Vous êtes sur la page 1sur 29

LITERATURE REVIEW on WATER ACTIVITY and GLASS TRANSITION

Dr. Ted Labuza University of Minnesota Dept. of Food Science and Nutrition 1354 Eckles Ave. St Paul MN 55108 612-624-9701 tplabuza@epx.cis.umn.edu

water activity and glass transition

page # 2

In the mid 1970s, water activity came to the forefront as a major factor in understanding the control of the deterioration of reduced moisture and dry foods, drugs and biological systems.1,2 It was found that the general modes of deterioration, namely physical and physicochemical modifications, microbiological growth, and both aqueous and lipid phase chemical reactions, were all influenced by the thermodynamic availability of water (aw, relative vapor pressure(p/po), or water activity) as well as the total moisture content of the system. It is the difference in the chemical potential of water () between two systems that results in moisture exchange and above a certain chemical potential as related to the aw of a system there is enough water present to result in physical and chemical reactions. Water activity or the equilibrium relative humidity of a system is defined as:

aw = p/po= %ERH

(1)

where p = vapor pressure of water in equilibrium with the dry system po = saturation vapor pressure of pure water at the same temperature. It must be noted that the use of the water activity term signifies that the system is at equilibrium with its surrounding vapor which may not be true in all cases. The water activity is related to chemical potential by:

= o + RT ln aw

(2)

Note that the chemical potential is the true driving force for the transfer of energy or matter and in this case at constant temperature is exponentially related to the water activity. The physical structure of a food or biological product, important from both functional and sensory standpoints, is often altered by changes in water activity due to moisture gain or loss. For example, the caking of powders is attributed to the amorphous-crystalline state transfer of sugars and oligosaccharides that occurs as water activity increases above the glass transition point which is close to 0.3 to 0.4 at room temperature.3 This caking interferes with the powder's ability to dissolve or be free flowing and phase transitions can lead to volatile loss or oxidation of encapsulated lipids. The desirable crispiness of crackers, dry snack products such as potato chips, and breakfast cereals is lost if a moisture gain results in a water activity above 0.40-0.45.4, again above the glass transition. Conversely, raisins and other dried fruits may harden due to the loss of water associated with decreasing water activity, usually below 0.55. Thus, raisins or other fruits in breakfast cereals are sugar coated to reduce the moisture loss rate or are modified with glycerol to reduce the water activity thereby preventing moisture loss. These procedures inhibit the net moisture transfer rate from the raisins to the cereal, therefore maintaining the cereal's crisp nature and the softness of the fruit pieces in the presence of a chemical potential driving force. Finally, as aw increases, the permeability of packaging films to oxygen and water vapor increases, due to swelling in the rubbery state. Like physicochemical phenomena, the growth and death of microorganisms are also influenced by water activity. It has been repeatedly shown that each microorganism has a critical water activity below which growth cannot occur.5 For example, Aspergillus parasiticus does not grow below a water activity of 0.82 while the production of aflatoxin, a potent toxin, from the same organism is inhibited below a water activity of 0.87.6 For growth or toxin production to cease, key enzymatic reactions in the microbial cell must cease. Thus, the lowering of water

water activity and glass transition

page # 3

activity inhibits these biochemical reactions which in turn restricts microbial functioning as a whole. With spores, the lower the water activity, the more resistant they are to heat kill.. Microbially stable dry foods generally are defined as those with a water activity of less than 0.6, a level below which no known microbe can grow.5 Water activity has been shown to influence the kinetics of many chemical reactions such as the loss of vitamin C in the dry state.7 Except for lipid oxidation reactions where the rate increases as water activity decreases at very low water activities, the rates of chemical reactions generally increase with increasing water activity as depicted in Figure 1.8
50
relative rate moisture

40

20

10

relative reaction rate

30

0 0.0

0.2

0.4

0.6

0.8

1.0

Water Activity

Figure 1 The influence of water activity on chemical reaction rates in the aqueous phase overlaid on a sorption isotherm. Generally, the minimum reaction rate for aqueous phase reactions is found at the monolayer moisture content which is the point where theoretically all polar groups have adsorbed one molecule of water vapor.8 It is at or just above this point that an aqueous phase just begins to form such that chemical species can dissolve, diffuse and react. From moisture sorption isotherm data, the monolayer can be determined from either the BET equation, using data below a water activity of 0.5 or the GAB equation which is based on the thermodynamics of multilayer moisture adsorption and uses all the data. The GAB equation can be written as m = mokbCaW/[(1 - kbaW)(1 - kbaW + kbCaW)] (3)

water activity

where m is the moisture content, aW is the water activity, mo is the monolayer, kb is a multilayer factor, and C is a heat constant. Either a polynomial solution or non-linear regression is used to solve for the constants in this equation so as to determine the monolayer value. This is currently the most accepted equation to generate an isotherm. It can generate the whole curve from five data points at different water activities It is thus important from a shelf life testing standpoint to evaluate the moisture sorption isotherm of each system and then to calculate the monolayer water activity to determine the optimum point for stability. from either the BET equation and the GAB equation. From this, a window for initial moisture content can then be set to be used as a production goal so as to maximize shelf life. The rationale for this determination will be discussed and then the methods to determine the monolayer will be presented. There are many examples of the influence of water activity above the monolayer value on the rates of chemical reactions in the dry state which lead to loss of shelf life or biological activity.

Moisture g water/100 g solids

water activity and glass transition

page # 4

For example, the rate of ascorbic acid degradation increases dramatically as water activity increases from the dry state.7 Similarly, the rate of glucose utilization during non-enzymatic browning increases as water activity increases.9 Below the monolayer, the rate of lipid oxidation increases again, thus for systems which also contain unsaturated lipids, such as in the membranes of all biological systems, the shelf life is at a maximum near the monolayer. 8 Since many of the reactions in the aqueous phase are acid-base catalyzed, the amount of water and the availability of H+ and OH- ions in the reduced moisture state will also significantly influence reaction rates. One of the most studied reactions in our laboratory which can be used as an excellent example of reaction kinetics and stability is that of the acid-base catalyzed degradation of aspartame in both liquid and dry state systems.10 The influence of the system water activity state on aspartame degradation can be used to illustrate many shelf life related chemical phenomena in reduced moisture systems. Figure 2 shows the simple case where water activity is the only factor which is varied to determine the effect on aspartame degradation in the dry state. An increase in water activity from the dry state, increases the aspartame degradation rate.
100
Aw = 0.34

% Aspartame remaining

Aw = 0. 57

Aw = 0.66

10 0

Days

100

200

Figure 2. Degradation of aspartame in model systems as a function of water activity at pH 5 and 30 C These results can be depicted as a water activity versus half-life plot as shown in Figure 3, which also shows the complex influence of the initial pH of the system. From the slope of the line in Figure 3, the QA at each pH can be determined. The QA is defined as the ratio between the half-life at one water activity and the half-life at a water activity 0.1 units higher. The QA's for aspartame degradation range from 1.3 to 2, which means that for an increase in water activity of 0.1 the rate of aspartame degradation increases by 30-100%. For most aqueous phase chemical reactions in the dry state for aw= 0 to 0.7 , the QA is between 2 to 3, which emphasizes the need for maintaining the water activity as close as possible to the monolayer.

water activity and glass transition

page # 5

1000

pH 3 pH 5 pH 7 pH 3

100

10

1 0.4 0.5 Aw 0.6 0.7 0.8

Figure 3. Log half life vs water activity plot for aspartame degradation at aw 0.33 and 30 C. As was seen in Figure 1, above about aw 0.6 to 7, reaction rates fall again . In general the pattern of a linear rise and fall of reaction rate with water activity on a semilog plot is found as shown in Figure 4 .

