Journal of Food Engineering 218 (2018) 14e23

Contents lists available at ScienceDirect

Journal of Food Engineering

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

Experimental data and estimation of sucrose solubility in impure

solutions

Bernardo b, Caliane B.B. Costa c, *, Marco Giulietti 1
Carlos E. Crestani a, Andre
a ~o Paulo, Rua St

Federal Institute of Education, Science and Technology of Sa ~o, SP, 15991-502, Brazil
efano D'Avassi, 625 - Nova Cidade, Mata
b
Department of Chemical Engineering, Federal University of Sa ~o Carlos, Rodovia Washington Luís, Km 235, Sa~o Carlos, SP, 13565-905, Brazil
c
Department of Chemical Engineering, State University of Maringa, Avenida Colombo, 5790, Bloco D90, Jd. Universita
rio, Maringa
, PR, 87020-900, Brazil

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

Article history: Although sucrose solubility has been widely studied, the presence of impurities, which is invariably true
Received 1 May 2017 in sucrose industrial crystallization, affects its solubility and the extent of this influence is not fully
Received in revised form addressed in the literature. This study aimed to obtain experimental data of sucrose solubility in aqueous
11 July 2017
pure and impure solutions from 10 to 100
C and evaluate prediction ability of SLE models. Experimental
Accepted 21 August 2017
Available online 29 August 2017
data shows that impurities in sugarcane juice, i.e. reducing sugars and salts, in levels studied (with 92.8%
of sucrose up to pure solutions) increase sucrose solubility. Evaluating existing models for SLE showed
that they generally provide poor prediction. However, a proposed modification in the Nývlt Equation of
Chemical compounds studied in this article:
Sucrose (PubChem CID: 5988)
solubility provides a small mean deviation (2.54%) between experimental and calculated data in pure and
Glucose (PubChem CID: 79025) impure aqueous solutions, making this SLE model useful for solubility calculation in crystallization
Fructose (PubChem CID: 5984) studies, especially when syrups containing typical impurities of sugarcane juice are used.
© 2017 Elsevier Ltd. All rights reserved.
Keywords:
Impurities
Mathematical modelling
Solid-liquid equilibrium
Solubility
Sucrose

1. Introduction
Sucrose (IUPAC name (2R,3R,4S,5S,6R)-2-[(2S,3S,4S,5R)-3,4-
dihydroxy-2,5-bis(hydroxymethyl)oxolan-2-yl]oxy-6-(hydrox-
ymethyl)oxane-3,4,5-triol) is a very important ingredient in the
food industry, used in numerous products from cured meat to
candies and ice-cream. Sugarcane industry is the oldest industry in
Brazil and it is one of the main pillars of Brazilian agroindustry
nowadays, with a production higher than 35 million tonnes of
sugar (UNICA, 2016). Thus, an accurate modelling of process steps is
of major importance in order to use process systems engineering
tools to optimize process operations and economics. Sucrose pro-
duction inevitably goes through a crystallization step, and sucrose
crystallization in sugarcane industries invariably occurs in the
presence of impurities commonly found in syrups, such as reducing
* Corresponding author.
E-mail addresses: cecrestani@ifsp.edu.br (C.E. Crestani), abernardo@ufscar.br
(A. Bernardo), cbbcosta@uem.br (C.B.B. Costa).
1
In memoriam.
sugars and salts, which may affect the crystallization process,
Therefore, the study of sucrose Solid-Liquid Equilibrium (SLE) in the
presence of these typical impurities is of great importance for the
crystallization process.
Authors have been publishing research concerning sucrose
solubility in pure aqueous solutions has been published since the
1940s (Taylor, 1947) and both experimental data and mathematical
modelling can be found in the literature. On the other hand, sucrose
solubility in impure aqueous solutions is a more complex issue and
it is not clearly presented in the literature, although solubility data
from beet syrups have been present in the literature since the 1930s
(Brown and Nees, 1933; Dahlberg and Bennett, 1951 and others).
Experimental data are limited and the available data are conflicting.
Martins et al. (2009) cite a great difficulty in developing a generic
solubility model, able to calculate sucrose solubility in impure so-
lutions from different sources (beet or sugarcane) and from
different regions of the world, and therefore, according to the au-
thors, a specific study of each solution is needed. Authors presented
solubility data of sucrose in impure syrups at temperatures be-
tween 62 and 66
C using a methodology described in a previous

