Mlll“H11NHW1NHIWIHIWIHIHHWWIMHI 114 370 _THS_ I ll WI Iflllfllfllfll fl" Ill III III II III ”I I2! I This is to certify that the thesis entitled DEVELOPMENT OF A SIMULATION MODEL FOR SHELF-LIFE PREDICTION OF PACKAGED MODEL LIPID FOOD SYSTEMS presented by Kiyonori Kogashiwa has been accepted towards fulfillment of the requirements for M. S . degree in Packaging Dr. Jack R. GiraCIn‘ Major professor Date November 25, I980 0-7639 (15’- \\\\ L I III' v 1 I\ :IWIIC .- V.“",l’” I 1;] II '- ’ FL 3' Mir?) gm 50m University OVERDUE FINES: 25¢ per day per item RETURNING LIBRARY MATERIALS: Place in book return to remOuc charge from circulation record: DEVELOPMENT OF A SIMULATION MODEL FOR SHELF-LIFE PREDICTION OF PACKAGED MODEL LIPID FOOD SYSTEMS BY Kiyonori Kogashiwa A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE School of Packaging 1980 ABSTRACT DEVELOPMENT OF A SIMULATION MODEL FOR SHELF-LIFE PREDICTION OF PACKAGED MODEL LIPID FOOD SYSTEMS BY Kiyonori Kogashiwa A simulation model was developed which considered the influence of both package permeability to oxygen and the product oxidation rate on product quality. Relative humidity and temperature were held constant and the package permeability and product oxidation rates were determined as a function of oxygen partial pressure within the package. A computer-aided simulation model was developed, based on these kinetic data, to predict the extent of oxygen uptake by the food product, under selected package storage conditions. The oxygen uptake levels by the model food system, obtained from experiment and by calculation were compared and a fair agreement was obtained. Knowledge of the accept- able level of oxygen uptake of the product and the package permeability can be used for an estimate of product shelf- life. ACKNOWLEDGMENTS The author expresses his sincere appreciation and gratitude to Dr. Jack R. Giacin, chairman of his thesis committee and major professor, for his valuable assistance and guidance throughout this study. The author expresses his appreciation to the member of the thesis committee, Dr. Steven W. Gyeszly, for his unselfish contribution of time and his valuable help. He is extremely grateful to his thesis committee member, Dr. J. Ian Gray and also to his graduate assistants, for their valuable advice in the field of food science. ii TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES INTRODUCTION LITERATURE REVIEW . THEORETICAL BACKGROUND Mechanism in Lipid Oxidation Oxidation Rate Permeation Rate SIMULATION MODELING Assumptions Simulation Model MATERIALS AND METHODS Preparation of Model Lipid Food System . Oxidation Rate Studies Actual Storage Studies RESULTS Oxidation Rate Studies Results Prediction Results . Actual Storage Studies Results DISCUSSION Consideration of Possible Error in the Oxidation Rate Studies Discussion of Simulation Model SUMMARY AND CONCLUSIONS iii Page vi 10 11 16 16 17 21 21 22 26 28 28 32 35 38 38 42 46 TABLE OF CONTENTS (continued) APPENDICES A. TABLES FOR COMPOSITION AND DISTRIBUTION OF MODEL LIPID FOOD SYSTEM . B. HEADSPACE VOLUME DETERMINATION C. OXIDATION RATE STUDY DATA PROCESSING D. ACCEPTABLE QUALITY LEVEL OF MODEL LIPID FOOD SYSTEM . . . . . . . . . LIST OF REFERENCES iv Page 48 48 49 51 SS 56 Table 10. 11. 12. 13. 14. 15. LIST OF TABLES Oxygen Partial Pressure (Atm.) Change within the FLEX-CAN as a Function of Time Oxidation Rate as a Function of Oxygen Partial Pressure Predicted Change in the Oxygen Partial Pressure (Atm.) within the Respective Packages Predicted Oxygen Consumption within the Respective Packages Predicted Shelf— Life of Packaged Model Lipid Food System . . . . . Package Conditions for Actual Storage Test Change in Oxygen Partial Pressure (Atm.) within the Respective Packages (Actual Storage Test) Effect of Error Factors on the Oxidation Rate Effect of the Time-Step Function on the Computed Oxygen Partial Pressure Change Effect of Permeation of Nitrogen on the Oxygen Partial Pressure Change within the Package Composition of Model Lipid Food System Lipid Distribution in Model Lipid Food System . Change in the Concentration of the Carbon Dioxide within the FLEX- CAN by Sorption as a Function of Time . . . . Comparison of the Determined Headspace Volume to the Actual Volume Oxidation Rate Study Data Processing for FLEX-CAN Sample 1 Page 29 3O 33 33 34 35 36 42 43 45 48 48 50 50 54 LIST OF FIGURES Figure 1. Factors for the Package System Design 2. Boundary Conditions for Gas Permeation of Plastic Film . 3. Relationship Between Oxygen Consumption and Oxygen Permeation . . . 4. Illustration for Determination of Oxidation Rate vi Page 13 18 41 INTRODUCTION Packages affect the quality of food products, mainly by controlling moisture, oxygen and light transfer. In addition, packaging provides protection from biological attack as well as providing protection of the product in the mechanical environment. Package selection can therefore determine or influence the shelf-life of a food product. The most desirable package would maintain product quality for the required shelf-life period most economically. Traditionally, shelf-life has been determined by storage tests, carried out under either real or accelerated storage conditions. Storage tests under actual environmental con- ditions of time, temperature and relative humidity, are more accurate, but time-consuming and costly. The accel— erated storage tests would be less time-consuming and less costly. However, prediction errors may be introduced. In comparing these traditional techniques for shelf-life determination, shelf-life prediction by calcu- lation, based on a simulation model of environment-package- product interaction, has a number of advantages. For example, this technique is less costly and much less time- consuming. In some applications, the calculation method was reported to be even more accurate than the accelerated 2 test method (Clifford et al., 1977). It may also be an advantage that the calculation technique expresses shelf- life prediction mathematically, especially with the recent development