Half Life in Days

100

ln k

10

Q a= 3

.1

0.0

rate constant

0.2

0.4

0.6

0.8

1.0

water activity

Figure 4. Influence of water activity on shelf life- QA Plot Figure 5 shows the effect of the initial pH before drying on the stability of aspartame in a system during storage at a water activity of 0.34. It is clear that even at low water activities, the initial pH before drying is another important factor that influences acid-base catalyzed degradation rates in the dry state. This phenomenon should be important in controlling the stability of many dry food, drug and biological systems.

water activity and glass transition

page # 6

Figure 5 . Degradation of aspartame as a function of pH at an aw of 0.34 and 30 C Figure 6 shows the pH rate profile plot for aspartame degradation at a water activity of 0.34. The slope of the inclines are much less than found in solution (slopes ~ 1 above the isoelectric point and -1 below in solutions) indicating that the actual pH in the aqueous condensed phase may be different than that in the wet state or the buffer concentration is changed due to the reduction in total water content.
.1
-1

.01

.001 2 3 4 5 6 7 8

Rate constant [day]

pH

Figure 6. pH rate profile plot at an aw of 0.34 Upon drying and rehydrating the model system, the buffer forms a super-saturated solution which changes the true pH in the condensed aqueous phase. Studies in our lab indicate that a wet system (i.e. before drying and rehydration) containing phosphate buffer at pH 5 will end up at a pH of 3.5 after drying, if the same amount of dry buffer salts are present in the amount of water found after drying and then rehydrating the model system.11 Thus a super-saturated buffer solution is formed. Therefore, the actual pH of reduced-moisture systems is lower than the initial pH used for preparation of the model system. Given the shape of the pH-rate profile this can either increase or decrease acid-base catalyzed reactions. It has also been demonstrated that the rates of acid-base catalyzed reactions are a function of total buffer concentration.10 Figure 7 shows this effect also holds true in a reduced-moisture system for aspartame stability at a water activity 0.34. If the buffer concentration increases because of the drop in water content, the rate of aspartame degradation will increase. However, as the buffer concentration approaches saturation, the pH will drop, which could then increase or decrease the rate of degradation depending upon which side of the isoelectric minimum the initial pH lies. Thus, the understanding of reaction rates of reactions which are pH and buffer dependent is much more complex at reduced moisture contents.

water activity and glass transition

page # 7

Figure 7. Aspartame loss at 30 C and an aw of 0.34 in phosphate buffer at two concentrations. The important effect of temperature on reaction rates has long been recognized.12 Generally, reaction rates increase with increasing temperatures. The most prevalent and widely used model to explain this effect is the Arrhenius relation, in which the rate constant is exponentially related to the reciprocal of the absolute temperature. This relationship can be derived from thermodynamic laws as well as from statistical mechanics principles and takes the form: EA k = kA exp (-RT ) (4)

In this equation, k is the reaction rate constant, kA is equal to the Arrhenius preexponential constant and EA is the excess energy barrier that the reacting substance needs to overcome to proceed to degradation products, generally referred to as the activation energy . In practical terms this relationship shows that if values of the rate constant k ,determined at different temperatures, are plotted on semilog graph paper against the reciprocal absolute temperature, 1/T, a straight line is obtained with a slope of -EA/2.303 R (R=1.987 cal/ mol, the universal gas constant). An example is shown in Figure 8 for aspartame degradation in solution. The value of the activation energy is the slope of this line times 2.303 R.

water activity and glass transition

page # 8

Figure 8. Typical Arrhenius plot for aspartame degradation To insure reliable results, the reaction rate should be determined at three or more temperatures and then k is plotted vs. 1/T on a semilog graph and a linear regression fit is employed to determine EA. Statistical analysis, is used to determine the 95% confidence limits of the Arrhenius parameters. If only three k values are available (ie. only three test temperatures were used), the confidence range is usually wide since the Student ta/2 value is large. To obtain meaningfully narrower confidence limits in EA and k , rates at more temperatures are required or methods to apply non-linear regression of all the data simultaneously can be used. Between 5 or 6 experimental temperatures would be the optimum but is not practical especially if several water activities are also to be tested. An alternative way of expressing the temperature dependence, which has been extensively used by the food industry and in the food science and biochemistry literature, is the shelf life or Q10 approach. Q10 is defined as the ratio of the shelf lives at two temperatures differing by 10 C. In essence, Q10 is defined as the relative change of shelf life qs, i.e., the time to reach an unacceptable quality level, when the product is stored at a temperature higher by 10 C. The Q10 approach introduces a temperature-dependence equation of the form

k(T) = ko e bT or qs = qo e-bT log k(T) = log ko + bT or log qs =l og qo - bT

(5) (6)

water activity and glass transition

page # 9

which shows that if k or the shelf life, qs is plotted on semilog graph paper vs. temperature (instead of 1/T of the Arrhenius equation) a straight line is obtained. Figure 9 shows such a semilog plot of qs vs. temperature in which the time for 25% loss of thiamin in pasta was chosen as the end of shelf life.
1000

Half life vitamin C [weeks]

Aw 0.4 Q 2.2
10

100
Aw 0.5 Q 10 1.93

Aw 0.6 Q 10 1.3

10 20 30 40 50

Temperature C

Figure 9. Shelf life plot for thiamin loss as a function of temperature and water activity. Such plots are often called shelf life plots, where -b is the slope of the shelf life plot and qo is the intercept. The shelf life plots are true straight lines only for narrow temperature ranges of 20 to 30 C. For such a narrow interval, data from an Arrhenius plot will give a relatively straight line in a shelf life plot. In other words Q10 and b are functions of temperature and depend on the temperature range.. The activation energy of a food quality loss reaction, Q10 and b (slope of shelf life plot) are interrelated through the following equation: EA 10 ln Q10 = 10 b = R T(T+10) (7)

The variation of Q10 with temperature for reactions of different activation energies is shown in Table 1. Table 1. Q10 Dependence on EA and Temperature EA kcal/mol 10 20 30 Q10 Q10 at 5 C at 20 C 1.87 1.76 3.51 3.10 6.58 5.47 Q10 at 40 C 1.64 2.70 4.45

water activity and glass transition

page # 10

There are many factors relevant to quality loss reactions that can cause significant deviations from an Arrhenius behavior with temperature.13 The most important is a temperature caused change in the reaction conditions such as changes in pH and water activity that usually are assumed to be the same at all temperatures. Our lab has shown the magnitude of this upward shift in the water activity for products in sealed packages going to higher temperatures.14 This shift would result in both a water activity and temperature acceleration of the chemical reactions leading to increased loss of shelf life. It is thus important in the understanding of most biological systems to determine the moisture sorption isotherm at least at two temperatures and then use the Clausius-Clapeyron equation to predict changes for temperature shifts. Phase changes may also be involved which can affect reaction rates. The influence of glass transition state will be covered in a latter section. Fats may change into the liquid state contributing to the mobilization of organic reactants. In frozen systems, the effect of phase change of the water to ice can cause a pronounced rate increase in the immediate sub-freezing temperature range because of increased reactant and H+ concentration and increased mobility of the proton through ice. The rate increase is especially notable for reactants of low initial concentration and is related basically to the freeze-concentration effect. It is prominent in the temperature zone of maximum ice formation, the width of which will depend on the type of food, but generally will be in the range of -1 C to about -10C. Other phase change phenomena are also important. Carbohydrates in the amorphous state may collapse or crystallize, creating more free water for other reactions. Denaturation of proteins as temperature increases can increase or decrease the susceptibility to chemical reactions depending upon the stereochemical factors that affect these reactions, another factor that can cause non-Arrhenius behavior. In general the rates of degradation reactions increase as temperature increases as was shown in Figure 9 but the temperature sensitivity, ie Q10 or b from the shelf life plot decreases. Table 2 lists the activation energies for aspartame degradation for a dry model system made to different pH values initially and held at different water activities. As seen, the activation energy decreases as water activity increases , at both pHs. Table 2. Activation Energies for Aspartame Degradation. pH aw Activation Energy initial Kcal/mole 3 3 3 5 5 5 0.33 0.55 0.65 0.33 0.55 0.65 25.4 22.1 22.3 27.8 24.9 22.3