http://dx.doi.org/10.1016/j.jfoodeng.2017.08.023
0260-8774/© 2017 Elsevier Ltd. All rights reserved.
 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23
study (Martins et al., 2005). Nature and the composition of impu-
rities influence sucrose solubility. Chen and Chou (1993) cite that
most inorganic salts tend to increase solubility and reducing sugars
tend to decrease it in mass basis. A model proposed by Frew (1973)
is widely used in the literature by adjusting some parameters to the
experimental data (Van der Poel et al., 1998; Georgieva et al., 2003;
Rozsa, 2000; Steindl et al., 2001). However, as the experimental
data are conflicting, parameters of the same equation differ
depending on the author. There are also some thermodynamic
activity coefficient models for sugar systems in the literature, based
on both UNIQUAC and UNIFAC. Despite the relatively large number
of equilibrium models based on activity coefficients available in the
literature, to the best of the authors’ knowledge none of them were
tested in SLE calculation of sucrose impure aqueous solutions.
These models would only be properly evaluated in systems in
which impurities are completely characterized.
Aiming to fill this gap, the first aim of this study is to present
new experimental data of sucrose solubility both in pure aqueous
solutions and in aqueous solutions containing impurities
commonly found in sugarcane syrups. With these experimental
data, the second aim of this study is to evaluate and compare
models of sucrose solubility in the presence of impurities. One of
these models is an original modification of the Nývlt Equation of
solubility (Nývlt et al., 1985), which proves to be very useful. This
same equation is called the Apelblat Equation in some more recent
studies referring to an article of 1993 (Apelblat, 1993), which aimed
to calculate activity and osmotic coefficients in electrolyte
solutions.
2. Theory
This section presents mathematical models proposed in the
literature for the calculation of sucrose solubility in pure and
impure solutions for the best understanding of the applications in
this work.
Sucrose solubility in pure aqueous solutions can be widely found
in the literature as well as mathematical modelling of these solu-
tions. Equations (1) and (2) are recommended by the International
Commission for Uniform Methods of Sugar Analysis (ICUMSA) to
calculate sucrose solubility in aqueous solutions, respectively, up to
100
C and from 100 to 145
C:
wS ¼ 64:447 þ 0:08222$T þ 1:66169  103 $T 2  1:558
 106 $T 3  4:63  108 $T 4
wS ¼ 71:0615 þ 0:053625$T þ 0:000655303$T 2
where ws is the percentage in mass of soluble sucrose and T is the
temperature in
C.
Sucrose solubility in impure aqueous solutions is a more com-
plex issue and it is not consolidated in the literature. Frew (1973)
proposed the calculation of sucrose solubility in impure solutions
using the variable D, which is dependent on the concentration of
impurities in syrup and it is calculated by Equation (3):


I ematical modelling of SLE.
D ¼ 1  0:088$
W
where I is the mass of impurities and W is the mass of water. Sol-
ubility is calculated as:
Z ~o
ðS=WÞsat ¼ D$
100  Z
15
where (S/W)sat is the solubility in the impure solution, expressed as
mass of sugar per mass of water, and Z is the solubility of pure
solution. Equation (5) presents the equation proposed by Vavrinecz
(1978), which has been used in several studies in the literature:


mI m
Ssat ¼ m$ þ b þ ð1  bÞ$exp  c$ I (5)
mW mW
where Ssat is the solubility expressed as gram of solute per gram of
water, mI and mW are, respectively, the mass of impurities and
water and m, b and c are parameters, determined by the impurities
present in the solution. Values for these parameters were deter-
mined in some studies in the literature, with conflicting values
among them. Van der Poel et al. (1998) cited values of 0.178, 0.820
and 2.1, respectively; Georgieva et al. (2003) used values of 0.1, 0.4
and 0.24 in a crystallization study; and Rozsa (2000) cited that the
values of 0.178, 0.820 and 2.1 are typical for beet syrups while the
values of 0.06265, 0.982 and 2.1 are the best ones for sucrose
solubility in solutions derived from sugarcane.
Apart from these equations, some thermodynamic activity co-
efficient models for sugar systems have been proposed since the
1990s. Peres and Macedo (1996) estimated interaction parameters
of a UNIQUAC model, with the modification in combinatorial
contribution proposed by Larsen et al. (1987) in order to calculate
the equilibrium of solutions containing sugars. This method is
called P&M-UNIQUAC. Furthermore, when no experimental data is
available, predictive methods can be alternatively used. Thus, some
UNIFAC based methods were developed to predict the equilibrium
of sugar solutions. These methods included new functional groups
aiming to represent the complex behavior of sugar in a solution. As
sugar molecules should not be dissociated because of the vicinity of
functional groups, interaction of groups inside the molecule is
crucial to represent the interaction of the whole molecule in solu-
tion. Besides, inclusion of new functional groups allows for differ-
entiating isomers such as fructose and glucose. Abed et al. (1992),
Catte
et al. (1995), Kuramochi et al. (1997) and Spiliotis and
Tassios (2000) are examples of authors who introduced different
decomposition groups to represent sugar molecules. Interaction
groups introduced by Catte
et al. (1995) and Peres and Macedo
(1997) (P&M-UNIFAC) were used by Ferreira et al. (2003) to
develop a method called A-UNIFAC, which combines the traditional
UNIFAC contributions (residual and combinatorial) to the associa-
tive one, related to the interaction forces of hydroxyl groups in
solution. The method introduced by Kuramochi et al. (1997) called
(1) Bio-UNIFAC was proposed for calculating the activity coefficients of
biochemical compounds in water, including sugars. The contribu-
(2) tions are divided in short-range, which are comprised in the
modified UNIFAC of Larsen et al. (1987), and long-range, which
describe ion-ion interaction and enable the method to be used in
electrolyte solution calculations. Tsavas et al. (2004) modified and
extended S-UNIFAC, the method developed by Spiliotis and Tassios
(2000), creating the mS-UNIFAC method.
3. Experimental
This section is divided into two parts: experimental and math-
(3)
3.1. Experimental methodology
3.1.1. Chemicals
Deionized water type 2 produced by a Millipore Elix 10 water
(4) purification system, commercial refined sugar trademark Unia
with a mass fraction of 99.8% of sucrose ((2R,3R,4S,5S,6R)-2-
[(2S,3S,4S,5R)-3,4-dihydroxy-2,5-bis(hydroxymethyl)oxolan-2-yl]
16 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23

oxy-6-(hydroxymethyl)oxane-3,4,5-triol) and commercial brown Table 2

sugar trademark Vitaçucar with a mass fraction of sucrose of 92.8% Brown sugar composition.

