs. (35)
and (36) rewritten for Bi / ∞ resulted in,

sinðln Þ
cosðln Þ þ ¼0 (35a)
aln

Fig. 2. Mass fractions of vanilla solutes in underflow and extract obtained at equilib-
rium during solid-liquid extraction in 60% w/w ethanol-water. Experimental points at sinðln Þ cosðln Þ sinðln Þ
(o) 30
C, (x) 40
C and (þ) 50
C. u0 ðln Þ ¼  þ (36a)
ln al2n al3n

were evaluated by exhaustive extraction (vanilla characterization in
Section 3), the determination was destructive and therefore the
samples for extraction were different of samples used for exhaus-
tive extraction. However, the probability of null hypothesis (Ho):
Keq ¼ 0 is p < 0.001. Therefore the estimated apparent equilibrium
constant Keq ¼ 0.86 is statistically significant even with the referred
dispersion, and the results are representative of the whole vanilla
used for experiments as it was explained in vanilla samples sub-
section (section 3). However, the results did not show statistic ev-
idence that temperature has effect on equilibrium constant (Keq).
Therefore, the equilibrium constant was considered Keq ¼ 0.86 at
temperatures between 30 and 50
C. It is important to note that
analytical solutions of mass transfer equation during solid-liquid
extraction process (Crank, 1975; Mikhailov, 1977; Chilev et al.,
2014) assume implicitly or explicitly a distribution constant be-
tween phases.
Knowing the apparent distribution constant, the mass transfer
relations experimentally obtained for vanilla solid-liquid extraction
kinetics are listed in the first part of Table 1 (diffusivity estimation).
The Biot mass number was estimated by considering the external
mass transfer coefficient (kcg) and effective diffusivity (Dg)
The firsts 64 eigenvalues (l1, l2, …, l64) of the analytical solution
(Eqs. (34) and (38) or 40) for the a values listed in Table 1 were
calculated with Eq. (35a), and the first (l1) eigenvalue at each
condition was also listed in Table 1. Like l1 < l2 < l3 …, Eq. (40) can
be written only in terms of l1,
!
  2ð1 þ aÞsinðl1 Þ l21 Db
ln Jg ðtÞ ¼ ln  t (42)
au0 ðl1 Þl31 l2
Eq. (43) is valid when ln(Jg(t)) vs t is linear, like it can be
observed in the semilog representation of Fig. 1 at t > 0.2.
As it was discussed, the extract concentration has more confi-
dence and therefore the effective diffusivities in underflow (Db)
were estimated form extract concentration kinetics in Eq. (42)
dominion, by linear regression fit of Eq. (42) on dimensionless
experimental concentration kinetics of solute in extract. The
experimental vanilla extractable solids kinetics in extract at 0.1 g/
mL (vanilla/solvent) and 30, 40 and 50
C, joint with fitted Eq. (42)
in the linear zone are plotted in Fig. 3. The diffusivities were
calculated with,

Table 1
Process variables and mass transfer characteristics of vanilla solid-liquid extraction kinetics in extract.

g/mL T(
C) mb0 (g)a mg0 (g)a mbe (g)a mge (g)a xge  102a a l1d
Diffusivity estimation
0.1 30 10.2 ± 0.10 87.8 ± 0.34 11.7 ± 0.54 86.3 ± 0.78 2.50 ± 0.02 6.37 1.6648
0.1 40 10.2 ± 0.03 88.6 ± 1.46 12.4 ± 0.32 86.4 ± 1.11 2.65 ± 0.26 6.01 1.6701
0.1 50 10.2 ± 0.06 88.2 ± 0.12 12.0 ± 0.06 86.4 ± 0.12 2.95 ± 0.04 6.19 1.6674
Model validation
0.3 30 30.0 ± 0.02 88.2 ± 0.15 33.3 ± 0.30 84.9 ± 0.13 7.67 ± 0.19 2.19 1.8170
1.75b 1.8677
2.63c 1.7811
0.4 50 40.1 ± 0.02 87.9 ± 0.34 48.9 ± 1.26 79.12 ± 1.66 10.0 ± 0.04 1.39 1.9280
1.11b 1.9950
1.67c 1.8793

l¼(1.5 ± 0.07)103 m for all samples (average ± standard deviation of 20 samples).