The decrease in activation energy (a smaller negative value) as water activity increases may be based in part on the enthalpy/entropy compensation theory.15 For compound A, undergoing reaction to an activated intermediate state AX* before reacting to produce product B, the change in Gibbs free energy DG* for the activated state is : A + X --> AX* ---> B + X (8) where:

water activity and glass transition

page # 11

DG* = DH* - T DS* (9) * In this case DH is the system enthalpy change (a negative value for exothermic spontaneous reactions), T is absolute temperature, and DS* is the system entropy. The enthalpy change is directly proportional to the activation energy by: Ea = DH* + RT (10)

Thus the entropy should increase with increasing water activity as found since as water becomes less bound it is more available for reaction. Thus, as entropy increases to a more positive value, the activation energy should become a smaller negative value for the Gibbs free energy to remain unchanged. However, in some food systems the opposite occurs; the reason for this is unknown. Physical Properties of the rubbery and glassy state and stability Glass transition theory from the study of polymer science may help to understand textural properties of food systems and explain changes which occur during food processing and storage such as stickiness, caking, softening and hardening as was noted earlier. Figure 10 shows a glass rubber transition diagram, the line representing the glass temperature, Tg.

Figure 10. Representative glass-rubber transition diagram for an amorphous material From polymer science, if an amorphous material exists in the glassy state, it is hard and brittle, eg for cereals it would represent a crisp product. In the rubbery state the material is soft and elastic, for a fried snack or cereal this would represent an undesirable soggy state. Texture is an important sensory attribute for many cereal based foods and the loss of desired texture leads to a loss in product quality and a reduction in shelf life (Nielsen, 1979). The texture of crisp foods has been studied as a function of water activity by Katz and Labuza (1981). They determined that saltine crackers, popcorn, puffed corn curls and potato chips lost crispness if the water activity exceeded 0.35 - 0.50. They attributed crispness to intermolecular bonding of starch forming small crystalline-like regions when little water was present. These regions require force to break apart which gives the food a crisp texture. Above a certain water activity, the water was presumed to disrupt these bonds allowing the starch molecules to slip past each other when chewed. Hsieh, Hu, Huff and Peng (1990) also observed that puffed rice cakes lost crispness at a water activity just above 0.44. Vickers and Bourne (1976) showed that crisp perception of dry cereal snacks was the result of sounds generated when chewed which diminished as the water activity was

water activity and glass transition

page # 12

increased. Our lab has shown that loss of crispness is well explained by the transition from the glassy to the rubbery state. Caking is another property that can be related to the glass diagram as noted above. When a sugar is in solution and is dried, it is in the amorphous glassy state and the powder is free flowing.. At a high enough moisture or temperature, is crossing over the Tg line in Figure 10, the material can enter the rubbery state. In the rubbery state, dried amorphous sugars tend to crystallize rapidly because of increased diffusion rates above Tg, , a condition resulting in undesirable caking which inhibits free flow. Caking follows the steps shown below in Figure 11 for particles that are wetted by water vapor.
Wetting

Sticking

Drying

Wetting

Wetting Drying Hard Caking

Figure 11. Steps involved in caking and agglomeration Thus glass transition theory provides a clearer approach to understanding the physical and texture changes of crisp cereals or snacks as water content increases. The elastic modulus which is related to flow, extendibility and bending is approximately 103 times higher in the glassy state than the modulus in the rubbery state (Sperling, 1986). Dry cereal products when baked extruded or fried generally are in the glassy state and are best described as hard, stiff, brittle and crisp. As these foods gain moisture or increase in temperature, they may enter the rubbery state and become soft. It is desirable to retain crisp foods such as fried snack products and breakfast cereals in the glassy state and to retain high sugar foods in the glassy state to prevent caking. The choice of ingredients and level of plasticizers such as water and other small molecular weight components influences the glass transition temperature of a food product. In general, as the molecular weight of a polymer increases within a homologous series, the glass transition temperature increases (Sperling, 1986). The addition of plasticizers decreases the glass transition temperature (Roos and Karel, 1991). Sugars also cake at different rates with glucose and fructose being most susceptible thus it is hard to predict behavior without having established a glass transition diagram.

water activity and glass transition

page # 13

Study of texture by glass transition theory The key to understanding texture using glass transition theory lies in the ability to measure the glass transition temperature (Tg) for the food and to correlate this transition with a change in texture of the food as determined by sensory analysis or other means. The measurement of glass transition temperature of multicomponent foods, however, is quite difficult. Methods include differential scanning calorimetry (DSC) and dynamic mechanical thermal analysis (DMTA). Food systems under analysis may yield multiple glass transitions, and the melting or crystallization of components such as fats or sugars within the system may obscure the observation of the glass transition. In the case of multiple glass transitions, it is important to identify the specific transition which correlates with the change in food texture. Glass transition temperature measurement There are several methods which are used to measure the glass transition temperature for a material. The most common method reported in the literature for glass transition determination in food is differential scanning calorimetry (DSC). DSC measures the change in heat capacity between the glassy and rubbery states and is indicated by a change in baseline in a DSC thermogram as shown below in Figure 12. DMTA on the other hand measures changes in elastic behavior at the transition. It is much more difficult to perform but has a much higher sensitivity.

Tg

endothermic

T melt Temperature
Figure 12. Typical DSC thermogram Point of Tg determination In the literature, researchers report glass transition temperature data as either the midpoint or the onset Tg. As shown below in Figure 13 , the point chosen for the determination of Tg can affect the value of the glass transition temperature. The onset Tg is generally considered the most appropriate temperature to report. However, many researchers report midpoint Tg values since a plot of the first derivative of the glass transition curve shows a peak at the midpoint glass transition temperature making this point easy to identify.

water activity and glass transition

page # 14

T g onset T g midpoint

Temperature

Figure 13. Determination of Tg from DSC curve The graph below (Figure 14) shows results using the DSC at 10C/min and determining the midpoint temperature for native and pre-gelatinized starch.
140 120 100 80 60 40 20 0.05 0.15 0.25 Native Wheat Starch Pre-Gelatinized Wheat Starch

Wheat Starch

Temperature (C)

Moisture (g water/ g solid)

Figure 14. Zeleznak, K. J. and Hoseney, R.C. 1987. Glass transition in starch. Cereal Chem. 64(2):121-124.