without any additional purification steps were used in the experi- Test Result Unit
ments. Table 1 summarizes the main information of chemicals Thermotolerant Coliforms <10 UFC/g
used, while Table 2 presents the composition of the brown sugar Salmonella spp. Absent per 25 g
used. The main impurities of brown sugar are glucose and fructose Humidity <2.5 g/100 g
whose IUPAC names are, respectively, (2S,3R,4S,5S,6R)-6-(hydrox- Mould and yeast <10 UFC/mL.g
Copper <0.2 mg/kg
ymethyl)oxane-2, 3, 4, 5-tetrol and (3S,4R,5R)-1,3,4,5,6-
Calcium 666.5 mg/kg
pentahydroxyhexan-2-one. Sodium 202.1 mg/kg
Potassium 342.1 mg/kg
Phosphorous 22.9 mg/kg
3.1.2. Equipment
Iron 30.9 mg/kg
Experiments were carried out in a 10-L jacketed reactor equip- Magnesium 589.3 mg/kg
ped with a refractometer (MPR E-Scan Electron Machine Corpora- Sucrose percent 92.8 % mass
tion, with a precision of 0.0001 units of refraction index), and Reducing sugars 5.4 % mass
temperature (±0.1
C) inline measurement probes. Moreover, a
Programmable Logic Controller (PLC), able to maintain solution in
the reactor at constant temperature and/or to perform its heating
or cooling, was available. A constant stirring speed of 80 rpm and a !
1 Pur
double anchor propeller were used in all experiments. The pres- D ¼ 1  F$ x0S $ (7)
ence of a solid phase was always monitored using a G400 Mettler 1  x0S
Toledo FBRM® (Focused Beam Reflectance Measurement).
xS ¼ D$x0S (8)
3.1.3. Experimental procedure
3.1.3.1. Reference curve of refractometer. Three sets of aqueous so- where x0s is the mass fraction of the pure solution, T is the tem-
lutions with sugar mass fractions of 0.2, 0.4, 0.65, 0.70, 0.75, 0.80 perature in
C, RI the refraction index, a, b, c, d, e, f, g, h, i, j and F are
and 0.85 were prepared. In the first set, only refined sugar was adjustable parameters, D is a variable proportional to the concen-
used; in the second one, only brown sugar was used; finally, in the tration of impurities, Pur is the purity of sugar in solution obtained
third set, solutions were prepared with 50% in mass of refined and by HPLC analysis and expressed in mass fraction and xS is the mass
50% of brown sugar. Each solution was separately added to the fraction of sugar dissolved in solution.
reactor, heated to 100
C and slowly cooled until 20
C (with cooling Parameters of Equations (6) and (7) were fitted using the Lev-
rates varying from 10 to 35
C/h). The refractive index and tem- enbergeMarquardt algorithm for minimization of an objective
perature of solution were measured every 30 s in both heating and function (sum of squared deviation between experimental and
cooling and an eventual birth of crystals was monitored with the calculated data of solution concentration) in Scilab® software.
FBRM. The birth of crystals was noticed only in the brown sugar
solution with mass fraction 0.85 at temperatures below 42
C and, 3.1.3.2. Solubility. Solubility experiments were carried out using a
therefore the data related to this solution with temperatures below procedure based on the one proposed by Myerson (2001) and used
42
C were discarded. The measured refraction index was the same in a previous study (Crestani et al., 2013). A suspension of sugar in
at the same temperature, when the temperature was achieved both water was prepared and added to the reactor at the experiment
during heating and during cooling. This fact indicates that temperature. Excess of sugar was added based on the literature
composition of solutions was the same before and after heating and data of sucrose solubility and presence of crystals was monitored by
there was no decomposition of any of the substances present in the FBRM to guarantee the presence of undissolved sugar. Sus-
solution. pensions were maintained under agitation at a constant tempera-
A calibration curve was then prepared to estimate the solution ture until refraction index (registered each 30 s) did not vary
concentration from refractive index and temperature data anymore. In the experiments, a maximum of 8 h of variation in the
measured in the reactor. Equations (6)e(8) present the empirical refraction index was observed. However, in order to assure com-
relation used: plete dissolution up to the concentration corresponding to solubi-
lity, at least 12 h of experiments were provided. The procedure was
x0S ¼ a þ b$T þ c$lnðRIÞ þ d$T 2 þ e$lnðRIÞ2 þ f $T$lnðRIÞ þ g$T 3 carried out in triplicate for pure sucrose aqueous solutions (pre-
þ h$lnðRIÞ3 þ i$T$lnðRIÞ2 þ j$T 2 $lnðRIÞ pared with refined sugar) at temperatures of 10, 25, 30, 40, 50, 60,
70, 80, 90 and 100
C; for impure aqueous solutions prepared with
(6)
brown sugar at temperatures of 30, 40, 50, 60, 70, 80, 90 and 100
C;
and for solutions prepared with 50% of brown sugar and 50% of

Table 1
Sample table.