a
Average ± standard deviation of two replicates.
b
Calculated with lower limit of 80% confidence interval for Keq in Eq. (16).
c
Calculated with upper limit of 80% confidence interval for Keq in Eq. (16).
d
Calculated with Eq. (35a).
42 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44

Fig. 3. Experimental extract kinetics during vanilla solid-liquid extraction in 60% w/w
ethanol-water at 0.1 g/mL. Experimental points at (o) 30
C, (x) 40
C, (þ) 50
C.
Continuous lines represent Eq. (42) fitted by linear regression. Fig. 4. Arrhenius graph of vanilla extractable solids average diffusivity in underflow
during vanilla solid-liquid extraction in 60% w/w ethanol-water.

d l2 In the first paragraph of section 4.3 it was discussed that Keq is

Db ¼  (43) the estimated property with more uncertainly. Therefore, in order
l21
to evaluate the effect of Keq dispersion, the 80% confidence interval
was calculated with the parameter variance. The result was
where, d is the slope of Eq. (43) fitted to experimental results.
0.86 ± 0.17. Applying the lower and upper limits in Eq. (16) pro-
Calculated diffusivities are listed in Table 2 jointly with the prob-
duced lower and upper limits for a and in consequence for eigen-
ability of null hypothesis (Ho): Db ¼ 0. These probabilities were
values ln. a and l1 limits are listed in the second part of Table 1.
calculated with the slope (d) variance. Fig. 4 shows that calculated
Eq. (40) was solved with the first 64 eigenvalues calculated with
diffusivities follows an Arrhenius relation with respect to
expected value and 80% confidence interval limits of a (Table 2) and
temperature.
Db taken from Table 2. The experimental and simulated results are
plotted in Figs. 5 and 6. It is evident that Eq. (40) is robust with
4.4. Extraction kinetics predictions
respect to Keq dispersion, and the model has the capacity to predict
extraction kinetics at different vanilla/solvent ratios. Applying the
In order to validate the dynamic models capacity for solid-liquid
same criterion, Eq. (41a) was solved with expected value and 80%
extraction kinetics prediction, the punctual analytical solution (Eq.
confidence interval limits of a (Table 2) and Db taken from Table 2.
(40)) and averaged analytical solution (Eq. (41)) were used to pre-
dict the vanilla solid-liquid extraction kinetic at different condi-
tions than those used for diffusivity estimation. Two conditions
were selected at two different vanilla/solvent ratios and at the
extreme temperatures: 30
C with 0.3 g/mL, and 50
C with 0.4 g/
mL. The experimental kinetics at these conditions were developed
as detailed in experimental kinetics sub-section (section 3). The
experimental results for initial and equilibrium masses and
experimental mass fraction at equilibrium are listed in second part
(model validation) of Table 1.
Similarly to the case of punctual analytical solution, averaged
solution (Eq. (41)) simplified for Bi / ∞ was written as,
 
fn 1þa1 t
〈J〉b ðtÞ ¼ Jg ðtÞ ¼ e (41a)

Table 2
Effective solute diffusivity in vanilla underflow during solid-liquid extraction at
0.1 g/mL (vanilla/solvent).