Application of glass transition theory to the texture of cereal foods If textural changes in a cereal system can be correlated with a glass transition, and the state diagram for the cereal food is known, then the processing and environmental conditions can be controlled such that the desired state for the food is achieved and is also retained during distribution and storage. Little research has been performed to show the correlation between glass transition and the textural changes in foods. LeMeste, Huang, Panama, Anderson and Lentz (1992), for example, studied the texture of white pan bread by thermal mechanical analysis. The glass transition associated with cookies has been studied (Doescher, Hoseney and Milliken, 1987), but was not related to the texture of cookies. Much research can still be done in this area.

water activity and glass transition

page # 15

Several examples can be used to relate the glass transition state diagram to the texture of cereal foods. The observed transition from a crisp to a soggy texture as moisture is gained is obviously a glass transition. A benefit of knowing the state diagram for crisp snack foods includes the possibility of formulating a snack food with a higher glass transition temperature curve. A higher glass transition curve allows a food to reach a higher temperature or higher moisture content without losing its crisp texture. This would provide a longer shelf life for the crisp snack food, improve the snack quality, necessitate less stringent control over storage and distribution conditions, and reduce package film moisture barrier requirements. As a breakfast cereal picks up moisture from milk, it loses its crisp texture and becomes soggy. This is known as a loss of bowl life and may be the result of a glass to rubber transition. In this case, if it were possible to raise the glass transition curve for the cereal, more moisture absorption would be required before the cereal would become soggy and unacceptable. Highsugar cereals consumed by children tend to have short bowl lives. This may be the result of a lower glass transition temperature for such cereals as the result of plasticization by the higher level of small molecular weight sugars which may lower the Tg curve (Roos and Karel, 1991). For some cookie products, a soft texture makes the cookie seem freshly baked. A soft texture may be achieved through a number of means, including formulation such that the final baked cookie is in the rubbery state at room temperature. An interesting application of this is the dual- textured cookie based on a patent by Hong and Brabbs (1984). In this case, the outer cookie layer containing sucrose ends up in the glassy state when baked giving a crisp outer crust. The inner cookie dough is based on fructose and has a lower glass transition curve than the outer dough. Thus, the inner cookie ends up in the rubbery state following baking and seems soft and fresh. In the rubbery state, however, there is the possibility that crystallization of the sugars in the cookie could lead to a harder or crisp texture over time. Much work has been done to utilize sugar crystallization for the formation of a crisp texture for cookie products. This is discussed in patents by Martin and Furia (1990) and Gage and Mishkin (1990). Roos and Karel (1991) studied the rate of crystallization for several sugar solutions as a function of temperature above the glass transition temperature. They found that crystallization rate increased as T-Tg increased. Thus, if the state diagram for a sugar within a cookie product were known, the moisture and temperature conditions required to promote crystallization of the sugar to produce a crisp cookie would be also be known. A soft texture can be retained by adding ingredients to delay or prevent crystallization. With respect to caking, Figure 15 shows the DSC measured Tg curve vs the measurement of caking for a protein - sugar mixture made by spray drying (infant formula) as compared to the ampoule method (sealed sample heated slowly in a sealed glass ampoule at 1C per 3 minutes and shaken every few minutes to see if it sticks to the wall) and the propeller method (same heating rate but sample stirred with a propeller ). Generally the DSC Tg curve is below the other two as seen but does represent the moisture temperature line at which a sample will cake during long term storage and thus can be used as a guideline.

water activity and glass transition

page # 16

200

150

100

50

0 0 5 10 15 20

moisture (g H O/ 2 100 g solids)

Figure 15. Caking phenomenon for sugar protein system


m represents the Tg curve done by DSC at 1C/3 minutes s represents the ampoule method l represents the propeller method

Glass transition approach and reaction rates Glass transition theory applies to amorphous polymers and has also been found to be applicable to low molecular weight sugars (Slade et al., 1989). The glass transition temperature, Tg, is the temperature at which a glass to rubber transition takes place. Water, the most common plasticizer in foods, acts to decrease the glass transition temperature. The system properties above and below the glass transition temperature differ quite dramatically and such differences have been studied relative to the crystallization (Roos and Karel, 1991b) and the viscosity (Soesanto and Williams, 1981) of sugar solutions. The relationship between the rate of chemical reactions which occur in food systems and the state of the system, either glassy or rubbery, has also been studied. The properties of glassy and rubbery systems may contribute to differences in chemical reaction rates in each of these states. There is a dramatic change in the local movement of polymer chains at the glass transition temperature, resulting in a number of property differences between glassy and rubbery systems. As a system moves from the glassy to the rubbery state, viscosity drops dramatically from approximately 1012 to 103 Pasec at the glass transition temperature (Sperling, 1992). The reduced viscosity allows for greater polymer chain and reactant mobility. Free volume, defined as the amount of space associated with a system which is not taken up by polymer chains themselves, also changes between the glassy and rubbery states. The free volume available within a glassy system has been estimated to be between 2% and 11.3% of the total volume (Ferry, 1980), and is believed to increase substantially at the glass transition temperature due to a dramatic increase in the thermal expansion coefficient (Ferry, 1980). This increase in free volume should allow for faster diffusion of reactants. Based on the free volume required for diffusion, the size of a diffusing molecule may also be an important factor affecting diffusion rates. Diffusion is a function of the probability of creating a hole within a matrix which is sufficiently

water activity and glass transition

page # 17

large for a molecule to occupy. When a molecule is large compared to the available free volume, the probability of creating such a hole is low. Thus, a greater degree of free volume redistribution is required in order for diffusion of large molecules to take place compared to smaller molecules (Mauritz et al., 1990). This makes diffusion of large molecules within a region of limited free volume very slow. Reactions which are dependent upon the diffusion of such molecules may also be very slow. In addition to translational mobility, short range mobility may also be important for chemical reactions. Using electron spin resonance, Roozen and coworkers (1990, 1991) found a significant increase in the rotational mobility of spin probes within sucrose-water, glycerol-water and maltodextrin-water mixtures at a temperature which corresponded to the glass transition temperature, as measured by differential scanning calorimetry. The mobility of protons, as measured by nuclear magnetic resonance, has also been found to be higher in the rubbery state compared to the glassy state (Kalichevsky et al., 1992). Based on the properties associated with the glassy state, it might be expected that chemical reaction rates would be quite slow within the glassy state, or would not occur at all, but would substantially increase in rate within the rubbery state. In fact, this may be the explanation for cessation of reactions at the monolayer determined from moisture sorption isotherms. It is possible that the observed monolayer is not actually a monolayer, but rather a moisture content at which the glass transition is observed at a particular temperature. Lim and Reid (1991) studied the rate of diffusion-controlled processes within frozen systems. Maltodextrins, carboxymethylcellulose (CMC) and sucrose were used as model systems to study the rate of protein aggregation, ascorbic acid degradation and enzymatic hydrolysis at temperatures above and below the Tg' value for each model system. Figure 16 shows the results of enzymatic hydrolysis above and below the Tg' of a maltodextrin DE10 system.
5 4

Relative absorbance

Temperature (C)

3 2 1 0 0 50 100 150 200 250


Time (hours)

-3.5 -5.5 -8.5 -13 -19

Figure 16 Enzymatic hydrolysis of maltodextrin in the frozen state. Lim and Reid (1991) Note that at temperatures below the system Tg' (-10C), the reaction was very slow, but the rate increased above the Tg', with a large increase in hydrolysis rate occurring 4.5C above the reported Tg'. Similar results were found for other DE maltodextrins. Protein aggregation and ascorbic acid degradation within the maltodextrin systems were also slow within the rubbery state,

water activity and glass transition

page # 18

but increased in rate above the system Tg'. It was concluded that the maltodextrin system provided good cryostability below Tg'. For the CMC systems, Lim and Reid (1991) found that ascorbic acid degradation was observed below the system Tg', possibly due to the porous structure of the CMC matrix which allowed oxygen diffusion even within the glassy state. It may also be possible that the limited free volume within the glassy state was sufficient for oxygen diffusion to occur. CMC also failed to stabilize the system against protein aggregation within the glassy state. This was unexpected considering the high Tg' value for CMC (-11C) and that in order for aggregation to take place, the diffusion of large protein molecules through the matrix would be required. Lim and Reid (1991) found that the Tg' for a sucrose model system was very low (-33C) and that all study temperatures were above Tg', within the rubbery state. Despite this, it was observed that sucrose effectively stabilized the system against protein aggregation. The glass transition approach failed to predict the stability of the sucrose system in this case. As expected, however, the sucrose system showed rapid rates of ascorbic acid degradation within the rubbery state of this system. Note, however, that as the temperature above the Tg' increased, the water content in the system probably increased due to the melting of ice. Thus, it is difficult to attribute these observed results, as well as the results from other studies involving frozen systems, solely to a temperature effect. Karmas et al. (1992) found that nonenzymatic browning within the glassy state of carbohydrate model systems was very slow below the system Tg. Depending on the composition and moisture content of the carbohydrate system, the rate of nonenzymatic browning increased substantially at about 20-75C above the Tg. Figure 17 shows typical results of their work.
10 8
a w =0.12

