You are on page 1of 9

Journal of Food Engineering 77 (2006) 126–134

www.elsevier.com/locate/jfoodeng

Modeling temperature dependence of honey viscosity and of


related supersaturated model carbohydrate systems
M.P. Recondo, B.E. Elizalde, M.P. Buera *

Departamento de Industrias, Facultad Ciencias Exactas y Naturales, Universidad de Buenos Aires, 1428 Ciudad Universitaria, Buenos Aires, Argentina

Received 21 March 2005; accepted 18 June 2005


Available online 19 August 2005

Abstract

The viscosities of a unifloral honey and supersaturated sugar solutions were measured between 5 and 70 C. All systems exhib-
ited Newtonian behavior with reducing viscosity as increasing temperature. Four models (Arrhenius, VTF, WLF and Power Law)
were investigated to describe the temperature dependence of viscosity. Among the different ways of using the WLF model, the
method of reduced variables was the most suitable way to calculate coefficients. Oppositely, the WLF with ‘‘universal coefficients’’
badly predicted the temperature dependence of viscosity.
When the calculated and experimental points were plotted as a function of (T  Tg), WLF (with coefficients calculated by the
reduced variables method), VTF and power law models fitted the experimental data in a better trend than the Arrhenius equation.
Also, the extrapolation of fitted curves into the glass transition region, showed that the Arrhenius model predicts the lowest viscosity
values, while the WLF model (with coefficients calculated by the reduced model method) predicts the highest viscosity values in that
region. VTF and Power Law models provided curves with intermediate solutions between Arrhenius and WLF model.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Viscosity; Honey; Sugar systems; WLF; Arrhenius; VTF; Power Law models; Glass transition

1. Introduction Tolstoguzov, 2000; Roos, 1995; Slade & Levine, 1991).


The primary property that describes mechanical relaxa-
Dynamic behavior of supersaturated or supercooled tion of mobile components is viscosity (g).
aqueous sugar solutions has important implicances in The simplest equation to describe temperature depen-
technological processes. In the research area, they exhi- dence of viscosity is the Arrhenius equation. However,
bit interesting properties that are of difficult experimen- sugar–water systems show a departure of this model in
tal access in the glass transition region (Angell et al., the proximities to the glass transition temperature
1994). (Tg), and therefore they can be classified as ‘‘fragile’’
Particularly, there has been increasing interest in dif- glass-forming liquids (Angell et al., 1994). The devia-
fusion and relaxation phenomena in the rubbery and tions from the Arrhenius behavior may be accounted
glassy states of food systems and components, due to by several semiempirical models, such as Williams–
their influence in processing, storage and handling Landel–Ferry (WLF) (Willams, Landel, & Ferry,
(Bhandari, DÕArcy, & Kelly, 1999; Blanshard & Liliford, 1955), Vogel–Taumman–Fulcher (VTF) (Angell et al.,
1993; Kerr, Lim, & Reid, 1993; Matveev, Grinberg, & 1994) and Power Law (Hill & Dissado, 1982).
The WLF equation was frequently applied in poly-
*
Corresponding author. Tel.: +54 11 4794 3344; fax: +54 11 4576
mer science, employing Tg as a reference temperature
3366. and fixed constants (‘‘universal constants’’: C1 and C2)
E-mail address: pilar@di.fcen.uba.ar (M.P. Buera). obtained by averaging values for several polymers.

0260-8774/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jfoodeng.2005.06.054
M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134 127