T(
C) Db  1011(m2 s1) pa

30 1.221 <0.001
40 1.975 <0.001
Fig. 5. Experimental (o) and simulated extract kinetics during vanilla solid-liquid
50 2.433 <0.002
extraction in 60% w/w ethanol-water at 30
C and 0.3 g/mL. Continuous line repre-
a
Null hypothesis (Ho: Db ¼ 0) probability from parameter variance calculated as sents the simulated kinetic with Eq. (40), discontinuous lines represent the predictions
detailed in Rodríguez-Jimenes et al. (2013). at lower and upper limits of 80% confidence interval of Keq.
 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44
Fig. 6. Experimental (o) and simulated extract kinetics during vanilla solid-liquid
extraction in 60% w/w ethanol-water at 50
C and 0.4 g/mL. Continuous line repre-
sents the simulated kinetic with Eq. (40), discontinuous lines represent the predictions
at lower and upper limits of 80% confidence interval of Keq.
The experimental and simulated results are plotted in Figs. 7 and 8.
Eq (41a) show similar robustness than Eq. (40) with respect to Keq
dispersion.
It is evident that both Eqs. (40) and (41a) reproduce the exper-
imental kinetics with good approximation. Eq. (40) represents
moderately better the experimental kinetics that Eq. (41a), but the
two equations predict the same equilibrium time. This demon-
strates that punctual Eq. (40) are better for diffusivity estimation
but averaged equations can be used for process simulation. Aver-
aged equations in differential form (9)e(12) may be particularly
useful in the case of continuous contact process simulation in
which the analytical solution is not possible (Veloso et al., 2005).
The proposed models describe the solid-liquid extraction kinetics
knowing the equilibrium concentration in extract. In this study the
Fig. 7. Experimental (o) and simulated extract kinetics during vanilla solid-liquid
extraction in 60% w/w ethanol-water at 30
C and 0.3 g/mL. Continuous line repre-
sents the simulated kinetic with Eq. (41), discontinuous lines represent the predictions
at lower and upper limits of 80% confidence interval of Keq.
43
Fig. 8. Experimental (o) and simulated extract kinetics during vanilla solid-liquid
extraction in 60% w/w ethanol-water at 50
C and 0.4 g/mL. Continuous line repre-
sents the simulated kinetic with Eq. (41), discontinuous lines represent the predictions
at lower and upper limits of 80% confidence interval of Keq.
equilibrium concentration in extract was evaluated experimentally.
In order to predict equilibrium concentration in one single solid-
liquid extraction stage or in a multi-stages process, the proposed
model must be complemented with the design model proposed by
Castillo-Santos et al. (2016).
5. Conclusion
The developed analysis in the present paper resulted in explicit,
continuous and analytical maps that represent the solid-liquid
extraction process dynamics. That is, It was mathematically
deduced an explicit relationship between local and space-averaged
solute mass fractions in underflow and extract with respect to time
in terms of mass transfer driving forces. Mathematical relations
obtained were successfully applied as mechanistic models for so-
lute diffusivity estimation in underflow and for the prediction of
process kinetics in a vanilla-(60% w/w ethanol-water) extraction
system. The proposed models (punctual or averaged) can be
applied to any solid-liquid extraction system in which mass transfer
and equilibrium properties are known or evaluated as described in
this study.
Acknowledgements
The authors express their acknowledgments to the Tecnolo
gico
Nacional de Me
xico by the financial support through the project
5177.13-P, and Mexican Consejo Nacional de Ciencia y Tecnología
(CONACyT) by the scholarship assigned to Castillo-Santos, K.
References
Amendola, D., De Faveri, D.M., Spigno, G., 2010. Grape marc phenolics: extraction
kinetics, quality and stability of extracts. J. Food Eng. 97, 384e392.
Cacace, J.E., Mazza, G., 2003. Mass transfer process during extraction of phenolic
compounds from milled berries. J. Food Eng. 59, 379e389 (Diffusivities and
mass transfer coefficients from literature models).
Castillo-Santos, K., Aguirre-Alonso, R.O., Rodríguez-Jimenes, G.C., Robles-
Olvera, V.J., Salgado-Cervantes, M.A., García-Alvarado, M.A., 2016. An optimi-
zation based algorithm for solving design problems of counter-current multi-
stage batch solid-liquid extractors for complex systems: application to vanilla
extract. Comput. Chem. Eng. 89, 53e61.
Chilev, C., Koleva, V., Simeonov, E., 2014. A new empirical model for calculation the
44 K. Castillo-Santos et al. / Journal of Food Engineering 192 (2017) 36e44
effective diffusion coefficient for solidliquid extraction from plants. Industrial
Eng. Chem. Res. 53, 6288e6296.