d(OD)/dt

6 4 2 0 0 25 50 75

a w =0.33

100

T-T g

Figure 17 Nonenzymatic browning in the glassy and rubbery state Karmas et al (1992) For systems at aw=0.12 and aw=0.75, dramatic changes in the rate of nonenzymatic browning occurred at 50C and 75C above the Tg, respectively. They stated that the temperature at which a large increase in browning rate occurred could be related to the temperature for which the diffusion coefficient within the system rapidly increased based on changes in free volume. This

water activity and glass transition

page # 19

would be expected to occur at the glass transition temperature, but could occur at temperatures above the Tg depending on the system free volume and the size of the diffusing molecules. A study of the oxidation of orange oil within a maltodextrin M100 encapsulating matrix showed that oxidation occurred rather rapidly within the maltodextrin glassy state (Ma et al., 1992; Nelson and Labuza, 1992). As the water content of the matrix increased, but was still within the glassy state, reaction rates increased, as shown in Figure 18.
20
water activity 0.56

15
Limonene oxide (mg/g orange oil)
0.44

10
0.33

0.11 0 0.75

0 0 5 10 15
Time (days)

20

25

30

Figure 18 Oxidation of limonene as a function of water activity Nelson and Labuza (1992) In contrast to the expected rate increase within the rubbery state, the system at 75% relative humidity was within the rubbery state, but showed the greatest stability to oxidation. These findings were attributed to collapse of the matrix within the rubbery state which prevented oxygen diffusion through the matrix. Collapse occurs when a matrix can no longer support itself against gravity and collapses upon itself forming a compact system. Caking and sticking are examples of collapse (Downton et al., 1982). A rubbery system is characterized as a metastable state. In addition to collapse, crystallization may occur which may subsequently influence chemical reaction rates. Crystallization can occur within the rubbery state where viscosity is low such that molecules have sufficient mobility for crystallization to take place. Roos and Karel (1991b) found that the time for crystallization decreased dramatically as temperature above the glass transition temperature (T-Tg) increased for sucrose and lactose solutions. Crystallization leads to a cross-linking effect which results in reduced polymer chain flexibility and mobility. Michaels et al. (1963) reported lower gas diffusion rates within glassy and crystalline synthetic polymer matrices compared to rubbery matrices. Crystalline regions were found to impede diffusion since molecules were required to pass through regions of low polymer chain mobility. Labrousse et al. (1992) found that the collapse and crystallization of a lactose-gelatin matrix containing methyl linoleate oil led to an increased rate of oil oxidation. They attributed this to the reduction in free volume upon crystallization which forced the oil to the surface where protection from oxidation was minimal. Numerous researchers have examined the relationship between the collapse and the retention of encapsulated oils and volatiles (Omatete and King, 1978; Flink and Karel, 1972). The extent and rate of collapse or crystallization has been found to play a part in the retention and the subsequent stability of these materials. Based on the properties of glassy and rubber systems, it would be expected that reaction rates would be very slow within a glassy state and increase within a rubbery system. However,

water activity and glass transition

page # 20

based on the studies reported in the food science literature, it should not be assumed that glassy systems are completely stable, or that rubbery systems are less stable than glassy systems. Molecular size, matrix porosity and changes within the metastable rubbery state such as collapse and crystallization, may contribute to other, more complicated results. Water activity and the glass transition approach both have strengths and weaknesses in defining the relationship between moisture content and chemical reaction rates. It is important to recognize, however, that it is a relatively simple task to measure the water activity of food systems. A number of instrumental methods are available for such determinations (Stamp et al., 1984). With the use of controlled relative humidity chambers, specific water activities can also be achieved for a particular food system. The measurement of glass transition temperature, on the other hand, requires more effort. Differential scanning calorimetry has been used extensively to measure the Tg of simple food systems, but this method lacks the sensitivity required to measure the Tg of complex food systems. Other methods have, therefore, been examined as a means to measure glass transition temperatures (Kalichevsky et al., 1992; Bruni and Leopold, 1991). Without the ability to accurately and rapidly determine the Tg for a food system as a function of moisture content, it is difficult to correlate glass transition temperatures and reaction rates within complex food systems. Much work can still be done in this area. Arrhenius vs WLF temperature model In addition to the effect of moisture on reaction rates, it is also useful to understand the temperature dependence of chemical reactions in order to predict product shelf life as we have noted. Often, shelf life studies are performed at elevated temperatures in order to expedite data collection. In order to predict the rates of degradative chemical reactions at other temperatures, a relationship between the reaction rate and temperature must be established. The Arrhenius relationship has traditionally been used to describe the temperature dependence of chemical reactions (Glasstone, 1946). However, when reactions are diffusion-limited, an alternative approach by Williams, Landel and Ferry (1955) may be appropriate. The Arrhenius relationship is the major mathematical model used to describe the temperature dependence of most chemical reactions (Glasstone, 1946). Arrhenius originally developed this empirical relationship for glucose hydrolysis (1889). Subsequently, the basis for the relationship has been derived from thermodynamic and quantum mechanical principles and has been found applicable to many chemical and physical processes such as diffusion. As noted earlier, the Arrhenius relationship shown in eq. 4 is:

k =k

exp( E

/ RT )

(4)

where k is the rate constant at temperature T, ko is a pre-exponential factor, R is the ideal gas constant and Ea is the activation energy. The reference temperature, according to this equation, is absolute zero. Of course, one cannot measure reactions at this temperature. Kinetic data at several temperatures over a fairly large temperature range are required in order to test the applicability of the Arrhenius model. A plot of ln(k) vs. 1/T, if a straight line, indicates the applicability of the Arrhenius model and that the activation energy over the specific temperature range is constant. If the plot deviates from a straight line, either the test methods were poor, or other reactions begin to become critical above some temperature and influence the reaction rate being studied.

water activity and glass transition

page # 21

It has been stated that the Arrhenius model is applicable for describing the temperature dependence of reactions within the glassy state of a food matrix and also at 100C above the glass transition temperature, but is not applicable within the rubbery state (Slade et al. 1989). This assumption can be tested by applying the Arrhenius relationship to rates of chemical reactions within glassy and rubbery systems to determine whether the Arrhenius model provides a good fit of the data. von Meerwall and Ferguson (1979b) studied the rate of oil diffusion in natural rubber using NMR spectroscopy. They found that an Arrhenius plot of the diffusion coefficient of oil within the rubbery state of the system showed the expected curvature above the glass transition temperature. A broad temperature range from -10C to 140C was used in their study. Based on several kinetic models, the Arrhenius plot would be expected to be curved if such a large temperature range were used (Labuza, 1980) due to the extra temperature dependence at high temperatures which is not accounted for in the Arrhenius model. Ollett and Parker (1990) reported that the viscosity of anhydrous glucose and fructose measured at temperatures above the system Tg did not show Arrhenius-type temperature dependence. Although the correlation coefficients for the Arrhenius plots for glucose and fructose were 0.98 and 0.99, respectively, slight curvature in the Arrhenius plots for both systems was observed, possibly due to the large temperature range for this data (approximately 50C). From data of Lim and Reid (1991), the Arrhenius plot for an enzyme hydrolysis reaction within a partially frozen maltodextrin DE 25 above the system Tg' was constructed and is shown in Figure 19.
0