of highly mathematical, statistical, decision- making systems. For example, the factors which must be taken into consideration for package design are shown in Figure 1 (Gyeszly, 1980). As shown, the total cost of the package system should be minimized by optimizing the package design. Package optimization can be obtained by mathematical pro- cessing, such as linear or non-linear programming. While storage stability studies carried out under real time- temperature conditions or under accelerated conditions might be applied in such model building, it would require that an extensive number of tests be carried out, due to the inflexibility of such studies to variable change. As described above, a major advantage of computer simulation is that it allows for a rapid assessment of the influence of environmental change in product or package conditions. This study will deal with the interaction of oxygen permeation through a package and oxygen consumption by a food product. Specifically, the factors: Time, External Environment, Permeability of the Package Material, Permeability of the Package, Permeation Rate Through the Package Wall, Internal Environment, Interaction Between the Product and Components of Headspace, and Quality of Product (c.f. Figure 1), will be considered. Several assumptions have been made in this study concerning shelf-life prediction, 3 as the critical acceptable quality level for the model food system was not determined. The oxidation rate plays an important role in developing a simulation model for lipid-containing food products, and thus, it is necessary to determine the rate of lipid oxidation under well defined storage environmental conditions. Rate data were therefore obtained in a system, where temperature and relative humidity were held constant and the rate of oxidation was monitored as a function of continuously changing oxygen partial pressure. A number of assumptions were also made for the simulation model. Computed results were then compared to the experimental data. Physical Permeability of Design of the Packaging the Package3 Material 10 I Permeability of the Package Permeation Rate Through the Package Wall / \ 2f 11 Environment Inside of Package l Environment Requirement I \ / \ Product Headspace I Environment Interaction 7 / f Quality of Product . L .f”‘_——\¢f" 12f Cost ‘ SHELF-LIFE Figure 1. Factors for the Package System Design. LITERATURE REVIEW Shelf-life prediction has been conducted by accel- erated storage tests or through calculation based on physical or chemical properties of the product and package. Clifford et al. (1977) compared the shelf-life prediction by accel- erated storage tests to that of the calculation method and evaluated the agreement with experimental data obtained from storage under ambient conditions. They indicated that the calculation method not only required much less time and re- sources, but was also more accurate. The calculation method has been used for about forty years; Felt et a1. (1945) used the linear relationship between the permeation rate of water vapor and the water vapor partial pressure difference to predict the shelf-life of a cereal product. In this study, shelf-life was determined solely by the moisture content of the product. Numerous studies have since been reported in the area of product storage stability and shelf-life pre- diction, following the publication of this work. In this regard, emphasis has been focused primarily on lipid oxidation of freeze-dried foodstuffs. Karel (1967) has reviewed the theoretical aspects of lipid oxidation, a principal deteriorative mechanism in the storage stability of dehydrated food products. Quast and Karel (1971) 6 determined the rate of oxygen uptake and the effective diffusivity of oxygen in some dehydrated food products. Maloney et a1. (1966) studied the autoxidation of methyl linoleate as a function of water activity. Labuza (1971) reported on the effects of both water activity and glycerol content on the oxidation of lipids. Zirlin and Karel (1969) described the oxidation of a freeze-dried model system con- sisting of methyl linoleate and gelatin. Cabral et a1. (1979) determined the performance of various packages for potato chips, a very oxygen and moisture sensitive product, in terms of selected chemical and physical properties, such as peroxide value, hexanal formation, and texture evaluation. Martinez and Labuza (1968) studied several deteriorative mechanisms affecting the quality of freeze-dried salmon, such as lipid oxidation and non—enzymatic browning. Tuomy and Walker (1970) dealt with the effects of storage time, moisture level, and headspace oxygen concentration on the quality of a dehydrated egg mix. The storage stability studies have also been ex- tended to include the development of simulation models for the prediction of product shelf-life. Karel (1967, 1974) and his co-workers (Karel, Mizrahi and Labuza, 1971; Quast and Karel, 1972a, l972d; Quast, Karel and Rand, 1972b; Labuza, Mizrahi and Karel, 1972) used computer—aided mathe- matical models for the evaluation of package requirements. These models were based on a combination of a kinetic equation of deterioration (Marcuse and Fredriksson, 1968; Labuza et al., 1969) and the permeability characteristics 7 of packages. The resultant differential equations were solved by the use of a computer. The computer-aided iter- ation technique enabled the prediction of deterioration of a product under defined package requirements and defined storage conditions. The procedure was applied to several ? deterioration mechanisms of a food product, such as lipid oxidation in dehydrated shrimp (Simon et al., 1971) and potato chips (Quast et al., 1972a, 1972b), and non-enzymatic 5 browning in dehydrated cabbage (Mizrahi, Labuza and Karel, 1970a, 1970b) and dehydrated tomatoes (Mizrahi and Karel, 1977). Singh (1974) applied a similar procedure to Vitamin C9 degradation in a liquid food product. Further, Quast, Karel I and Rand (1972b) introduced a new model for predicting oxi- dative deterioration