VTF and WLF equations are interconvertible, but Supersaturated solutions of glucose, fructose, glu-
Angell et al. (1994) indicated that VTF is more useful cose, and glucose/fructose (1.22) with as 81% w/w were
than the WLF form in sugar systems, since it has coeffi- prepared by addition of appropriate amounts of re-
cients that can be considered ‘‘more universal’’ to be ap- agents, up to complete dissolution in distilled water at
plied in different systems, under various environmental 70 C. D (+) glucose, D () fructose (Merck A.G.,
conditions. Darsmstadt, West Germany) were used to prepare
The WLF equation has the advantage that reference model systems.
temperature Tg, can be determined experimentally by
techniques such as differential scanning calorimetry. 2.2. Differential scanning calorimetry (DSC)
However, the determination of viscosity close to Tg is
of difficult experimental access, since it is extremely high The glass transition temperature (Tg) and change of
(gg = 1015 mPa s). It is thus necessary to define another specific heat capacity at the Tg region (DCp) were
reference temperature within the experimental range. determined using a Mettler T-A 4000 Thermal Analysis
Then in systems where gg is unknown or cannot be Systems with a TC11 processor and graph-ware TA72-
reliably measured, C1 and C2 are unknown and there thermal analysis software. The instrument was cali-
doubt whether the WLF model itself is at all applicable, brated with indium.
the use of another reference temperature is convenient A scanning rate of 10 K/min. was used to analyze
(Nelson & Labuza, 1994; Peleg, 1992). samples (10–25 mg in hermetically sealed 40 lL alumi-
There have been several studies that have looked at num pans; Mettler) and an empty pan was used as a ref-
the viscosity–temperature relationship in aqueous sugar erence. The scans were performed from 150 to 120 C
solutions, and allow experimentally determined and pre- and the onset temperature of glass transition was used
dicted viscosity by different models of sugar solutions to as Tg. Reported data are the average of three
be compared (Kerr & Reid, 1994; Maltini & Anese, determinations.
1995; Maltini & Manzocco, 1997). However these stud-
ies evaluated viscosities of sugar amorphous phases in 2.3. Steady state rheology
partially frozen systems or viscosity of liquid super-
cooled pure sugar systems. The viscosity of honey and related supersaturated su-
There is a lack of data on supersaturated aqueous su- gar systems was determined using a rotating viscometer
gar systems and in naturally occurring sugar mixtures (Brookfield LVT). Viscosity measurements were taken
such as honey. Honey is a supersaturated glucose– at several rotational speeds. Up to three different spindles
fructose solution of high viscosity; containing several sizes were required to cover the whole range of viscosity
enzymes and other carbohydrates (sucrose, dextrins) values. The cell containing samples was thermostated by
that can complement the current trend in cryopreser- a circulating fluid cooled by a cryostat (Forma Scientific,
vation of multicomponent systems and that are of great Model 2095/2100, Mallinckrodt, Inc., Ohio, USA).
importance for several applications. Sopade et al. (2002) Measurements were taken at temperatures ranging from
reported that WLF equation was suitable for modeling 5 to 67 C. Once the samples were introduced in the
the temperature dependence of viscosity in honey. They cell, a 30 min period was allowed to elapse before mea-
obtained C1 and C2 and gg by a nonlinear regression surements were made. Similar equilibration periods were
analysis and employed Tg as reference temperature. allowed when the sample temperature was changed.
However they recognized that the underlying assump- Sample temperature was monitored with an Hg ther-
tions demanded careful evaluation. mometer, graduated in 0.001 C intervals, immersed in
In this work the viscosities of honey and supersatu- the cell and removed just before viscosity measurements.
rated sugar systems of similar composition were mea- Reported data are the average of three determinations.
sured as a function of temperature. WLF, Arrhenius, Viscosity data for glucose were obtained from Hand-
VTF, and Power Law equations were compared for book of Sugars (Harry & Ray Junk, 1980, chap. 11).
their ability to describe experimental data.
2.4. Mathematical models

2. Experimental Four different models were used to study the temper-


ature dependence of viscosity: Arrhenius, WLF, VTF
2.1. Materials and preparation of models systems and Power Law.

A natural unifloral honey which botanical origin was 2.4.1. Arrhenius model
Bulnesia sarmientoi (algarrobo) from the province of  
Chaco (Argentina) was investigated. The moisture con- Ea
g ¼ g0  exp  ð1Þ
tent was of 17.15%. RT
128 M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134

where g0 is a constant, R is the gas constant and Ea is an 2.4.3. Vogel–Taumman–Fulcher model (VTF)
energy barrier to flowing. The VTF equation can be expressed for viscosity as
The logarithmic form of Eq. (1) allows to determine  
parameters involved and fitting of a particular set of B
g ¼ g1  exp ð9Þ
data. T  T0