rdova-Quiroz, A.V., Ruiz-Cabrera, M.A., García-Alvarado, M.A., 1996. Analytical
Co
solution of mass transfer equation with interfacial resistance in food drying.
Dry. Technol. 14, 1815e1826.
Crank, J., 1975. The Mathematics of Diffusion, second ed. Clarendon Press, Oxford.
Espinoza-Pe
rez, J.D., Vargas, A., Robles-Olvera, V.J., Rodríguez-Jimenes, G.C., García-
Alvarado, M.A., 2007. Mathematical modeling of caffeine kinetic during solid-
liquid extraction of coffee beans. J. Food Eng. 81, 72e78.
Franco, D., Sineiro, J., Pinelo, M., Nún ~ ez, M.J., 2007. Ethanolic extraction of Rosa
rubiginosa soluble substances: oil solubility equilibria and kinetic studies.
J. Food Eng. 79, 150e157.
Garcia-Perez, J.V., García-Alvarado, M.A., Carcel, J.A., Mulet, A., 2010. Extraction ki-
netics modeling of antioxidants from grape stalk (Vitis vinifera var. Bobal): in-
fluence of drying conditions. J. Food Eng. 101, 49e58.
Linares, A.R., Hase, S.L., Vergara, M.L., Resnik, S.L., 2010. Modeling yerba mate
aqueous extraction kinetics: influence of temperature. J. Food Eng. 97, 471e477.
McCabe, W., Smith, J., Harriott, P., 2004. Unit Operations of Chemical Engineering.
McGraw-Hill.
Mikhailov, M.D., 1977. General solution of diffusion processes in solid-liquid
extraction. Int. J. Heat. Mass Transf. 20, 1409e1415.
~ ez, M.J., 2006. Mass transfer during continuous solid-
Pinelo, M., Sineiro, J., Nùn
liquid extraction of antioxidants from grape byproducts. J. Food Eng. 77, 57e63.
Rakotondramasy-Rabesiaka, L., Havet, J.L., Porte, C., Fauduet, H., 2010. Estimation of
effective diffusion and transfer rate during the protopine extraction process
from Fumaria officinalis L. Sep. Purif. Technol. 76, 126e131.
Rodríguez-Jimenes, G.C., Vargas-Garcia, A., Espinoza-Pe
rez, D.J., Salgado-
Cervantes, M.A., Robles-Olvera, V.J., García-Alvarado, M.A., 2013. Mass transfer
during vanilla pods solid-liquid extraction: effect of extraction method. Food
Bioprocess Technol. 6 (10), 2640e2650.
Seikova, I., Simeonov, E., Ivanova, E., 2004. Protein leaching from tomato
seedeeexperimental kinetics and prediction of effective diffusivity. J. Food Eng.
61, 165e171.
Simeonov, E., Tsibranska, I., Minchev, A., 1999. Solideliquid extraction from
plantseexperimental kinetics and modelling. Chem. Eng. J. 73, 255e259.
Spigno, G., De Faveri, D.M., 2007. Antioxidants from grape stalks and marc: influ-
ence of extraction procedure on yield, purity and antioxidant power of the
extracts. J. Food Eng. 78, 793e801.
Veloso, G.O., Krioukov, V.G., Vielmo, H.A., 2005. Mathematical modeling of vege-
table oil extraction in a counter-current crossed flow horizontal extractor.
J. Food Eng. 66, 477e486.