to include the effect of equilibrium relative humidity. In this study, to better simulate actual storage conditions, the rate of oxidation was determined as a function of oxygen partial pressure, extent of oxidation, and equilibrium relative humidity. They reported that the experimental and predicted values agreed reasonably well. Henig et a1. (1973) applied the computer-aided simulation model for the permeation-respiration interaction in packaged bananas. The study predicted equilibrium oxygen and carbon dioxide concentrations in the package headspace. The calculated result was validated by experimental test. In addition, Henig and Gilbert (1975) have employed this technique to packaged tomatoes. In all the studies described above, the predictions were made under constant temperature conditions. However, 8 Kwolek and Bookwalter (1971) suggested that the quality change can be expressed as.a function of storage temper- ature. They applied their proposed technique to published data and reported exceptional agreement between the pre- dicted and experimental results. Labuza (1979) reviewed the mathematical models for fluctuating temperature se- quences. Further, he developed equations to calculate the quality change in a product undergoing either random, sine, or square wave time-temperature distribution for both zero and first order reactions. As reviewed, mathematical models have been employed for simulating environmental factors, as well as for product properties and package properties, in order to predict the quality index change of various products. THEORETICAL BACKGROUND Mechanism in Lipid Oxidation Oxidation of unsaturated fatty acids proceeds through a free-radical chain mechanism involving initiation, propagation and termination steps (Gray, 1978). These can be formulated as: Step 1. RH + 021—» R0 + -OOH (Initiation) Step 2. R- + O —+-ROO° ROO- + RH —.-ROOH + R- (Propagation) Step 3. R- + R--—- RR R- + ROO° —*-ROOR ROO~ + ROO~ —+-ROOR + 02 (Termination) where: RH unsaturated fatty acid, R- = alkyl radical, ROO- = peroxy radical, ROOH = hydroperoxide. As presented above, the degree of lipid deterioration, as a result of oxidation, can be determined by the concentration of hydroperoxide (i.e. peroxide value), the major reaction product formed during the initial stages of oxidation. 10 Further, there is a stoichiometric relationship between the peroxide value and oxygen uptake by an unsaturated fatty acid (lipid). If all of the oxygen, which was consumed during product oxidation, is incorporated into a hydro- peroxide, each molar equivalent of peroxide corresponds to half a mole of oxygen uptake by the product. Consequently, a peroxide value of one corresponds, stoichiometrically, to an oxygen uptake of 11.2 ul 0 (STP)/g dry product [or 2 0.5 u mole O2 (STP)/g dry product] (Quast and Karel, 1972a). Therefore, it is theoretically possible to determine the degree of oxidative deterioration by monitoring oxygen uptake by a lipid-containing food product. Assumption 1, in SIMULATION MODELING, was made based on this relationship. Oxidation Rate The rate of lipid oxidation has been shown by Marcuse and Fredriksson (1968) and Labuza et a1. (1969) to be dependent on the oxygen partial pressure, moisture content, and the extent of oxidation of product. The classic form for expressing the rate of lipid oxidation, as developed by Bollard (1949), is shown below, and considers the rate of oxidation as a function of only oxygen partial pressure. dVOz _ 1 dt ’ c + D/Po2 (1) where: VO2 = volume of oxygen uptake by the food, C, D = constants found by data curve fitting, PO2 oxygen partial pressure within the package. 11 More recently, Quast and Karel (1972b) developed an equation for expressing the rate of oxidation as a function of all three aforementioned variables. This relationship is given by Equation (2). dE _ K3 ‘I' KQE P02 a? ‘ RH1/2 + E K1 + K2P02 (2) where: E = extent of oxidation expressed as microliters of oxygen uptake per g of product [pl 02 (STP)/g product], RH = equilibrium relative humidity within the package, PO2 = oxygen partial pressure (atm) within the package, 1, 2, 3, K4 = constants found by data curve fitting. Permeation Rate The driving force for gases and vapors penetrating or diffusing through permeable packages is the concentration difference between the internal and external environment of the package. The rate of transfer of a diffusing substance can be expressed mathematically by Equations (3), (4), commonly referred to as Fick's first and second laws of diffusion. F = “D 57(- (3) fig. : D 32C (4) t 3X2 where: F = flux (the rate of transfer of diffusing substance per unit area), D = diffusion coefficient, 12 C = concentration of diffusing substance, t = time, X = space coordinate measured normal to the section. The boundary conditions for Fick's first and second laws of diffusion can be illustrated, as shown in Figure 2. In this analysis, it is assumed that at steady state, the con- centration of gas or vapor remains constant at all points of the sheet. Therefore, the diffusion Equation (4) reduces to: Q; N C.) = O (5) Q... >< N Assuming D is constant, Equation (6) is then obtained by integrating Equation (5) twice with respect to X, and intro- ducing the boundary conditions, X = 0 and X = 2. (6) ._L_. S1de l \\ _%% S1de 2 \ P1 \ p2 \\ C C1 2 X = 0 X = 2 where: C1, concentration of gas or vapor on the surfaces X = O, X = K, respectively (C1>C2), P1, partial pressure of gas or vapor at the surfaces X = 0, X = 2, respectively (P1>P2), film thickness. Figure 2. Boundary Conditions for Gas Permeation of Plastic Film. 13 14 Equations (3) and (6) can be combined to give Equation (7): dC C -c F = -D a? = D I, 2 (C1>C2) (7) Further, from the Henry's law relationship, Equation (8), the rate of transfer of a diffusing species can be expressed in terms of the permeant partial pressure at the surface X = O and X = 2, Equation (8a). C = S°P (8) D-S Plipz (P1>P2) (8a) ’11 II where: S = solubility. By definition, the permeability constant (P) = D-S. Therefore, F = P