where g1 is the viscosity at T = 1 and T0 is the temper-


2.4.2. Willams, Landel and Ferry (WLF)
ature at which the relaxation time relevant to molecular
!
displacements becomes infinite. The value T0 was fixed
g C 1 ðT  T g Þ
log ¼ ð2Þ at 184 K, which was estimated from the data reported
gg C 2 þ ðT  T g Þ
by Parker and Ring (1995) for aqueous sugar systems
The WLF equation (Soesanto & Willams, 1981; Willams of similar concentration, and B was calculated as the
et al., 1955) relates the viscosity (g) at T with the viscos- slope of the linearized form of Eq. (9).
ity (gg) at Tg, as a function of the difference between T
and Tg. 2.4.4. Power Law
Ferry (1980), one the authors of the WLF model, ad- Alternatively, Hill and Dissado (1982) formulated a
vised against the use of ‘‘universal constants’’ coeffi- Power Law description of relaxation:
cients (C1 and C2), obtained as average values from g ¼ jðT  T g Þm ð10Þ
several polymers. He stated that it was convenient
to determine the real constants for each system. The where j and m are constants. The values of j and m
method of reduced variables proposed by Peleg (1992) were estimated from linearization of Eq. (10).
was found to be an adequate procedure to calculate
the WLF constants in real systems, avoiding extrapola- 2.5. Statistical analysis
tions and on the basis of physical approaches. In this
method a reference temperature (Tr) is chosen within Standard statistical packages (Graph Pad Prism
experimental range and then Eq. (2) becomes: Version 3.1 Software and Excel 2000) were used for
  relevant analysis. Linear regression was performed for
g C 1r ðT  T r Þ
log ¼ ð3Þ all models, except for WLF models, where a nonlinear
gr C 2r þ ðT  T r Þ
regression was applied.
where Tr is the reference temperature, gr is the viscosity All determinations were performed in triplicate and
at Tr and C1r and C2r are constant coefficients. To eval- the average value was reported.
uate C1r, and C2r a temperature at which the viscosity
can be considered infinite (T1) has to be defined. As it
can be assumed that virtually no long-range molecular 3. Results
movements are observed below Tg, a very good approx-
imation for this value is: 3.1. Thermal properties
T 1 ¼ T g  50 K
The DSC curves recorded during heating in the dy-
This assumption has probed to confirm experimental namic runs, of supersatured sugar solutions and honey
evidences (Buera & Karel, 1993; Mazzobre, Soto, samples are shown in Fig. 1. They evidenced baseline
Aguilera, & Buera, 2001; Peleg, 1992). At this temperature shifts. The temperature shift for honey was at 43 C,
C 2r ¼ T r  T 1 ð4Þ and the DCp was of 0.89 J/g K. These values were in
the range of previously reported Tg and DCp values for
When Eq. (3) is combined with Eq. (4) it follows:
honey (Rubin, Wasylyk, & Baust, 1990; Sopade, Bhan-
 
g C 1r ðT  T r Þ dari, Halley, DÕArcy, & Caffin, 2001; Zoltán, Pitsi, &
log ¼ ð5Þ Thoen, 1999). In addition these values are coincident
gr ðT  T 1 Þ
with those predicted values from Couchman–Karasz
A plot of log(g/gr) vs. ((T  Tr)/(T  T1)) is linear equation (Couchman & Karasz, 1978) for ternary sys-
through the origin if Tr is correctly chosen. Then, to tems. Similarly, the values of Tg and DCp found for su-
evaluate C1r and C2r when Tr = Tg, the difference be- gar solutions were coincident with previous results
tween Tr and Tg is defined (Mazzobre et al., 2001; Peleg, (Roos & Karel, 1991).
1992) as:

d ¼ Tr  Tg ð6Þ 3.2. Rheological properties


C 1 ¼ ðC 1r  C 2r Þ=ðC 2r  dÞ ð7Þ
The viscosity of honey and supersaturated sugar solu-
C 2 ¼ C 2r  d ð8Þ tions change slightly (coefficient of variation 6 10%) with
M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134 129

Fig. 1. DSC heat flow curves of sugar model systems and honey.

increasing shear rate at all temperatures (data not shown), whether to make them ‘‘universal’’ constants or to
but it substantially decreased as temperature was increa- allow them to vary for a good fit to the experimental
sed. This is expected because honey are Newtonian liq- data (Peleg, 1992; Roos, 1995). In addition, there are
uids and their viscosity value is highly sensitive to temper- uncertainties, in most practical circumstances, related
ature (Junzgeng & Changying, 1998; Mossel, Bhandari, to the location of the reference Tg and the values of
DÕArcy, & Caffin, 2000; Munro, 1943; Sopade et al., viscosity at Tg (gg). Then, with the aim to provide a
2002; White, 1978). The results obtained for sugar solu- contribution to these issues, the fitting of experi-
tions were similar to those reported for supercooled liquid mental data by the WLF model, was performed by
pure glucose and fructose systems (Ollet & Parker, 1990). nonlinear regression analysis in different forms:

3.3. Comparison between the different models to predict (a) Fixing C1 and C2 to the ‘‘universal values’’.
temperature dependence of viscosity (b) Fixing C1 and C2 to the ‘‘universal values’’ and set
Tg at experimental value.
There are several conflicting reports on how to handle (c) Fixing C1 and C2 to the coefficients calculated by
the constants C1 and C2 in the WLF model (Eq. (2)); the reduced model method (Peleg, 1992).