Chemical name Source Initial Mass Fraction Purification Final Mass Fraction Analysis
Purity Method Purity Method
a
Deionized water 0.999 None e e
Sucrose ((2R,3R,4S,5S,6R)-2-[(2S,3S,4S,5R)-3,4-dihydroxy-2,5- ~o
Unia 0.998 None e e
bis(hydroxymethyl)oxolan-2-yl]oxy-6-(hydroxymethyl)oxane-
3,4,5-triol))
Brown sugar Vitaçucar Composition in Table 2 None e e
a
Deionized water type 2 with TOC < 30 ppb, conductivity < 0.2 mS,cm1 and resistivity (5e15) MU,cm, produced by Millipore Elix 10 water purification system.
 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23 17

refined sugar at the same temperatures from 30 to 100
C. In so-

  A
lutions containing impurities, a sample of the liquid phase was
ln x01 ¼ A1 þ 2 þ A3 $lnðTÞ (10)
taken for measurement of sucrose concentration by High Pressure T
Liquid Chromatography (HPLC) (analysis procedure is described by
!
Sonego et al., 2014). 1  Pur
D ¼ 1  N$ x01 $ (11)
1  x01

3.2. Mathematical modelling of SLE

Mathematical models of sucrose SLE available in the literature
were evaluated both for aqueous pure and impure sucrose solu-
tions. Expressions proposed by Frew (1973) and Vavrinecz (1978)
with the parameters cited by Van der Poel et al. (1998), Georgieva
et al. (2003) and Rozsa (2000) for beet and sugarcane syrups
were tested and calculated values of solubility were compared to
the experimental data obtained in this study. Mathematical models
based on activity coefficient calculation were evaluated as well. The
sucrose activity coefficient was calculated using P&M-UNIQUAC,
P&M-UNIFAC, Bio-UNIFAC, mS-UNIFAC and A-UNIFAC and solubil-
ity was obtained using Equation (9) (Myerson, 2001) and parame-
ters of Table 3.
x1 ¼ D$x01 (12)
where x01 is the molar fraction of sucrose in the pure solution, A1, A2
e A3 are adjustable parameters of Nývlt Equation, T is the temper-
ature in Kelvin, N is an additional adjustable parameter, Pur is the
purity of the sugar in solution and x1 is the solubility expressed as
the molar fraction of solute. Purity was calculated from the results
of HPLC analysis. Parameter fitting was performed using the
Levenberg-Marquardt algorithm for minimizing an objective
function (sum of squared deviation between experimental and
calculated solubility data, in the entire range of the concentration of
impurities studied) in Scilab® software.

 


DHfus T DCP Tm  T
ln xsug $gsug ¼  $ 1 þ $
R$T Tm R T


DCP T
þ $ln
R Tm
where xsug is sugar solubility at the solution temperature (T)
expressed in molar fraction; gsug is the activity coefficient of sugar
in solution calculated by the thermodynamic model; DHfus is the
melting enthalpy at the normal melting temperature (Tm) and DCp
is the difference between the heat capacity of pure solvent and
solid sugar.
Equation (9) is widely used to determine the experimental ac-
tivity coefficient (Myerson, 2001). In this work, this equation was
used to obtain sucrose activity coefficient from solubility experi-
mental data. For this purpose, physical properties of mS-UNIFAC
(Tsavas et al., 2004) was used. Values of DHfus, Tm and DCp should
not vary among methods, but as it happens in the literature data,
the choice of the values of these physical properties was made
based on the quality of estimations of VLE (Crestani et al., 2011) and
SLE (Crestani et al., 2013) of sugar solutions in earlier studies of the
authors of this paper.
Impurities were considered consisting of fructose and glucose
in the proportions of HPLC analysis. Composition presented in
Table 2 shows that glucose and fructose are responsible for 75% in
mass of the impurities of brown sugar (5.4% by weight are glucose
and fructose in brown sugar with 7.2% of impurities and
humidity).
A modification in Nývlt Equation of solubility (Nývlt et al., 1985)
is proposed here by including a variable (D), proportional to the
concentration of impurities in solubility calculations. Equations
(10)e(12) present the proposed mathematical model.
4. Results and discussion
In this section, the developed refractometer reference curve is
(9) presented followed by the solubility experimental results and dis-
cussion and mathematical models evaluation and comparison.
4.1. Reference curve
A single calibration curve for solution concentration calculation
from the refractive index and temperature measurement in
different concentration of impurities was developed. Table 4 pre-
sents the values of fitted parameters of Equations (6) and (7) and
the mean square deviation between experimental and calculated
data of all experimental points. The mean square deviation be-
tween the experimental and calculated data of the 3755 experi-
mental points (309 points of solution prepared with refined sugar,
1447 points of the solution prepared with brown sugar and 1999
points of the solution with mixed refined and brown sugar) is
3.04  103 (percent mean deviation of 0.34%).
Fig. 1 presents experimental versus calculated data of solution
concentration expressed as the mass fraction and a linear fit of
these values. As the coefficient of determination (0.9989) demon-
strates, experimental and calculated data are very close to each
other.
Table 4
Parameters a-j and F of Equations (6) and (7) fitted to sucrose aqueous solutions in
the presence of impurities.
Parameter Value Unit

a 4.19 (g/g)
b 1.45  102 (g/g)$
C1
Table 3 c 2.29  101 (g/g)
Physical properties of the different activity coefficient based models. d 1.43  105 (g/g)$
C2
e 3.49  101 (g/g)
Model DHfus/J$mol1 Tm/K DCp/J$mol1$K1 f 7.32  102 (g/g)$
C1
g 1.10  108 (g/g)$
C3
P&M-UNIFAC 46187 459.15 424.2026e3.47$ (T-T0)
h 2.21  101 (g/g)
Bio-UNIFAC 40356 459.15 88.0
i 9.24  102 (g/g)$
C1
P&M-UNIQUAC 46187 459.15 316.1153e1.1547$ (T-T0)
j 2.95  105 (g/g)$
C2
S-UNIFAC 57000 459.15 200.1
F 1.44  101 e
mS-UNIFAC 40356 459.15 88.0
A-UNIFAC 56946 458.15 254 Mean squared deviation 3.04  103 (g/g)2
18 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23

0.8
R² = 0.9989

Calculated mass fracƟon

0.7

0.6

0.5

0.4

0.3

0.2
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Experimental mass fracƟon

Fig. 1. Experimental versus calculated solution concentration and their linear fit.