g (9) where: (P1>P2). The relationship between the permeation rate through the defined film and the permeability constant (P) can be ex- pressed by the following equation: "UI P = % = (A (Pl-P2) = Pp (Pl-P2) (10) where: Q = quantity of diffusing substance transferring through the film, t = time, A = area, D P = film permeation rate (= by definition), f 15 Pp = package permeability (= FLA). In this study, Equation (10) was used for the simulation modeling, with P-A/R being defined as the package perme- ability. A more detailed discussion of these and other diffusion coefficients is given by Crank and Park (1963) and Talwar (1974). SIMULATION MODELING Assumptions The following assumptions were made prior to the model building. Assumption 1: The amount of oxygen consumed by a model lipid food system is directly proportional to the degree or extent of lipid oxidation (Quast and Karel, 1972a). Validity of Assumption 1: As previously discussed, theoretically a linear relationship between oxygen uptake and the extent of lipid oxi- dation is assumed in the early stage of oxidation. Assumption 2: The rate of oxidation can be expressed mathe- matically as a function of oxygen partial pressure. Validity of Assumption 2: The classical form of the oxidation rate equation is shown in Equa- tion (1). In this study, the simplest form, R0 = III'POZ 'I' b, was employed which was based on the results of the oxidation rate equation study, where: R0 = oxidation rate, PO2 = oxygen partial pressure within the package, m, b = constants experimentally obtained. 16 17 Assumption 3: Oxygen permeation rate of the package can be expressed mathematically as follows: OP = Pp (Pl‘Pz) where: OP = oxygen permeation rate of the package, Pp = package permeability, P1, P2 = oxygen partial pressure outside and inside the package, respectively. Validity of Assumption 3: The equation is widely used, proven, and accepted in the literature. Assumption 4: The headspace volume change is only due to oxygen volume change. (The effect of nitro- gen permeation on the headspace volume is ignored.) Validity of Assumption 4: The validity of this assump- tion will be further discussed. Assumption 5: There is no interaction between packaging material and a product. Validity of Assumption 5: Since both selected films (Mylar and Saran) have good oil resistance properties (Agranoff, 1977), it can be assumed that the interaction is too low to influence the oxygen permeability of these films within the experimental conditions of this study. Simulation Model The relationship between oxygen consumption and oxygen permeation can be represented schematically as follows (Figure 3): A \ \ . \ Perme- f Permeation ation | Rate #’/// Through the: Package I Oxygen Partial Pressure Inside the Package (atm.) . . Oxidation in Ox1dat10n . . F d Rate _. a L1p1d oo 4/// System \ Package 4// External Environment Where: ——*- represents influence on, .__. represents the flow of oxygen. Figure 3. Relationship Between Oxygen Consumption and Oxygen Permeation. 18 19 As shown in Figure 3, the various factors interact with each other and their relationship is time-dependent. Therefore, the system can be simulated, assuming the vari- ables remain constant during a short period of time (this is referred to as a time-step). Based on this assumption, the system can be simulated discretely by an iteration technique. The resultant simulation model is shown below in sequential order. Step 1: Step 2: Step 3: Quantity of oxygen consumed by a food system during a time-step (At) can be calculated as follows: OCAt = [a-P02(t) + b] - At where: OCAt = oxygen consumption during At, PO2 = oxygen partial pressure within the package at time = t, a, b = experimentally obtained constants, At = time-step. Quantity of oxygen permeated into the package head- space through the package wall during a time-step (At) can be calculated as follows: OPAt = Pp - [0.208 — POz(t)] . At where: OPAt = quantity of oxygen permeated through the package wall during At, Pp = package permeability, POZ(t) = oxygen partial pressure within the package at time = t. Change in the absolute quantity of oxygen within the package during a time-step (At) is as follows: = _ :8 OAt OPAt OCAt 20 where: OAt = change in the absolute quantity of oxygen within the package during a time-step. *Notation (-) is dependent on the direction of the oxygen flow. Step 4: Quantity of oxygen within the package at time (t + At) can be expressed as follows: TO(t + At) = TO(t) + OAt where: TO(t + At) = quantity of oxygen within the package at time, t + At, TO(t) = quantity of oxygen within the package at time, t. Step 5: Headspace volume at t + At is expressed as follows: V(t + At) = V(t) + OAt where: V(t + At) = headspace volume within the package at time, t + At, V(t) = headspace volume within the package at time, t. Step 6: Oxygen partial pressure within the package at time, t + At is expressed as follows: = TO(t + At) P02(t + At) V(t + At) where: P02(t + At) = oxygen partial pressure within the package at time, t + At. A computer program was written based on the above simulation model. Calculations were performed following Steps 1 to 6 for each package, using experimentally obtained data. MATERIALS AND METHODS Preparation of Model Lipid Food System The major components used in preparation of the model food system were: soybean oil (Hain Food Co., Inc., Los Angeles, California) and carboxymethyl cellulose (CMC) (Hercules Incorporated, Wilmington, Delaware; Type 7HF). Tween 20 (polyoxyethylene sorbitan monooleate, Fisher Scientific Co., Fairlawn, New Jersey) was used as an emulsifying agent. Soybean oil, 30 g, was mixed with CMC, 3 g, Tween 20, 1.5 g, and phosphate buffer of pH 6.0, 600 g. The resultant slurry was mixed for 1 minute in a laboratory blender (Waring Products Division, Dynamics Corporation, New Hartford, Connecticut). Two hundred grams (200 g) of the mixed slurry was then transferred to an aluminum pan (O.D. = 20 cm), and the sample was dehydrated by freeze— drying. The sample remained in the freeze