Table 1
Parameters calculated by fitting experimental data through different forms of WLF model
Models Obtained parametera Honey Fructose Fructose–glucose Glucose
14 14 14
Universal coefficients (a) gg (mPa s) (4.6 ± 0.2) · 10 (4.1 ± 0.3) · 10 (4.1 ± 4.5) · 10 (3.5 ± 1) · 1014
Tg C) 59.7 ± 1 61.8 ± 2 62.5 ± 2 83.9 ± 1.2
R2 0.997 0.993 0.990 0.999
Universal coefficients and gg (mPa s) (4.4 ± 0.4) · 1013 (1.8 ± 0.1) · 1013 (1.5 ± 0.05)1014 (7.1 ± 0.4) · 1014
experimental Tg (b) R2 0.899 0.925 0.972b 0.896
Reduced coefficientsc(c) gg (mPa s) (7.4 ± 0.4) · 1011 (7.9 ± 0.6) · 1011 (1.3 ± 0.6) · 1013 (4.1 ± 0.1) · 1011
Tg (C) 45.2 ± 1 47.3 ± 1.4 54.05 ± 1.8 62.6 ± 0.7
R2 0.997 0.991 0.990 0.9999
Reduced coefficientsc and gg (mPa s) (5.6 ± 0.1) · 1011 (7.9 ± 0.2) · 1011 (1.3 ± 0.03) · 1013 (1.3 ± 0.003)1011
experimental Tg (d) R2 0.997 0.991 0.990 0.999
Experimental Tg and C1 15.7 ± 0.2 16.10 ± 0.2 16.75 ± 0.25 16.80 ± 0.02
gg = 1015 mPa s (e) C2(K) 27.0 ± 1.2 29.95 ± 1.3 37.53 ± 1.7 31.26 ± 0.14
R2 0.989 0.985* 0.988 1
Experimental Tg (f) C1 13.4 ± 0.4 16.52 ± 2 15.45 ± 1.68 17.24 ± 2.46
C2(K) 88.3 ± 15 171.5 ± 51 118 ± 60 29.91 ± 7.26
gg (mPa s) (4.7 ± 2) · 109 (4.7 ± 3.7) · 108 (7 ± 9)109 (3 ± 7)1015
R2 0.988 0.996 0.992 1
a
± Standard deviation is indicated for the parameters.
b
Significant deviation from model.
c
Coefficients calculated by the reduced model method (Table 2).
130 M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134

(d) Fixing C1 and C2 to the coefficients calculated by


the reduced model method and set Tg at experi-

C2 (K)
50 ± 1
50 ± 1
50 ± 1
50 ± 1
mental value.
(e) Fixing gg at 1015 mPa s.
(f) Calculating C1, C2 and gg using Tg as reference

13.8 ± 0.6
13.9 ± 0.7
15.4 ± 0.8
13.4 ± 0.3
Parameters calculated through the linearization of Arrhenius, VTF, WLF (reduced variables) and Power Law models, for honey, fructose, fructose–glucose and glucose systems
temperature.

C1
There are other possible ways to obtain the WLF

0.996
0.990
0.990
0.999
parameters (C1, C2, gg and Tg), by nonlinear multiple

R2
correlation analysis, allowing all parameters to vary.
This method was not shown here because of the high

5.66 ± 0.09

5.76 ± 0.16
3.95 ± 0.03
5.57 ± 0.7
uncertainty of the obtained parameters, and of the lack
of any physical value for those variables. Table 1 and

WLF
C1r
Fig. 2a (honey) and b (fructose–glucose systems) show
the result of these fittings.

0.998
0.993
0.991
0.999
It can be observed that of the several ways to

2
R
manipulate the WLF equation, the highest deviation

7.75 ± 0.2
from experimental values was obtained employing the

7.5 ± 0.1

8.4 ± 0.2
6.5 ± 0.1
m
6.5

17

1.2 ± 0.5) · 1018

(1.2 ± 0.3) · 1015


(2.9 ± 0.53) · 10
6.0

(9.5 ± 4) · 1019
Experimental
(a) universal coefficients

Power Law
5.5 (b) Tg = –43 and universal coefficients

j (mPa s)
(c) reduced coefficients
(d) reduced coefficients, Tg = –43˚C
5.0
(e) Tg = –43˚C and log ηg =15
(f) Tg = –43˚C
log η (m Pa.s)

4.5

0.996
0.988
0.988
4.0

2
R

1
3.5

2

(2.2 ± 0.7) · 102


(2.5 ± 0.9) · 102
3.0

(1.7 ± 1) · 102
2.5 (1.4 ± 0.3) · 10
g1 (mPa s)

2.0

1.5
-10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
(a) T (°C)
1391 ± 37
1535 ± 26

1379 ± 43
1190 ± 4

5.5
B (K)
VTF

5.0 Experimental
(a) universal coefficients
0.994
0.996
0.991
0.998
± Standard deviation is indicated for the parameters.