4.2. Solubility experimental results Table 5

Sucrose (xS), glucose (xG) and fructose (xF) mean mass fraction in solutions of sol-
ubility experiments with aqueous solutions of brown sugar.a
Fig. 2 presents the experimental data of sucrose solubility in
aqueous solutions of refined sugar, brown sugar and the mixture of Sample xS xG xF
both. The presented data is the mean value of triplicates of each 30
C 0.9279 0.0268 0.0285
experimental point. Mean standard deviation (expressed as molar 40
C 0.9298 0.0241 0.0297
fractions) between the samples are 1.35  104 for refined sugar 50
C 0.9439 0.0156 0.0232
60
C
solution, 1.68  103 for brown sugar solution and 1.72  103 for
0.9561 0.0031 0.0229
70
C 0.9472 0.0097 0.0257
the solution prepared with a mixture of both. Because of the order 80
C 0.9507 0.0054 0.0262
of magnitude of deviation, the error bar is not visible in the chart of 90
C 0.9437 0.0136 0.0253
Fig. 2. The line is the sucrose solubility in aqueous solutions 100
C 0.9437 0.0135 0.0255
calculated by the ICUMSA recommended expression (Equation (1)). a
Uncertainties u are u(xS) ¼ u(xG) ¼ u(xF) ¼ 0.0001, u(T) ¼ 0.1
C.
Solution concentration of all experimental data was determined
using Equations (6)e(8) with parameters of Table 4. The sucrose
molar fraction in each solution (as well as glucose and fructose An additional measurement was performed to obtain concen-
concentration) was determined using the results of HPLC analysis. tration of sucrose, glucose and fructose from the original brown
These data are presented, for solutions prepared with brown sugar, sugar. Sucrose, glucose and fructose concentrations, expressed in
in Table 5. Each value in this table is the mean value of all mea- mass fraction, determined by this analysis were, respectively,
surements provided for each experimental point (triplicates of 0.9501, 0.0095 and 0.0228, which is in accordance with supplier
solubility points and duplicate for HPLC measurements). information presented in Table 2 (sucrose mass fraction excluding