dryer (Virtis Model II, Repp Industries Inc., Gardiner, New York) for 3 days. Operating conditions for freeze-drying were as follows: glycol temperature, -120°F; platen temperature, 100°F; vacuum, 5 u. 21 22 Freeze-dried samples were stored at -20°C prior to usage. Blank samples, which served as positive controls, were prepared in a similar manner, except that no soybean oil was added to the control samples. For determination of the composition of the freeze- dried model food system, the percent of oil was determined by the official method of A.O.A.C. 24.005. The water content was determined by drying the sample to constant weight in a conventional oven (50°C), and the percentage of the CMC, Tween 20 and buffer salt was determined based on the slurry composition. The resultant composition is presented in Table 11 (Appendix A). To determine whether the model food system had a homogeneous distribution of lipid, a sample was cut into four (4) pieces and the respective portions were weighed. The oil in each of the portions was extracted by the standard proce- dure (A.O.A.C. 24.005) and weighed. As shown in Table 12 (Appendix A), the oil appears to be fairly uniformly dis- tributed throughout the freeze-dried sample. Oxidation Rate Studies Package Design and Structure. The oxidation rate studies were carried out using the following flexible package: Reynolds FLEX-CAN Retort Pouch (polyester/aluminum/— polypropylene laminate film). This package design and struc- ture was used for the rate studies, since for all intents and purposes the pouch can be assumed to be impermeable. Prior to usage, a gas sampling septum was provided by application 23 of a globe of silicon rubber (General Electric Company, Waterford, New York) to the pouch surface, and it was cured overnight at 25°C. Water Activity Control. The water activity of the freeze-dried model system was carefully controlled by equil- ibration over a saturated salt solution (potassium nitrite; RH of 48% at 25°C) (Greenspan, 1977). In order to prevent potential oxidation during the equilibration period, the following procedure was employed. A sample was charged into the FLEX—CAN Retort Pouch, and the pouch was placed in a vacuum desiccator which contained the saturated salt solu- tion. The air within the desiccator was evacuated by a vacuum pump and replaced by pure nitrogen. The operation was repeated three times. The samples in the open pouches were then allowed to equilibrate for 3 days under nitrogen, at a temperature of 25°C, and a RH of 48%. The equilibration maintained the model lipid food system at an equilibrium RH of 15% at 50°C, the temperature at which the oxidation studies were conducted. Temperature Control. Oxidation rate studies were carried out at 50°C. Preliminary studies showed that the temperature within the pouch was 50°C, when the pouch was placed in a constant temperature water bath, maintained at 50:1°C. Monitoring Oxygen Consumption Levels. The model freeze-dried food system (approximately 15 g), equilibrated to defined water activity, was weighed and added to the FLEX-CAN Retort Pouch (flat dimension of 4%" x 7"). 24 Six pouch samples containing the model lipid system (Samples, No. 1 to 6, in Table l) and a pouch containing the positive control (blank in Table l) were employed in this study. The pouches were then sealed and placed in a constant temperature bath, maintained at 50:1°C. At predetermined time intervals, the pouches were removed from the constant temperature bath, and an aliquot of headspace gas within the FLEX-CAN Pouch was removed with a gas tight syringe through the silicon rubber septum. The gas samples were injected directly into a gas chromatograph and the concentration of oxygen in the sample determined. A Hewlett Packard Gas Chromatograph, Model 5830A, equipped with dual thermal conductivity detection, was used for determination of oxygen concentration. Gas chromato- graphic conditions: 3 feet x 1/4 inch O.D. stainless steel column, packed with Molecular Sieve 5A (Supelco, Inc., Bellefonte, Pennsylvania); helium flow rate of 30 ml/min.; injection port temperature of 150°C; detector temperature of 350°C; column temperature of 70°C; oxygen retention time of 1.78 minutes. The percent of oxygen was computed from the oxygen peak area, and this value was used directly for the concentration of oxygen. The experiment was terminated when the oxygen concentration within the FLEX-CAN Retort Pouch was reduced to a level of 1%. Headspace Volume Determination. Following termi— nation of the rate studies, the headspace volume of each FLEX—CAN Retort Pouch was determined by a modification of the procedure of Davis and Hunington (1978). The initial carbon dioxide concentration within the FLEX-CAN Pouch was determined 25 by gas chromatography using a 1.0 m1 headspace gas sample. A known volume (1.0 m1) of pure carbon dioxide was then introduced into the headspace of the FLEX-CAN, and after an equilibration period of 20 minutes, the carbon dioxide con- centration was determined again by gas chromatography. A Hewlett Packard Gas Chromatograph, Model 5830A, equipped with dual thermal conductivity detection, was used for determi- nation of carbon dioxide concentration. Gas chromatographic conditions were as follows: 6 feet x 1/8 inch O.D. stainless steel column, packed with Chromosorb 102, 80—100 mesh (Johns-Manville, Celite Division, Denver, Colorado); helium flow of 10 ml/min.; injection port temperature of 150°C; detector temperature of 350°C; column temperature of 70°C; carbon dioxide retention time of 0.97 minute. The concen- tration was obtained based on an external standard of 100% carbon dioxide. The volume of the FLEX-CAN was calculated from the relationship: _ Va x 100 v - W (11) where: V = headspace volume (ml), Va = volume of carbon dioxide added (ml), C1 = final carbon dioxide concentration (volume/volume %), C0 = initial carbon dioxide concentration (volume/volume %). 