-1

y = 79.497 - 2.1621E+4x R2 = 0.999

ln(k)

-2

-3

-4 0.0037

0.0038

0.0039

1/T (K-1 ) Figure 19 Arrhenius plot of Lim and Reid results (1991) The data spans a very limited temperature range of about 10C above the Tg' (13.3C). The curvature expected within the rubbery state was not observed, and the Arrhenius plot is quite linear (r2=0.99) for this reaction, with a calculated activation energy of 42.9 kcal/mole (10.3 kJ/mole). In order to observe the curvature expected in the Arrhenius plot within the rubbery state, data over a broader temperature range above the system Tg may be required. However, this is impossible for frozen systems since the system melts at slightly higher temperatures, and is essentially a liquid above 0C.

water activity and glass transition

page # 22

A shift in the slope of an Arrhenius plot indicates a change in the activation energy for a reaction. Due to the property differences between glassy and rubbery systems, a break in the Arrhenius plot at the glass transition temperature of a system would be expected. Figure 20 shows an Arrhenius plot which spans a temperature range which included a glass to rubber transition from work by Karmas et al. (1992) on nonenzymatic browning within a carbohydrate matrix.
2 0 -2
rubbery Ea = 45.1 kcal/mol

ln(k)

-4 -6

Tg glassy Ea = 20.8 kcal/mol

-8 -10
0.0027 0.0028 0.0029 0.0030 0.0031 0.0032 0.0033

1/T (K-1) Figure 20 Arrhenius plot of data of Karmas et al (1992)

They referred to the temperature at which a break in the Arrhenius plot was observed as the system-dependent critical temperature. They attributed this break in the Arrhenius plot to collapse of the carbohydrate matrix above the system glass transition temperature. In the case of the lactose/CMC/trehalose/xylose/lysine model system in Figure 6, the break in the Arrhenius plot is observed near the glass transition temperature (Tg=50C). As generally expected, the activation energy calculated for nonenzymatic browning within the rubbery system was higher than the activation energy within the glassy state. Note that despite the 40C temperature range within the rubbery state, the Arrhenius plot is quite linear within this region (r2= 0.99). If all the data, both below and above the Tg, are included in the regression, the activation energy is 42.3 kcal/mole (10.1 kJ/mole) and the r2=0.985, which is still very good. In some cases, knowing the glass transition temperature may be useful in understanding and explaining breaks observed in Arrhenius plots. However, a break in an Arrhenius plot may be due other changes in a system, and is not always the result of a glass to rubber transition. Because the activation energy for reactions in glassy and rubbery systems may be different, caution should be applied when extrapolating rates of reactions within glassy systems to systems within the rubbery state. Although it has been stated that the Arrhenius model is not applicable for describing the temperature dependence of reaction rates within systems in the rubbery state, it has been seen that over small temperature ranges, the Arrhenius model may adequately predict changes in reaction rate. However, the WLF model has been stated to be a more appropriate approach for modeling the temperature dependence within rubbery systems. An approach to model the temperature dependence of mechanical and dielectric relaxations within the rubbery state, where it is assumed that the Arrhenius model does not

water activity and glass transition

page # 23

theoretically apply, was suggested by Williams, Landel and Ferry (1955). This approach is known as the WLF approach and has the following form:

log aT =

C1 (T Tref ) C2 + T Tref

(11)

where C1 and C2 are system-dependent coefficients (Ferry, 1970) and aT is defined as the ratio of the relaxation phenomenon at T to the relaxation at the reference temperature, Tref (i.e., h/href for viscosity). It has been suggested in the literature that WLF kinetics may also describe the temperature dependence of chemical reaction rates within amorphous matrices above their glass transition temperature (Slade et al., 1989). For systems where diffusion is free volume dependent, it was shown by Sapru and Labuza (1993) that the rate of reaction can be expressed using the WLF equation with a shift factor of aT= (kref/k):

log

kref k

C1 (T T ref ) C2 + (T Tref )

(12)

Average values for the WLF coefficients were calculated by Williams et al. (1955) using the available values for many synthetic polymers. It is quite common in the literature to use the average coefficients, which have the values of C1=17.44 and C2=51.6, for establishing the applicability of the WLF model (Roos and Karel, 1991b; Soesanto and Williams, 1981). Peleg (1992) discussed several problems associated with the use of average coefficients in the WLF equation. He found disagreement between the use of the average coefficients and literature values for actual coefficients for prediction purposes 20-30C above the glass transition temperature. Typical proof of the applicability of the WLF model, as cited in the food science literature, is made based on fitting the WLF equation to kinetic data shown as in a plot of log(k) vs. T-Tg. The first step in such a proof involves determining the rate constant at the glass transition temperature, since this is generally not known. An average value of kg is calculated by solving equation 12 for all experimental data and using the measured Tg and the average WLF coefficients. The calculated average kg value is then used along with the average WLF coefficients to calculate values of the rate constant over the experimental temperature range using equation 12. A reasonable fit of the line to the data is stated as proof of the applicability of the WLF model. This was done for establishing the applicability of the WLF approach for the crystallization of sugars (Roos and Karel, 1991b) and for the viscosity of sugar solutions (Soesanto and Williams, 1981). The disadvantage of this approach is that it assumes that the average WLF coefficients are appropriate for a particular data set, although this may not be the case. Roos and Karel (1990) measured the time for lactose crystallization at several moisture and temperature conditions. They reported that the average WLF coefficients adequately described the temperature dependence of their lactose crystallization data. However, an adequate fit of the crystallization data in the temperature range of the measurements can also be made by utilizing other WLF coefficients. Several different values for the WLF coefficients were used to determine an average value for the time for crystallization at the Tg using lactose data for systems initially equilibrated to water activities between 0.11 and 0.44, as was done by Roos and Karel (1990) using the average WLF coefficients. The average value of the time for crystallization at Tg (g)

water activity and glass transition

page # 24

was then be used to predict the time to crystallization at a specific temperature. This procedure was followed for several different sets of WLF coefficients to construct the lines shown in Figure 21.
12
C1 C2 51.6 60.0 45.0 30.0

aw 0.11 0.22 0.33 0.44

10

17.44 20.0 15.0 10.0

log ( )

10

20

30

40

50

60

T-Tg (C)

Figure 21 WLF plot for lactose crystallization from data of Roos and Karel (1990) As observed in Figure 21, all predicted lines visually fit the crystallization data for the systems equilibrated to water activities between 0.11 and 0.44 about equally well. However, the time for crystallization at the glass transition temperature, qg, was very different, depending on the set of WLF coefficients used. This can be observed as the difference in the intersection of the predicted curves with the y axis at T-Tg=0. This is the danger of utilizing the average WLF coefficients, as well as the danger of making predictions at the glass transition temperature using data far above Tg (Peleg, 1992). Alternative approaches will be suggested for accessing the applicability of the WLF model. According to the WLF equation and using the average coefficients, a 10C increase above the Tg should correspond to a reaction rate increase of approximately 600 times (Slade et al., 1989). Table 3 shows the rate increase expected for other temperature changes above Tg. Karmas et al. (1992) found that the increase in nonenzymatic browning rate within a carbohydrate model system above the Tg was not as large as that predicted using the average WLF coefficients, as shown in Table 1. A simple check of this involves calculating kT values at temperatures above the Tg using equation 14 and the average coefficients, and comparing these to measured values. It is evident that the rate increases predicted from use of the average WLF coefficients did not describe the temperature dependence of nonenzymatic browning within this system. This may not mean that the WLF model is not applicable for this reaction, but rather that other WLF coefficients, rather that the average values, must be used for this system. Table 3

water activity and glass transition

page # 25

Rate increases predicted using the WLF equation with average values for C1 and C2, and observed by Karmas et al. (1992) for nonenzymatic browning within a lactose/CMC/trehalose/xylose/lysine model system at initial aw=0.12. Temperature range Tg Tg + 10C Tg + 20C Tg + 30C Tg + 10C Tg + 20C Tg + 30C Tg + 40C Relative rate increase predicted 678 110 3.5 1.58 Relative rate increase observed 21.8 7.39 4.06 2.82