0.30

Rebined sugar
0.25 Brown sugar
Mixture
Sucrose mass fracƟon

0.20 ICUMSA EquaƟon

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 2. Sucrose solubility from aqueous solutions with different impurities concentration.
 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23
humidity is 0.9528).
Results presented in Fig. 2 show that the higher the concen-
tration of impurities, the higher the sucrose solubility is. Martins
et al. (2009) report an increase in solubility coefficient in the
presence of fructose, but a decrease in the presence of glucose and
starch (at temperatures between 62 and 66
C and fructose, glucose
and starch concentrations of dried solids, respectively, from 0.02 to
0.25, from 0.005 to 0.100 and from 0.005 to 0.009). Table 6 shows
solubility experimental values (between 10 and 100
C for pure
solutions and between 30 and 100
C for the impure ones) and
experimental activity coefficient values of sucrose, calculated from
Equation (9), using physical properties obtained by Tsavas et al.
(2004), presented in Table 3. Sugar purity (Pur) was calculated
from the compositions in Table 5.
The combined presence of glucose and fructose, along with the
other impurities of brown sugar (Table 2), leads to a decrease in
sucrose activity coefficient values, which results in higher solubility
values. Glucose solubility is higher than that of sucrose at tem-
peratures up to 50
C (Peres and Macedo, 1996), while fructose is
more soluble than sucrose in the whole range of studied temper-
atures (Crestani et al., 2013).
4.3. Mathematical modelling of SLE
Figs. 3 and 4 present the experimental data of sucrose solubility
in impure aqueous solutions (with brown sugar and with a mixture
of brown and refined sugar respectively), calculated solubility using
Equation (5) (with literature values for the parameters m, b and c,
fitted from experimental data using beet syrups (Van der Poel et al.,
1998; Rozsa, 2000), European syrup of sugarcane (Georgieva et al.,
2003)), and calculated solubility using Equation (4), which was
developed from experimental data from Australian syrups (Frew,
1973).
Results presented in Figs. 3 and 4 show small differences in
solubility of sucrose from different sources (syrups of sugarcane or
beet and from different locations around the world, from Brazil to
Australia, passing through Portugal). Fig. 3, in which the sugar
purity in solution varied from 0.871 to 0.920 in molar fraction,
shows that Equation (5) with parameters fitted by Rozsa (2000)
leads to the best prediction of experimental data obtained in this
study. When impurities are in smaller concentrations (purity
varying from 0.934 to 0.959 in molar basis, Fig. 4), this ESL model,
with the same values of the parameters (Rozsa, 2000), provides
small deviations. However, with these concentrations of impurities,
the values of the parameters fitted by Georgieva et al. (2003) from
experimental data of European sugarcane syrups applied to Equa-
tion (5) led to the smallest deviations.
Table 6
Sucrose solubility (xs*) expressed in molar fraction and the experimental activity coefficient (gexp
fraction of sucrose in the initial amount of added sugar).a
T/
C xs*
Pur ¼ 0.996 Pur ¼ 0.958
10.0 0.090 e e
25.0 0.097 e e
30.0 0.102 0.101
40.0 0.110 0.107
50.0 0.119 0.121
60.0 0.132 0.135
70.0 0.145 0.149
80.0 0.162 0.169
90.0 0.180 0.192
100.0 0.213 0.227
a
Uncertainties u are u(T) ¼ 0.1
C, u(xS*) ¼ 0.002, u(gS) ¼ 0.002.
19
Activity coefficients of quaternary solutions sucrose/water/
glucose/fructose were calculated using the thermodynamic models
P&M-UNIQUAC, P&M-UNIFAC, Bio-UNIFAC, mS-UNIFAC and A-
UNIFAC. From these calculations, solubility values were obtained
within the consideration of impurities being composed of glucose
and fructose (considering sugar concentration presented in Table 7,
which keeps the proportions determined by HPLC). Figs. 5e7 show
the calculated solubility compared to experimental values obtained
in this work for solutions prepared, respectively, with refined sugar,
brown sugar and a mixture of brown and refined sugar.
Fig. 5 shows that some activity coefficients models from the
literature predict sucrose solubility with small deviations from
experimental data obtained in this study. A-UNIFAC and P&M-
UNIQUAC are the best choices for SLE calculation of aqueous solu-
tions of sucrose. Previous studies using these models for calcula-
tions of Vapor-Liquid Equilibrium (VLE) and SLE of aqueous fructose
solutions and with mixed water/ethanol solvents showed P&M-
UNIQUAC, mS-UNIFAC and A-UNIFAC were the best options for
equilibrium calculations with those solutions (Crestani et al., 2011,
2013). The results in Fig. 6 show a divergence not found in previous
studies when comparing experimental data of solubility to those
calculated with activity coefficient models. Such a divergence is less
significant in the results of Fig. 7, in which some of the models
predicted solubility values similar to experimental data. Mean
squared deviation between experimental and calculated values for
the best model (P&M-UNIQUAC) was 5.83  105 (mean percent
deviation of 4.58%) and 1.35  103 (16.98%), respectively, in data of
Figs. 7 and 6. These results suggest that the precision of calculations
using the consideration of impurities being composed of glucose
and fructose is limited. The estimation of sucrose solubility in
mixtures of sugars has been studied in the recent literature (Van
der Sman, 2017). As presented in Table 2, brown sugar (as well as
sugarcane syrup) presents a complex set of impurities, which
comprises not only other sugars but also salts and other substances.
It makes the efforts of estimating sucrose solubility in such solu-
tions quite complicated. For this purpose, the use of UNIFAC and
UNIQUAC based methods are limited and may require more suit-
able hypothesis than the consideration made in this research, with
a better description of impurities in this estimation. Therefore,
further and more detailed studies of sucrose SLE modelling with
activity coefficient models applied to impure aqueous solutions are
necessary.
Finally, a modification in Nývlt Equation (Nývlt et al., 1985) is
proposed here for calculating sucrose solubility in impure solu-
tions. Table 8 presents values of parameters of Equations (10) and
(11) fitted to experimental data. Fig. 8 shows the solubility calcu-
lated with proposed modification in the Nývlt Equation compared
s ) in solutions with different conditions of impurities (Pur stands for the molar
gexp
s
Pur ¼ 0.917 Pur ¼ 0.996 Pur ¼ 0.958 Pur ¼ 0.917
0.067 e e
0.107 e e
0.111 0.122 0.123 0.112
0.115 0.160 0.163 0.152
0.133 0.205 0.202 0.183
0.146 0.256 0.250 0.230
0.166 0.317 0.308 0.277
0.191 0.386 0.369 0.327
0.233 0.465 0.436 0.359
0.272 0.525 0.492 0.410
20 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23

0.30

0.25

Sucrose molar fracƟon

0.20

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 3. Experimental ( ) sucrose solubility in solutions prepared with brown sugar and its comparison with sucrose solubility curves calculated by different methods available in
the literature for sucrose from different sources: Georgieva et al. (2003), Frew (1973), Van der Poel et al. (1998), Rozsa (2000).

0.25

0.20
Sucrose molar fracƟon

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 4. Experimental ( ) sucrose solubility in solutions prepared with a mixture of brown and refined sugar and its comparison with sucrose solubility curves calculated by different
methods available in the literature for sucrose from different sources: Georgieva et al. (2003), Frew (1973), Van der Poel et al. (1998), Rozsa
(2000).