26 The validity of this procedure was established during development of the method, and the results of those studies are summarized in Appendix B. Actual Storage Studies Films with high oxygen-barrier properties were selected for the actual storage studies. These were a vinyl- idine chloride/vinyl chloride copolymer (Saran, 1 mil), and a Saran-coated polyethylene terephthalate (Mylar M24, 0.5 mil), respectively. Oxygen permeability of the films was measured on the Oxtran-100 (Modern Control, Inc., Minneapolis, Minnesota). The measurements were done under the conditions similar to those of the storage studies: temperature of 50°C (the average of upper and lower chamber temperatures) and 0% relative humidity (RH). Film samples were cut and sealed on three sides according to a master template (12 cm x 8 cm) using a Sentinal Impulse Heat Sealer (Packaging Industries, Montclair, New Jersey). Conditions for sealing were as follows: impulse time of 0.3 second for Saran, 0.4 second for Mylar, respectively; pressure of 30 psi. for both films. A sampling septum of silicon rubber was provided in a manner similar to that employed for the FLEX-CAN Pouch. The water activity of the model lipid system was also controlled by equilibration over the saturated salt solution, as previously discussed. Samples were triplicated for both Saran and Mylar film packages. (Results are repeated for Samples Mylar M24-1, 2 and 3, and for Saran—2 in the RESULTS Section.) 27 The packaged samples were placed in a conventional oven maintained at 50i2°C. The samples were removed from the oven, and the oxygen concentration within the package was monitored at predetermined time intervals. After 611 hours of storage, the packaged samples were removed from the oven, allowed to cool to ambient conditions, and the headspace volume of each package was determined. The determination of oxygen concentration and headspace volume was carried out in the same manner as for the FLEX-CAN Pouches. RESULTS Experiments were carried out according to the procedure described in the previous section. Results are shown as follows: Oxidation Rate Studies Results Oxygen partial pressure change within the FLEX-CAN as a function of time was obtained experimentally and the results summarized in Table 1. Oxidation rates, as a function of oxygen partial pressure, were then derived from the experimentally obtained data in the way presented in Appendix C. The results are tabulated in Table 2. 28 Table 1. Oxygen Partial Pressure (Atm.)(a) Change within the FLEX-CAN as a Function of Time. Time Sample Number (Hours) Blank l 2 3 4 5 6 0 .207 .207 .206 .207 .206 .207 .207 44 .205 .179 .175 .179 .187 .189 .192 69 .205 .164 .155 .160 .178 .182 .187 93 .205 .147 .137 .142 .171 .174 .178 119 .204 .128 .117 .123 .164 .163 .168 142 .204 .110 .094 .104 .156 .154 .160 161 .204 .096 .079 .149 .146 .152 186 .204 .077 .052 .139 .136 .142 196 .204 .065 .042 .135 .130 .137 208 .204 .055 .031 .130 .124 .132 219 .203 .044 .023 .124 .118 .126 234 .204 .030 .014 .118 .109 .119 246 .203 .018 .015 .111 .106 .114 258 .204 .013 .013 .108 .103 .108 269 .203 .012 .012 .100 .093 .099 (a) Oxygen partial pressure was obtained based on oxygen concentration (volume/volume %). National Weather Service in East Lansing, Michigan reported that the atmospheric pressure during the experiment period was 0.97 atm. with standard deviation of 0.0089 atm. the atmospheric pressure can be assumed to be 1 atm. 29 Therefore, nmoo. mmo. onoo. omo. veoo. Hmo. mnoo. woo. Nmoo. moo. vwoo. mmo. vooo. Nmo. mooo. moo. Hnoo. ono. onoo. coo. omoo. woo. onoo. ooH. mwoo. mma. mnoo. Baa. owoo. wma. onoo. NvH. Hooo. an. wuoo. sea. mwoo. ooH. mooo. mmH. onoo. voa. owoo. mnH. wooo. qua. Nnoo. oma. nnoo. mom. oooo. oom. whoo. mom. mfiflo mo anew A.Epmo mfiflo mo Ewyww. A.Epmo mHHo mo SHAW] m.aumo Hog .H:\No HEV ohdwmohm Mom .H:\~o HEV thmmon pom .H:\No HEW whammogm ovum Hmfiugmm opwm Hwfluhmm comm Hwflpmmm coapmezmcou comxxo soameSmcoo somxxo cOHumESmcou somxxo comxxo :omxxo somxxo m onEmm N mamawm H oamewm Honesz ofimewm oHSmmoem Hwflphmm comxxo mo coflpocSm e we comm :oflpwoflxo .N magma .3 30 .u xflocomm< CM oousomopm mfl opsmmogm Hmfluymm somxxo mo :ofipocsm m we come coflpmoflxo ozp campno ow oasoooogm mcflmmoUOHm wumo och moo mwoo. on. onoo. moa. oooo. on. vnoo. mHH. omoo. ooH. ovoo. HHH. mooo. oHH. wmoo. ooH. nnoo. wHH. whoo. omH. mnoo. wHH. mmoo. «NH. nnoo. NmH. Hooo. vmfi. mnoo. mNH. vooo. an. oooo. oNH. omoo. vmfi. mwoo. med. mwoo. omH. mooo. oma. vooo. mmfi. mmoo. ovH. Hooo. ovH. mnoo. ooa. vooo. va. mmoo. oma. mooo. on. wmoo. moH. «moo. vofi. wooo. wnH. nmoo. «AH. mvoo. find. oooo. Boa. mmoo. NwH. omoo. wha. wmoo. HmH. nwoo. owa. wmoo. Boa. vooo. mom. mooo. mom. Nnoo. oom. AHHO mo ammw: m.Epmo mfiflo mo Edam m.Epmv mHHo mo EwHw‘ m.Epmo Mom .H:\No Heo ohdmmowo pom .H:\No Hey magmmoym Mom .H;\~o Hag ohdmmonm opmm Hmfiupmm oumm Hwfluhmm comm Hmfluemo coameSmcou cowxxo cofiumezmcou :owxxo :OAHQESmcou comxxo cowxxo comxxo cowxxo o onEmm m ofimamm v omMEmm Accesz onEmm meoscfloeooo N mfinme 31 32 Obtained data, shown in Table 2, were plotted and linear regression analysis was carried out since data suggested that the oxidation rate and the oxygen partial pressure had a linear relationship. The following relation- ship was obtained. OR = .00453 x P02 + .00705 (12) where: OR = oxidation rate (m1 OZ/hr. per gram of oil), PO2 = oxygen partial pressure (atm.). Prediction Results Change in the oxygen partial pressure and the oxygen consumption was predicted using a computer program with the obtained Equation (12). Predictions were performed for the respective packages, which were used for the storage studies; namely, Samples Mylar M24-l, Mylar M24—2, Mylar M24-3 and Saran-2. The predicted change in the oxygen partial pressure is presented in Table 3, and the predicted oxygen consumption is shown in Table 4. The shelf-life of the packaged model lipid system was then determined for the individual packages. In this study, the shelf-life is defined as a period of time re- quired, until oxygen uptake by the packaged, model lipid system reaches an unacceptable quality level. The results are shown in Table 5. For this study, the product was assumed to be unacceptable after sorption of 2.73 ml 02 (STP/g oil). 