One approach to determine the applicability of the WLF equation for describing the temperature dependence of a chemical reaction would first be to determine the rate of reaction at the Tg, if possible. Once kg is known, equation 12, with Tg as the reference temperature, can be rearranged to the following form: 1 kg C2 log = C1 ( T Tg ) C1 k such that a plot of
1 1

(13)

1 kg log vs. T Tg k

(14)

gives a straight line with a slope equal to -C2/C1 and an intercept of -1/C1, if the WLF model is applicable. For many systems, however, it is not possible to utilize the Tg as the reference temperature because the phenomena at the Tg is so slow that good kinetic rate constants cannot be measured. In such situations, a reference temperature above the Tg should be used to evaluate the applicability of the WLF model. In fact, this approach was originally suggested by Williams et al. (1955) for the same reason. If the glass transition temperature is known, the WLF constants at the reference temperature can be transformed to correspond to the glass transition temperature, as shown by Peleg (1992): C1 =
'

C1C2 C2

C'2 = C2

(15)

In these relationships, d is the temperature difference between Tref and Tg, and C1' and C2' are the transformed WLF coefficients at the glass transition temperature. These values can then be compared to the widely used average WLF coefficients. Figure 22 was constructed using the approach of equation 14 and nonenzymatic browning rates obtained by Karmas et al. within a carbohydrate model system over a 50C temperature range above the glass transition temperature.

water activity and glass transition

page # 26

0
y = - 5.557E-2 - 10.015x r2 = 0.994

1/log(kref /k)

-1

C1 =

18.0

C2 = 180.2

-2

-3 0.0

0.1

0.2

0.3

1/(T-Tref)
Figure 22 Proper WLF linear plot for the browning data of Karmas et al (1992) Note that the line is quite linear (r2=0.99), indicating the applicability of the WLF model. From the linear regression for the plot, the values of C1 and C2 were determined to be 18.0 and 180.2, respectively, using 55C as the reference temperature. When transformed to correspond to the glass transition temperature, 50C, using equation 6, the values for C1' and C2' were 18.5 and 175.2, respectively. These are quite different from the average WLF coefficients often cited in the food science literature. An alternative method to determine the applicability of the WLF model for describing the temperature dependence of a chemical reaction is possible when kinetic data at many temperatures are available, the Tg is known, but kg cannot be measured. In this approach, it is first assumed that the WLF relationship describes the temperature dependence of the reaction. Then, initial estimates of kg, C1 and C2 are made and iterative nonlinear regression is used to optimize these values, according to the relationship of equation 12. If kinetic data are available at many temperatures over a wide range (at least 30C), it is probable that only one line fits the data set and that the values determined for kg, C1 and C2 are probably quite close to actual values. von Meerwall and Ferguson (1979a) used this three parameter optimization method to fit the diffusion of a diluent within a rubbery matrix with the WLF equation. They found that experimental precision was very important for the accuracy of such a fit. An additional disadvantage of this method is that it assumes, rather than proves, that the WLF model applies to the system. Understanding the physical properties of foods using glass transition theory may still be in its infancy; however, the potential for utilizing this theory for the improvement of food texture is great. It is important to note, however, that food ingredients cannot be chosen based solely on their affect on the glass transition temperature of a particular food. As an example, it may be desirable to formulate soft foods to the rubbery state. However, the rate of degradative chemical reactions and crystallization may be higher in the rubbery state compared to the glassy state which may adversely affect food stability and quality during distribution, especially caking. In reviewing these concepts several principles are clear:

water activity and glass transition

page # 27

1. water activity is an important concept for stability 2. moisture sorption isotherms help to understand stability 3. the monolayer optimal stability moisture can be obtained from the isotherm 4. the glass transition curve is a critical factor needed to understand physical changes 5. a glass transition in the middle of the temperature range for a shelf life test could lead to significant error. In addition, if the sorption parameters for an equation for the isotherm is known, eg the BET, GAB, or a simple linear model of m = ba +c over a small water activity range, then moisture gain or loss for any given distribution system is possible.16 This greatly enhances the food product development scientist in predicting product shelf life. General Water Activity Literature References
1. R. B. Duckworth, ed., "Water Relations in Foods," Academic Press, London (1975). 2. Labuza, T.P. Sorption phenomena in foods. 1975 in "Theory, determination and control of physical properties of foods". C. Rha ed. D. Reidel Pub. Dordrecth Holland pp 197-219. 3. M. Saltmarch and T. P. Labuza, Influence of relative humidity on the physicochemical state of lactose in spray- dried sweet whey powders, J. Food Sci. 45(5):1231-1236 & 1242 (1980). 4. E. E. Katz and T. P. Labuza, The effect of water activity on the sensory crispness and mechanical deformation of snack food products, J. Food Sci. 46:403-409, 81 5. L. Beuchat, Microbial stability as affected by water activity, Cereal Foods World. 26:345 (1981). 6. M. D. Northolt, C. A. H. Verhulsdonk, P. S. S. Soentoro, and W. E.Paulsch, Effect of water activity and temperature on aflatoxin production by Aspergillus parasiticus, J. Milk Food Technol. 39:170-174. 7. S. Lee and T. P. Labuza, Destruction of ascorbic acid as a function of water activity, J. Food Sci. 40:370-373 (1975). 8. T. P. Labuza, The effect of water activity on reaction kinetics of food deterioration, Food Technol. 34(1):36-41 & 59 (1980). 9. J. A. Kamman and T. P. Labuza, A comparison of the effect of oil vs. shortening on the rates of glucose utilization in non - enzymatic browning, J. Food Proc. and Preserv. 9:217-222 (1985). 10. Bell, L.N. and Labuza, T.P. Aspartame Degradation in limited water conditions in "Water Relationships in Foods" H. Levine and L. Slade eds. Plenum Press N.Y. pp 337-349 (1991) 11. Bell, L.N. and Labuza, T.P. Potential pH implications in the dry state. Cryo Letters 13:335-344 12. Labuza,T.P. Applications of chemical kinetics to deterioration of foods. J. Chem. Ed. 61:348-358 (1985) 13. Labuza, T.P. and Schmidl, M.K. Accelerated shelf life testing of foods. Food Technol 39(9):57-62 14. Labuza, T.P., Kaanane, A. and Chen. J. Effect of temperature on moisture sorption isotherms and aw shift of two foods. J. Food Sci. 50:385-391(1985) 15 . T. P. Labuza, Enthalpy entropy compensation in food reactions, Food Technol. 34(2):67-77 (1980). 16. Taoukis, P.S, A. ElMeskine and Labuza, T.P. Moisture transfer and shelf life of packaged foods. in "Food Packaging Interactions" J. Hotchkiss ed. ACS Symposium Series ACS Press pp 244-261