(b) universal coeficients , Tg = –53.9˚C


2
R

(c) reduced coefficients


4.5
(d) reduced coefficients, Tg = –53,9ºC
(e ) η g = 1015, Tg = –53, 9ºC
11

(3 ± 1) · 1011
(5 ± 4) · 1011
(6 ± 5) · 107

(f) Tg = –53, 9ºC


log η (m Pa.s)

(2 ± 1) · 10

4.0
g0 (mPa s)

3.5
Ea (kJ/mol)

3.0
Arrhenius

82.8 ± 2
81.3 ± 2
79.9 ± 2
50.3 ± 2

2.5
Fructose–Glucose

2.0
-3 2 7 12 17 22 27 32 37 42 47 52 57
(b) T (°C)
Fructose
Models*

Glucose
Table 2

Fig. 2. Experimental viscosity data and different WLF fits calculated


System
Honey

by nonlinear regression as a function of temperature for (a) honey and


*

(b) fructose–glucose system.


M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134 131

universal constants and the experimental Tg value (fit b, of the ‘‘universal’’ ones and also similar to those calcu-
full line in Fig. 2a and b). When the universal constants lated by the reduced model method (Table 2). However,
were employed, but the Tg values were allowed to vary, C2 coefficient values were significantly different to the
a better fitting was obtained, although the predicted Tg ‘‘universal’’ and ‘‘reduced’’ coefficients and also presented
values were far from the real ones (Table 1). A superfi- high uncertainty. The predicted viscosity values (gg) were
cial analysis of Fig. 2a and b may lead to the conclusion three to four orders of magnitude lower than 1015 mPa s,
that the fitting methods (a), (c), (d), (e) and (f) are suggested for sugar glass systems (Soesanto & Willams,
equally suitable to describe the experimental behavior. 1981) (Table 1), except for the glucose system.
However, they differ in their ability to approximate mea- Finally, when gg was fixed at 1015 mPa s, the model
sured/expected values, and their validity can be con- presented a high deviation from experimental data
firmed by the calculation of expected/measured gg and (lower R2), except the glucose system (Table 1).
Tg values, as discussed in next paragraphs. The WLF The higher deviations between calculated and experi-
fitting with the coefficients calculated by the reduced mental values obtained with method (b) (universal coef-
model method (c) was compared to that employing ficients, Tg fixed to the experimental values) are evident
‘‘universal values’’ (a) in their capability to estimate in Fig. 2a and b. However, although methods (a, c, d, e,
the Tg values. Although correlation coefficients were f) seem to be equally suitable to fit experimental values
similar, Table 1 shows that the model with ‘‘universal as could be concluded from the graphs, the analysis of
values’’ (a) predicts high Tg values in comparison with the obtained gg and/or Tg values that are predicted
the experimental values, whereas the estimated Tg values through them allowed to discriminate and verify their
by the model with the coefficients calculated by the re- validity. This analysis indicated that the calculation of
duced model method (c), were close to the experimental C1 and C2 coefficients through the reduced variables
values. method was the best way to describe temperature depen-
When the fitting of experimental data was performed dence of viscosity by the WLF equation.
by nonlinear regression with experimental Tg as the only Arrhenius, VTF, WLF (with reduced variables) and
fixed parameter (f), C1 coefficients values were in the order Power Law models were then compared, to describe

3
Experimental
12.5 Experimental
linear regression linear regression
2
lnη(m Pa.s)

10.0
1
log(η/η r)

0
7.5

-1

5.0
-2
-0.5 -0.4 -0.3 -0.2 -0.1 -0.0 0.1 0.2 0.3 0.00300 0.00325 0.00350 0.00375 0.00400
(a) (T-Tr)/(T-T ∞ ) (b) 1/T(K-1)

6.0
14
Experimental 5.5 Experimental
linear regression linear regression
5.0
11
log η(m Pa.s)

4.5
lnη(m Pa.s)

4.0
8
3.5

3.0
5

2.5

0.00625 0.00750 0.00875 0.01000 0.01125 0.01250 2.0


(c) 1.5 1.6 1.7 1.8 1.9 2.0 2.1
1/(T-T0 ) (K-1) (d)
log (T-T g)