33 A detailed discussion concerning derivation of this value is presented in Appendix D. Table 3. Predicted Change in the Oxygen Partial Pressure (Atm.) within the Respective Packages. Time Oxygen Partial Pressure (Atm.) (Hours) Mylar M24-l Mylar M24-2 Mylar M24-3 Saran—2 0 0.208 0.208 0.208 0.208 60 0.178 0.175 0.180 0.187 120 0.149 0.143 0.153 0.165 180 0.121 0.113 0.128 0.144 240 0.097 0.084 0.104 0.122 300 0.073 0.057 0.081 0.099 360 0.052 0.033 0.061 0.076 420 0.034 0.011 0.042 0.054 480 0.017 - 0.025 0.032 Table 4. Predicted Oxygen Consumption within the Respective Packages. Time Oxygen Consumption (ml Oz/gram oil) (Hours) Mylar M24-l Mylar M24-2 Mylar M24-3 Saran—2 0 0 0 0 0 60 0.48 0.48 0.48 0.48 120 0.94 0.94 0.94 0.95 180 1.02 1.40 1.41 1.41 240 1.86 1.85 1.86 1.87 300 2.30 2.29 2.31 2.32 360 2.74 2.73 2.75 2.77 420 3.18 3.16 3.19 3.21 480 3.61 - 3.62 3.65 u I‘lflIIIIIIIQIICI‘qliIIIT HIIl‘l-WquI . I I1 IRRIII. 34 Table 5. Predicted Shelf-Life of Packaged Model Lipid Food System. Packages Mylar—l Mylar—2 Mylar-3 Saran-2 Shelf-Life 358 360 357 360 (Hours) It can be seen from Table 5 that there is little or no difference in the predicted shelf-life of the model food system packaged in the respective test materials, based on the simulation model developed. These results can be rationalized, if the rate of oxygen consumption (m1 OZ/hr./gram product) is slower than the rate of O2 permeation through the respective packages. The oxygen permeation rates of the two test packages are of the same order of magnitude; namely, 2.25 and 5.16 ml 02 (STP)/mZ/hr. for Saran and Mylar, respectively. From these permeation rates and the total package surface area of approximately 192 cm2 (package dimensions of 12 cm x 8 cm), it can be estimated that for the Mylar package approximately 0.1 m1 (STP) of oxygen will permeate through per hour. This represents a minimum value, since the studies were conducted at 50°C. The total oxygen con— sumption will be approximately 0.076 ml OZ/hr. at this temperature. Further, the oxidation rate is assumed to be independent of oxygen partial pressure. 35 Actual Storage Studies Results Actual storage tests were carried out according to the procedure described in the previous Section, under the following package conditions (Table 6). Table 6. Package Conditions for Actual Storage Test. Packages Variables Mylar-l Mylar-2 Mylar-3 Saran-2 Initial Headspace 55 57 60 78 Volume (ml) Weight of Model 4.40 4.98 4.48 4.32 Lipid System (g) Surface Area (m2) 0.0199 0.0189 0.0200 0.0153 Oxygen 5.16 5.16 5.16 2.35 Permeability Rate (m1 02/m2 hr. atm.) For these studies the initial headspace volume was not determined directly in order to eliminate any effect of the carbon dioxide gas, introduced to the package for the headspace determination. The initial headspace volume was, therefore, calculated based on the final headspace volume, using the following Equation (13). Hi = (Hf + 4.0) x (1 - POzf)/(l - P021) (13) where: H. = initial headspace volume (ml), Hf = final headspace volume (ml), 0 = the amount of gas evacuated from the headspace for sampling (m1), 36 P02. initial oxygen partial pressure 1 (atm.), PO - final oxygen partial pressure 2f (atm.). The change in the oxygen partial pressure within the respective packages is shown as a function of time in Table 7. Table 7. Change in Oxygen Partial Pressure (Atm.) within the Respective Packages (Actual Storage Test). Time Packages (Hours) Mylar-1 Mylar-2 Mylar-3 Saran-2 0 0.206 0.207 0.207 0.207 46 0.196 0.194 0.196 0.175 115 0.174 0.170 0.175 0.162 161 0.162 0.157 0.164 0.149 203 0.148 0.142 0.151 0.133 252 0.139 0.131 0.144 0.119 296 0.126 0.116 0.132 0.108 329 0.114 0.099 0.121 0.094 417 0.096 0.074 0.107 0.067 466 0.080 0.053 0.094 0.049 537 0.044 0.025 0.061 0.013 587 0.015 0.012 0.029 0.012 611 0.013 0.013 0.015 0.011 As shown in Tables 3 and 7, a fairly good agreement was obtained for the oxygen partial pressure change within the packages between predicted results and actual storage 37 test results. This indicates that oxygen consumption by the model lipid system is also simulated reasonably well by the developed computer program. The shelf-life determination can then be considered fairly reliable. DISCUSSION Consideration of Possible Error in the Oxidation Rate Studies In the system which was employed to determine the oxidation rate and the oxidation rate equation, the temper- ature and relative humidity were held constant, and the rate of oxidation was monitored as a function of continuously changing oxygen partial pressure. The oxidation rate was determined by a calculation based on the oxygen partial pressure change and the headspace volume of the system. In this procedure, a certain degree of error is introduced due to both sampling and instrument (i.e. gas chromatograph) accuracy. The oxidation rate was calculated, based on the change in the amount of oxygen within the package as a func- tion of time. The oxidation rate is assumed equal to the slope, where oxygen quantity within the package is plotted as a function of storage time. The effect of both instrumental and sampling error on the oxidation rate studies are discussed below. The amount of oxygen within the package headspace is calculated as follows: 02(t) = V(t) X P02(t) (14) 38 39 actual amount of oxygen present where: 02(t) within the headspace volume at time, t, V(t) = actual headspace volume at time, t, P02(t) = actual oxygen partial pressure at time, t. In the present studies, both headspace volume and oxygen partial pressure were determined by gas chromatography measurements. It is assumed that the percent error of both sampling and instrumentation is iA% for the headspace volume measurement, and iB% for the oxygen partial pressure measure- ment. The measured headspace volume at time, t, with maximum possible error can be described as follows: V*(t) = (113) x V(t) (15) where: V*(t) measured headspace volume with maximum possible error at time, t, V(t) actual headspace volume, a =J1. 