Table 7 proportional to the concentration of impurities in solution. Mean

Molar fraction of sucrose, glucose and fructose in the different sucrose solutions squared deviation between experimental and calculated data is of
used in ESL calculations.a
2.45  105 (mean percent deviation of 2.54%) considering all data.
Sugar Molar fraction This result shows that the Nývlt Equation, which is already widely
Sucrose Glucose Fructose used in solubility calculations, may be extended to calculations of
equilibrium in impure solutions or in any other solution in which
Refined 0.9962 0.0016 0.0022
Brown 0.9172 0.0352 0.0476 there is no control over concentration of one component in relation
Mixture 0.9578 0.0179 0.0243 to another (more than one solute for example). This extension
a
Uncertainties u are u(xS) ¼ u(xG) ¼ u(xF) ¼ 0.0001. makes this equation a very useful tool for crystallization studies in
which concentration varies over time.
to experimental values obtained in this work.
Results of Fig. 8 show a good correlation between the experi- 5. Conclusions
mental data and solubility calculated using the proposition of
modifying Nývlt Equation with the variable D, which is A reference curve was developed to provide the calculation of
 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23 21

0.30

0.25

Sucrose molar fracƟon 0.20

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 5. Experimental ( ) sucrose solubility and solubility curves calculated from activity coefficient models in solutions prepared with refined sugar: mS-UNIFAC,
A-UNIFAC, P&M-UNIQUAC, Bio-UNIFAC, P&M-UNIFAC.

0.30

0.25
Sucrose molar fracƟon

0.20

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 6. Experimental ( ) sucrose solubility and solubility curves calculated from activity coefficient models in solutions prepared with brown sugar: mS-UNIFAC,
A-UNIFAC, P&M-UNIQUAC, Bio-UNIFAC, P&M-UNIFAC.

solution concentration from refractive index and temperature
measured inline in pure and impure sucrose aqueous solutions. The
developed calibration curve is able to calculate solution concen-
tration with a mean percent deviation of 0.34%.
Sucrose solubility in pure and impure aqueous solutions was
determined over a temperature range from 10 to 100
C for pure
solutions and from 30 to 100
C for solutions containing typical
impurities of sugarcane syrup. These data, especially concerning
those of impure solutions, is scarce in the literature and, to the best
of the authors’ knowledge, there is no previous study encompass-
ing such a range of temperatures and impurities. These data are not
only useful but also essential for crystallization studies and
industrial processes improvement. The concentration of impurities
increases sucrose solubility in the studied range of temperatures
and concentration.
Comparing experimental sucrose solubility data to calculated
ones using models available in the literature allows for evaluating
the applicability of such models in sucrose solubility determination
from impure sugarcane syrups. The equation proposed by
Vavrinecz (1978) with parameters fitted by Rozsa (2000) resulted in
the best correlation between experimental and calculated data. In
addition, activity coefficient based models P&M-UNIQUAC, P&M-
UNIFAC, Bio-UNIFAC, mS-UNIFAC and A-UNIFAC were used to pre-
dict sucrose solubility from pure and impure aqueous solutions,
22 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23

0.30

0.25

Sucrose molar fracƟon

0.20

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 7. Experimental ( ) sucrose solubility and solubility curves calculated from activity coefficient models in solutions prepared with a mixture of brown and refined sugar:
mS-UNIFAC, A-UNIFAC, P&M-UNIQUAC, Bio-UNIFAC, P&M-UNIFAC.

Table 8 A model for calculating sucrose solubility in impure syrups was

Parameters of modified Nývlt Equation for sucrose solubility from impure solutions. proposed by modifying the Nývlt Equation of solubility. The pro-
Parameter Value Unit posed mathematical model was able to determine sucrose solubi-
A1 134.76 e
lity in the studied sugarcane syrups and in the range of
A2 5545.45 K temperatures and concentration of impurities studied, and pro-
A3 19.98 e vided calculations with the smallest deviation from experimental
N 23.03 e solubility values (mean squared deviation of 2.45  105 and
Mean quadratic deviation 2.45  105 e
percent deviation of 2.54%), when all evaluated models are
compared. It demonstrates that this model is a reliable equation to
be used in the mathematical modelling of crystallization from so-
considering impurities composed of fructose and glucose. A-UNI- lutions in which there is no control over components’ concentra-
FAC and P&M-UNIQUAC are the best choices for solubility calcu- tion. This model extends the application of the original Nývlt
lation of aqueous solutions of sucrose, although the activity Equation, and, as in the original equation, parameters may be fitted
coefficient based models did not provide predictions with small to any system.
deviations from experimental values when impurities are present.

0.30

0.25
Sucrose molar fracƟon

0.20

0.15

0.10

0.05

0.00
0 20 40 60 80 100
Temperature (°C)

Fig. 8. Experimental values of sucrose solubility ( stands for refined sugar, brown sugar and the mixture of both) compared to solubility curves calculated by the modified
Nývlt Equation ( ) in aqueous solutions with different concentrations of impurities.
 C.E. Crestani et al. / Journal of Food Engineering 218 (2018) 14e23
Acknowledgements
The authors would like to offer a special word of thanks to
Professor Marco Giulietti, who was the supervisor of this study and
is now in a better place. Thanks to PPG-EQ UFSCar and CAPES for
the support and PhD scholarship.
Funding
~o Paulo Research Foundation,
This work was supported by Sa
FAPESP (grant number 2011/51902-9).
References
Abed, Y., Gabas, N., Delia, M.L., Bounahmidi, T., 1992. Measurement of liquid solid-
phase equilibrium in ternary-systems of water sucrose glucose and water su-
crose fructose, and predictions with UNIFAC. Fluid Phase Equilibria 73, 175e184.
Apelblat, A., 1993. Activity and osmotic coefficients in electrolyte solutions at
elevated temperatures. AIChE J. 39 (5), 918e923.
Brown, R.J., Nees, A.R., 1933. Solubility of sucrose in beet house sirups. Ind. Eng.
Chem. 25 (5), 555e558.