Fig. 3. Experimental viscosity data and linear regression fits calculated by (a) WLF, (b) Arrhenius, (c) VTF and (d) Power Law models for honey
system.
132 M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134

the temperature dependence of viscosity with tempera- (glucose); m ranged from 7.75 (fructose) to 6.5 (glucose).
ture. Fig. 3a–d shows experimental data and regression Both values were in the same order of that reported by
lines obtained for honey system. In the analyzed temper- Kerr and Reid (1994) for glucose, sucrose and malto-
ature ranges, the experimental values could be ade- dextrins, but they were various orders of magnitude
quately described through all the analyzed models. higher those than reported for Australian honeys
The sugar model systems presented a similar behavior (Sopade et al., 2002).
(data not shown). Table 2 shows the calculated coeffi- Although the R2 values obtained for all systems were
cients and goodness of fit (R2). It can be observed that higher that 0.98, independently of model used (Table 2),
the values of C1, calculated for the WLF model with when the models and experimental points, were plotted
the reduced model method, were approximately of same as function of (T  Tg), (Fig. 4a and b), it can be seen,
(13) for all systems except for fructose–glucose system, that WLF, VTF and Power Law models fitted the exper-
that presented a significantly different value (15). imental value in a better way. The extrapolation of fitted
Although C1 values were in the order of the ‘‘universal’’ curves into the glass transition region, also showed, that
C1 coefficient (17.44), they were significantly different the Arrhenius model predicts the lowest viscosity values,
from it. Kerr and Reid (1994) and Nelson and Labuza while the WLF model (with coefficients calculated by the
(1994) reported similar C1 values for honey and sugar reduced model method) predicts the highest viscosity
systems respectively. However, both of them evaluated values close to Tg. The VTF and Power Law models
C1 and C2 through a nonlinear regression of WLF equa- provide curves with intermediate solutions between
tion, using Tg as the reference temperature, with gg al- Arrhenius and WLF model (Fig. 4a and b).
lowed to vary. In the reduced variable method Table 3 shows the gg (glass viscosity) calculated for
employed C2 coefficient fixed at 50 K for all systems all systems studied through the different models,
(Peleg, 1992), which is close to the ‘‘universal value’’ employing the experimentally determined Tg values. It
(51.6 K). is to be noted that gg is not mathematically defined in
The Ea values, calculated by the Arrhenius model
were close to 80 kJ/mol, except for glucose system that
presented a value significantly lower (50.3 kJ/mol). The 16
Experimental
Ea value obtained for honey, was in the order of that re-
Arrhenius
ported for Australian honey (Sopade et al., 2002). The 14
VTF
value for glucose was similar to that reported for dehy- WLF (reduced cofficients)

drated lactose systems (Mazzobre et al., 2001) but three- 12 Power Law
ln η (mPa.s)

fold lower than that found by Kerr and Reid (1994).


This difference could be attributed to the fact that in 10
the last work the measurements were performed in a sys-
tem totally different (sugars in coexistence with ice) and 8
the Ea associated with viscous drag molecules depends,
upon both the types of molecules involved and of their 6
relative concentration.
30 40 50 60 70 80 90 100 110
For the VTF model the value of constants B were in
(a) T-Tg (°C)
the same order (1200–1500 K) than those reported for
viscosity at different temperatures of sucrose (Parker & 15.0
Experimental
Ring, 1995), fructose and glucose (Ollet & Parker, Arrhenius
1990) and crystallization kinetics of lactose and lac- VTF

tose–trehalose systems (Mazzobre et al., 2001). Sopade 12.5 WLF (reduced cofficients)

et al. (2002) found B values in the range from 4 to Power Law


ln η (mPa.s)

13 K, for Australian honey. This difference is due to


the fact that these authors used the value of Tg for the 10.0

constant T0. The g1 values calculated for the VTF


model, were close to 102 mPa s and similar to that re-
ported for Parker and Ring (1995) and Ollet and Parker 7.5

(1990) for sugar systems. In addition these results


confirm Angell et al.Õs hypothesis (1994), that for sugar
systems g1 is a universal value and corresponds a relax- 5.0
40 50 60 70 80 90 100 110
ation time of the 100 s at T = 1. (b) T-Tg (°C)

The derived constants estimated from the lineariza- Fig. 4. Experimental data and Arrhenius, WLF, VTF and Power Law
tion of Power Law (Eq. (10)) show that j constant ran- plotted as a function of (T  Tg) for (a) honey and (b) fructose–glucose
ged from 2.9 · 1018 mPa s (fructose) to 1.2 · 1015 mPa s systems.
M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134 133