Glass Transition References


Bailley, R.T., North, A.M. and Pethrick, B. 1981. Molecular Motion in High Polymers. Clarendon Press. Oxford. Bruni, F. and Leopold, A.C. 1991. Glass transitions in soybean seed. Plant Physiol. 96:660-663. Doescher, L.C., Hoseney, R.C. and Milliken, G.A. (1987). A mechanism for cookie dough setting. Cereal Chem. 64(3),158-163 Downton, G.E., Flores-Luna, J.L. and King. C.J. 1982. Mechanism of stickiness in hygroscopic, amorphous powders. Ind. Eng. Chem. Fundam. 21:447-451. Ferry, J.D. 1980. Viscoelastic Properties of Polymers. 3rd edition. John Wiley & Sons, Inc.: New York. 264-320. Flink, J. and Karel, M. 1972. Mechanisms of retention of organic volatiles in freeze-dried systems. J. Fd. Technol. 7:199-211. Gage, D.R. and Mishkin, M.A. (1990). Texture equilibration in cookies. U.S. Patent 4,892,745 Glasstone, S. 1946. Textbook of Physical Chemistry. 2nd edition. Van Nostrad: Princeton, NJ. Hsieh, F., Hu, L., Huff, H.E., and Peng, I.C. (1990). Effects of water activity on textural characteristics of puffed rice cake. Lebensmittel Wissenschaft und Technologie. 23(6),471-473. Hong, C.A., and Brabbs, W.J. (1984). Doughs and cookies providing storage-stable texture variability.

water activity and glass transition

page # 28

U.S. Patent 4,455,333 Kalichevsky, M.T., Jaroszkiewicz, E.M., Ablett, S., Blanshard, J.M.V., and Lilliford, P.J. 1992. The glass transition of amylopectin measured by DSC, DMTA and NMR. Carbohydrate Polymers. 18:77-88. Karmas, R., Buera, M.P. and Karel, M. 1992. Effect of glass transition on rates of nonenzymatic browning in food systems. J. Agric. Food Chem. 40(5):873-879. Katz, E.E. and Labuza, T.P. (1981). Effect of water activity on the sensory crispness and mechanical deformation of snack food products. J. Food Science. 46(2), 403-409 Labrousse, S., Roos, Y. and Karel, M. 1992. Collapse and crystallization in amorphous matrices with encapsulated compounds. Science des Aliments. In press. Labuza, T.P., Maloney, J.F., Karel, M. 1966. Autoxidation of methyl linoleate in freeze-dried model systems. II. Effect of water on cobalt-catalyzed oxidation. J. Food Sci. 31:885-891. Labuza, T.P. 1971. Kinetics of lipid oxidation in foods. CRC Critical Reviews in Food Technology. 10:355-405. Labuza, T.P. 1980. Enthalpy/entropy compensation in food reactions. Food Tech. 2:67-77. LaJolo, F.M. Tannenbaum, S.R. and Labuza, T.P. 1971. Reactions at limited water concentration. II. Chlorophyll degradation. J. Food Sci. 36:850-853. LeMeste, M., Huang, V.T., Panama, J., Anderson, G. and Lentz, R. (1992). Glass transition of bread. Cereal Foods World. 37(3), 264-267 Martin, A.J. and Furia, T.E. (1990). Shelf stable cookie. U.S. Patent 4,965,077 Lim, M.H. and Reid, D.S. 1991. Studies of reaction kinetics in relation to the Tg' of polymers in frozen model systems. In Water Relationships in Foods. H. Levine and L. Slade, editors. Plenum Press: New York. 103-122. Ma, Y., Reineccius, G.A., Labuza, T.P. and Nelson, K.A. 1992. The stability of spray-dried microcapsules as a function of glass transition temperature. Presented at the IFT Annual Meeting, 1992. New Orleans, LA. Mauritz, K.A., Storey, R.F. and George, S.E. 1990. A general free volume based theory for the diffusion of large molecules in amorphous polymers above Tg. 1. Application to di-n-alkyl phthalates in PVC. Macromolecules. 23:441-450. Michaels, A.S., Vieth, W.R. and Barrie, J.A. 1963. Diffusion of gases in polyethylene terephthalate. J. Applied Physics. 34(1):13-20. Nelson, K.A. and Labuza, T.P. 1992. Relationship between water and lipid oxidation rates: Water activity and glass transition theory. In Lipid Oxidation in Foods. A.J. St. Angelo, ed. American Chemical Society: Washington, DC. 93-103. Nielsen, A.C. Company. 1979. Product and package performance: The consumers view. A.C. Nielsen Co IL Omatete, O.O. and King, C.J. 1978. Volatile retention during rehumidification of freeze dried food models. J. Fd. Technol. 13:265-280. Peleg, M. 1992. On the use of the WLF model in polymers and foods. CRC Critical Reviews in Food Sci. and Nutr. 32:5966. Roos, Y. and Karel, M. 1991a. Phase transitions of mixtures of amorphous polysaccharides and sugars. Biotechnol. Prog. 7:49-53. Roos, Y. and Karel, M. 1991b. Plasticizing effect of water on thermal behavior and crystallization of amorphous food models. J. Food Sci. 56(1):38-43. Roos, Y. and Karel, M. (1991). Apply state diagrams to food processing and development. Food Technology. 45(12), 66-71 Roozen, M.J.G.W. and Hemminga, M.A. 1990. Molecular motion in sucrose-water mixtures in the liquid and glassy state as studied by spin probe ESR. J. Phys. Chem. 94:7326-7329. Roozen, M.J.G.W., Hemminga, MA. and Walstra, P. 1991. Molecular motion in glassy water-malto-oligosaccharide (maltodextrin) mixtures as studied by conventional and saturation-transfer spin-probe e.s.r. spectroscopy. Carbohydrate Research. 215:229-237. Sapru, V. and Labuza, T.P. 1993. Glassy state in bacterial spores predicted by polymer glass transition theory. J. Food Sci. 58(2):445-448. Slade, L. and Levine, H. 1988. Non-equilibrium behavior of small carbohydrate-water systems. Pure and Appl. Chem. 60(12):1841-1864. Slade, L., Levine, H., and Finey, J. 1989. Protein-water interactions: Water as a plasticizer of gluten and other protein polymers. In Protein quality and the effects of processing. Phillips, R.D., Finlay, J.W., Editors. Marcel Dekker: New York. 9-123. Soesanto, T. and Williams, M.C. 1981. Volumetric interpretation of viscosity for concentrated and dilute sugar solutions. J. Phys. Chem. 85:3338-3341. Sperling, L.H. (1986). Introduction to Physical Polymer Science, pp. 224-295. New York, John Wiley & Sons Stamp, J.A., Linscott, S. Lomauro, C. and Labuza, T.P. 1984. Measurement of water activity of salt solutions and foods by several electronic methods as compared to direct vapor pressure measurement. J. Food Sci. 49:1139-1142. von Meerwall, E. and Ferguson, R.D. 1979a. Diffusion of hydrocarbons in rubber, measured by the pulsed gradient NMR method. J. Applied Polymer Sci. 23:3657-3669. von Meerwall, E. and Ferguson, R.D. 1979b. Pulsed-field gradient NMR measurements of diffusion of oil in rubber. J. Applied Polymer Sci. 23:877-885. Warmbier, H.C., Schnickles, R.A. and Labuza, T.P. 1976. Effect of glycerol on non-enzymatic browning in a solid intermediate moisture model food system. J. Food Sci. 41:528-531.

water activity and glass transition

page # 29

Williams, M.L., Landel, R.F. and Ferry, J.D. 1955. The temperature dependence of relaxation mechanisms in amorphous polymers and other glass-forming liquids. J. Chem. Eng. 77:3701-3707. Vickers, Z.M. and Bourne, M.C. (1976). A psychoacoustical theory of crispness. J. Food Science. 41, 1158-1164

Vous aimerez peut-être aussi