, M., Dussap, C.G., Gros, J.B., 1995. A physical-chemical UNIFAC model for
Catte
aqueous-solutions of sugars. Fluid Phase Equilibria 105, 1e25.
Chen, J.C., Chou, C.C., 1993. Cane Sugar Handbook: a Manual for Cane Sugar Man-
ufacturers and Their Chemists, twelfth ed. John Wiley & Sons, Inc, United States
of America.
Crestani, C.E., Bernardo, A., Costa, C.B.B., Giulietti, M., 2011. Analysis of the influence
of fructose in Vapor-Liquid Equilibrium of ethanol and water solution (in Por-
tuguese). In: Proceedings of VI Congresso Brasileiro de Termodina ^mica Apli-
cada. CBTermo 2011, Salvador, BA, Brazil, November 22e25.
Crestani, C.E., Bernardo, A., Costa, C.B.B., Giulietti, M., 2013. Fructose solubility in
mixed (ethanolþwater) solvent: experimental data and comparison among
different thermodynamic models. J. Chem. Eng. Data 58, 3039e3045.
Dahlberg, H.W., Bennett, A.N., 1951. Effects of nonsugars on solubility of sucrose in
sugar beet sirups. Ind. Eng. Chem. 43 (3), 660e662.
Ferreira, O., Brignole, E.A., Macedo, E.A., 2003. Phase equilibria in sugar solutions
using the A-UNIFAC model. Ind. Eng. Chem. Res. 42, 6212e6222.
Frew, J.A., 1973. Optimal control of batch raw sugar crystallization. Ind. Eng. Chem.
Process Des. Dev. 12 (4), 460e467.
Georgieva, P., Meireles, M.J., Azevedo, F., 2003. Knowledge-based hybrid modelling
of a batch crystallisation when accounting for nucleation, growth and
agglomeration phenomena. Chem. Eng. Sci. 58, 3699e3713.
Kuramochi, H., Noritomi, H., Hoshino, D., Nagahama, K., 1997. Representation of
23
activity coefficients of fundamental biochemicals in water by the UNIFAC
model. Fluid Phase Equilibria 130, 117e132.
Larsen, B.L., Rasmussen, P., Fredenslund, A., 1987. A modified UNIFAC group-
contribution model for prediction of phase-equilibria and heats of mixing.
Ind. Eng. Chem. Res. 26, 2274e2286.
Martins, P.M., Rocha, F.A., Rein, P., 2005. Modeling sucrose evaporative crystalliza-
tion. Part 2. Investigation into crystal growth kinetics and solubility. Industrial
Eng. Chem. Res. 44 (23), 8865e8872.
Martins, P.M., Ferreira, A., Polanco, S., Rocha, F., Damas, A.M., Rein, P., 2009. Un-
steady-state transfer of impurities during crystal growth of sucrose in sugar-
cane solutions. J. Cryst. Growth 311, 3841e3848.
Myerson, A.S., 2001. Chapter 1: solution and solution properties. In: Handbook of
Industrial Crystallization, second ed. Elsevier Science and Technology Books,
pp. 12e16.
Nývlt, J., So€hnel, O., Matuchov
a, M., Broul, M., 1985. The Kinetics of Industrial
Crystallization. Elsevier.
Peres, A.M., Macedo, E.A., 1996. Thermodynamic properties of sugars in aqueous
solutions: correlation and prediction using a modified UNIQUAC model. Fluid
Phase Equilibria 123, 71e95.
Peres, A.M., Macedo, E.A., 1997. A modified UNIFAC model for the calculation of
thermodynamic properties of aqueous and non-aqueous solutions containing
sugars. Fluid Phase Equilibria 139, 47e74.
Rozsa, L., 2000. Sucrose solubility in impure cane sugar solutions. Int. Sugar J. 102
(1217), 230e235.
Sonego, J.L.S., Lemos, D.A., Rodriguez, G.Y., Cruz, A.G., Badino, A.C., 2014. Extractive
batch fermentation with CO2 stripping for ethanol production in a bubble
column bioreactor: experimental and modeling. Energy & Fuels 28 (12),
7552e7559.
Spiliotis, N., Tassios, D., 2000. A UNIFAC model for phase equilibrium calculations in
aqueous and nonaqueous sugar solutions. Fluid Phase Equilibria 173, 39e55.
Steindl, R.J., Broadfoot, R., Miller, K.F., 2001. Development and application of a
model for low-grade cooling crystallization. Proc. Int. Soc. Sugar Cane Technol.
24, 116e123.
Taylor, M., 1947. The solubility at high temperatures of pure sucrose in water.
J. Chem. Soc. 1678e1683.
Tsavas, P., Voutsas, E., Magoulas, K., Tassios, D., 2004. Phase equilibrium calculations
in aqueous and nonaqueous mixtures of sugars and sugar derivatives with a
group-contribution model. Ind. Eng. Chem. Res. 43, 8391e8399.
UNICA, 2016. Uni~ ao da Indústria de Cana-de-Açúcar. http://www.unica.com.br.
(Accessed 12 May 2016).
Van der Poel, P.H., Schiweck, H., Schwartz, T., 1998. Sugar Technology: Beet and
Cane Sugar Manufacture. Dr. Albert Bartens, Berlin.
Van der Sman, R.G.M., 2017. Predicting the solubility of mixtures of sugars and their
replacers using the FloryeHuggins theory. Food Funct. 8, 360e371.
Vavrinecz, G., 1978. Formation and composition of beet molasses, I- the equation for
solubility. Sugar Tech. Rev. 6, 117e129.