Table 3 ficients calculated by the reduced variables method


Predicted gg values determined through Arrhenius, VTF and WLF showed to be suitable for modeling the temperature
(reduced variables) models
dependence of viscosity.
System gg (mPa s) As discussed before, in the analyzed temperature
WLF VTF Arrhenius range, all the models can be employed quite satisfacto-
Honey 5.62 · 1011
4.3 · 1012
1.5 · 108 rily to describe the experimental behavior. However,
Fructose 7.9 · 1011 5.3 · 1012 1.9 · 109 no extrapolation can be made out of that particular
Fructose/glucose 1.3 · 1013 2.8 · 1015 5.8 · 108 range for which the models coefficients were calculated.
Glucose 1.3 · 1011 2.5 · 1015 1.2 · 106
The Arrhenius model predicts the lower temperature
dependence, while the WLF predicts the higher temper-
ature dependence close to the Tg value. The VTF and
the Power Law model. As discussed when analyzing Power Law models provide curves whit intermediate
Fig. 4a and b, WLF and VTF models predicted gg values solutions between Arrhenius and WLF. However, the
substantially higher than the Arrhenius model (about six Power Law model is mathematically undefined for
orders of magnitude) for all the systems. The gg values T = Tg and, besides its coefficients lack of any physical
shown in Table 3, were similar to that found by extrap- meaning. Thus, VTF model seems to be the best alterna-
olation by Sopade et al. (2002) when analyzing viscosity tive to predict viscosity values if any extrapolation has
of Australian honeys and for viscosity of supercooled to be performed. In addition, VTF was proposed by An-
fructose and glucose systems (Ollet & Parker, 1990). It gell et al. (1994) as the best model to fit temperature
is commonly believed that the gg should have a constant dependence of relaxation phenomena in sugar systems.
value close to 1015 mPa s, thus characterizing the glassy
state and its interpretation as condition of limiting free
volume fraction. However, even if glassy state is Acknowledgments
achieved for all systems at an average free volume frac-
tion of about 0.025, the distribution free volume fraction The authors acknowledge financial support from
can be altered in mixtures and can differ for molecules of Universidad de Buenos Aires (Project EX 274 and EX
different geometry. This affects the local mobility of mol- 226), CONICET (PIP 2734).
ecules and leads to variations of viscosity with composi-
tion even in the glassy state (Soesanto & Willams, 1981).
The results of Table 3, confirm this hypothesis. In fact, References
the gg predicted by the WLF equation, varied according
to the systems and was lower to 1015 mPa s. Angell, C. A., Bressel, D. R., Green, J. L., Kanno, H., Oguni, M., &
Sare, E. J. (1994). Liquid fragility and the glass transition in water
aqueous solutions. Journal of Food Engineering, 22, 115–142.
Bhandari, B., DÕArcy, B., & Kelly, C. (1999). Rheology and crystal-
4. Conclusions lization kinetics of honeys. International Journal Food Properties, 2,
217–226.
As expected, honey and supersaturated sugar solu- Blanshard, J. M. V., & Liliford, P. J. (1993). The glassy state in foods.
Loughborough, UK: Nottingham University Press.
tions exhibited Newtonian behavior and inverse rela-
Buera, M. P., & Karel, M. (1993). Application of the WLF equation to
tionship of viscosity with temperature. The WLF describe the combined effects of moisture and temperature on non-
coefficients calculated by the method reduced variables enzymatic browning en foods. Journal of Food Processing and
were significantly different from the ‘‘universal values’’, Preservation, 17, 31–45.
but close to them. Couchman, P. R., & Karasz, F. E. (1978). A chemical thermodynamic
discussion of the effect of composition in glass transition temper-
The VTF, Power Law and WLF (with coefficients
atures. Macromolecules, 11, 117–119.
calculated by the reduced variables method), models fit- Ferry, J. D. (1980). Viscoelastic properties of polymers. New York:
ted the experimental values in a better way than the John Wiley, pp. 264–272.
Arrhenius model within the studied experimental range Harry, M.P. & Ray Junk, W. (1980). Handbook of sugars (2nd ed., p.
of temperature. The values of gg, obtained by extrapola- 271). Westport, CT: Avi Publishing Company, Inc.
Hill, R. M., & Dissado, L. A. (1982). The temperature dependence of
tion to Tg through WLF, VTF and power law models
relaxation processes. Journal of Physics C. Solid State Physics, 15,
were different from the constant value (1015 mPa s) gen- 5171–5193.
erally accepted for the glassy state and depended on the Junzgeng, P., & Changying, J. (1998). General rheological model for
system. natural honeys in China. Journal of Food Engineering, 36, 165–168.
The use of the WLF equation with ‘‘universal coeffi- Kerr, W. L., & Reid, D. S. (1994). Temperature dependence of the
viscosity of sugars and maltodextrins in coexistence with ice.
cients’’ may lead to considerable error in the viscosity
Lebensmittel-Wissenchaft und -Technologie, 27, 225–231.
value calculated at a given temperature, particularly Kerr, W. L., Lim, M. H., & Reid, D. S. (1993). Chemical kinetics in
when employed to calculate the viscosity value at Tg, relation to glass transition temperature in frozen food polymers
through extrapolation. Oppositely, the WLF with coef- solutions. Journal of Food Science and Agriculture, 61, 512–526.
134 M.P. Recondo et al. / Journal of Food Engineering 77 (2006) 126–134

Maltini, E., & Anese, M. (1995). Evaluation of viscosities of Roos, Y., & Karel, M. (1991). Phase transition of mixtures of
amorphous phases in partially frozen systems by WLF kinetics amorphous polysaccharides and sugars. Biotechnology Progress, 7,
and glass transition temperatures. Food Research International, 28, 49–53.
367–372. Roos, Y. H. (1995). Phase transition in foods (Chapters 2, 5, 7). New
Maltini, E., & Manzocco, L. (1997). Experimental and predicted York: Academic Press.
viscosities of model solutions at temperatures above the glass Rubin, C. E., Wasylyk, J. M., & Baust, J. G. (1990). Investigation of
transitions. Proceedings of ISOPOW, 7, 57–60. vitrification by nuclear magnetic resonance and differential mag-
Matveev, Y. I., Grinberg, V. Y., & Tolstoguzov, V. B. (2000). The netic resonance and differential scanning calorimetry in honey and
plasticing effect of water on proteins, polysaccharides and their model carbohydrate systems. Journal of Agriculture and Food
mixtures. Glassy state of biopolymers, food and seeds. Food Chemistry, 38, 1824–1827.
Hydrocolloids, 14, 425–437. Slade, L., & Levine, H. (1991). Beyond water activity: recent advances
Mazzobre, M. F., Soto, G., Aguilera, J. M., & Buera, P. (2001). on an alternative approach to the assessment of food quality and
Crystallization kinetics of lactose in systems co-lyophilized with safety. Critical Reviews in Food Science and Nutrition, 30, 115–360.
trehalose. Analysis by differential scanning calorimetry. Food Soesanto, T., & Willams, M. C. (1981). Volumetric interpretation of
Research International, 1–9. the viscosity for concentrated and dilute sugars solutions. Journal
Mossel, B., Bhandari, B., DÕArcy, B., & Caffin, N. (2000). Arrhenius of Physics Chemistry, 85, 3338–3341.
model to predict rheological behavior in some Australian honeys. Sopade, P. A., Bhandari, B., Halley, P., DÕArcy, B., & Caffin, N.
Lebensmittel-Wissenchaft und -Technologie, 33, 545–552. (2001). Glass transition in Australian honeys. Food Australia, 53,
Munro, J. A. (1943). The viscosity and thixotropy of honey. Journal of 399–404.
Economic Entomology, 36, 769–771. Sopade, P. A., Halley, P., Bhandari, B., DÕArcy, B., Doebler, C., &
Nelson, K. A., & Labuza, T. P. (1994). Water activity and food polymer Caffin, N. (2002). Application of the Willam–Landel–Ferry model
science: implications of state on Arrhenius and WLF models to the viscosity-temperature relationship of Australian honeys.
in predicting shelf life. Journal of Food Engineering, 22, 271– Journal of Food Engineering, 56, 67–75.
289. White, J. W. (1978). Honey. Advances in Food Research, 24, 287–374.
Ollet, A. L., & Parker, R. (1990). The viscosity of supercooled fructose Willams, M. L., Landel, R. F., & Ferry, D. H. (1955). Temperature
and its glass transition temperature. Journal of Texture Study, 21, dependence of relaxation mechanisms in amorphous polymers and
355–362. other glass-forming liquids. Journal of the American Chemical
Parker, R., & Ring, S. G. (1995). A theoretical analysis of diffusion Society, 77, 3701–3706.
controlled reactions in frozen solutions. Cryoletters, 16, 197– Zoltán, K., Pitsi, G., & Thoen (1999). Glass transition temperature of
208. honey as a function of water activity as determined by differential
Peleg, M. (1992). On the use of the WLF model in polymers and foods. scanning calorimetry. Journal of Agriculture and Food Chemistry,
Critical Reviews in Food Science and Nutrition, 32, 59–66. 47, 2327–2330.

You might also like