10° The measured oxygen partial pressure at time, t, with maximum possible error also can be shown: P02*(t) = (lib) x P02(t) (16) where: P02*(t) measured headspace volume with maximum possible error at time, t, P02(t) = actual headspace volume at time, t, b =__B_, 1 0 40 From Equations (14), (15) and (16), the relation- ship between 02(t) and 02*(t) can be obtained as follows: 02*(t) = V*(t) X P02*(t) 02*(t) = (lia)(1:b) x V(t) x P02(t) 02*(t) = (lia)(lib) x 02(t) (17) measured amount of oxygen present within the headspace volume at time, t. where: 02*(t) The amount of oxygen within the headspace was obtained as a function of time. The measured amount of oxygen falling within the region of (lia)(lib) ° 02(t) can be represented graphically for oxygen quantity values, determined at t1 and t2. Amount of Oxygen Present within the Headspace Volume (m1) (1+a) (1+b) ~02(t1) -— q 02(t1)-1r (1-a)(1-b)° 02(t1)‘” (1+a) (1+b) ° 02(112)" 02(t2)" (l-a)(1-b)° 02(t211 Time (Hours) where: I = actual amount of oxygen, 0 = minimum or maximum possible measured amounts of oxygen, Gmax = maximum possible calculated oxidation ° rate, 6min = minimum possible calculated oxidation ° rate, 6a = actual oxidation rate. Figure 4. Illustration for Determination of Oxidation Rate. 41 42 The results presented in Table 8 illustrate how such error factors can affect the oxidation rate, using as an example, data from the FLEX-CAN Sample 1. Data: t1 = 119 hours, t2 = 142 hours; 02(t1) = 11.72 ml 02, 02(t2) = 9.99 ml 02; Oxidation Rate = 0.076 (ml Oz/hr.). Table 8. Effect of Error Factors on the Oxidation Rate. Possible Measured Value . . with Error (m1) Ox1dat10n Error Rate Factor 02(t1) 02(t2) (m1 Oz/hr.) a b Max. Min. Max. Min. Max. Min. 0.005 0.005 11.84 11.60 10.09 9.89 0.085 0.066 0.005 0.01 11.90 11.54 10.14 9.84 0.090 0.061 0.01 0.005 11.90 11.54 10.14 9.84 0.090 0.061 0.01 0.01 11.96 11.49 10.19 9.79 0.094 0.057 As shown, the effect of error factors involving both sampling and instrumentation on the oxidation rate calculation can be significant. In this study, a number of measurements were made and linear regression analysis was carried out on these measurements. Discussion of Simulation Model In most examples of predicting shelf-life by cal- culation, the time-step selected can influence the results. However, there is a critical value for the time-step function, below which it has no influence on the computed 43 shelf—life value. The magnitude of the time-step, therefore, has to be below this critical value for appropriate appli- cation of the iteration technique. The effect of the time- step function on the calculated change in oxygen partial pressure, as a function of time, was determined using data obtained experimentally for the Mylar M24-1 package. The results are summarized in Table 9, where oxygen partial pressure change was computed for time-step values of 0.1, 1.0 and 5.0 hours, respectively. Table 9. Effect of the Time-Step Function on the Computed Oxygen Partial Pressure Change. Oxygen Partial Pressure (Atm.) Time Time—Step Time—Step Time—Step (Hours) 0.1 Hour 1.0 Hour 5.0 Hours 0 .208 .208 .208 100 .159 .159 .158 200 .114 .113 .113 300 .074 .073 .073 400 .040 .039 .039 500 .012 .012 .011 As shown, there is an average of about 1% differ- ence between the computed values of oxygen partial pressure change as a function of storage time for the three time-step levels evaluated. This indicates that the value of 5.0 hours is already below the critical value. In this study, a time- step of 1.0 hour was used, assuming that the value was well below the critical time-step level. 44 Another factor which may influence the calculated results is the permeation of nitrogen from the headspace within the package to the external package environment, as a result of a nitrogen partial pressure difference between the internal and external package environments. The nitro— gen partial pressure (i.e. concentration) within the package will increase as the oxygen partial pressure decreases because of oxygen consumption by the model lipid food system, resulting in a decrease in the internal volume of the pack— age. Since the nitrogen partial pressure remains constant (= 0.792 atm.) outside the package a nitrogen partial pres— sure difference occurs, resulting in the permeation of nitrogen through the package into the external environment. Although the permeation rate of nitrogen was not determined, it is possible to estimate the nitrogen permeation rate of the respective films, based on the permeability constants of nitrogen and oxygen, which are available in the literature (Agranoff, 1977). The estimated permeation rates for the respective films are listed below: Mylar: 62.0 [ml N2 (STP)/m2 - 24 hr. - atm.] at 50°C, Saran: 28.3 [ml N2 (STP)/m2 ° 24 hr. - atm.] at 50°C. The effect of the nitrogen permeation was evaluated using these data. The computer program which was written to calculate the oxygen partial pressure change as a function of storage time was modified to include the permeation of nitro- gen through the respective permeable materials. The computed oxygen partial pressure values for the Mylar M24-l are 45 summarized in Table 10 and compared to the values previously determined, where the quantity of nitrogen within the package was assumed constant. Table 10. Effect of Permeation of Nitrogen on the Oxygen Partial Pressure Change within the PackageIa). Oxygen Partial Pressure (Atm.) Storage Time With the Effect Without the Effect (Hours) of Nitrogen of Nitrogen 0 .208 .208 100 .159 .159 200 .114 .113 300 .075 .073 400 .040 .039 500 .011 .012 (a) Data of Mylar M24—1 Sample are used for the computation. As shown, there appears to be a negligible effect on the computed oxygen partial pressure change, due to the permeation of nitrogen. Therefore, Assumption 4 for the simulation model was considered valid. SUMMARY AND CONCLUSIONS The shelf—life prediction study was carried out by using a mathematical model to simulate the interaction mechanism of oxygen permeation through a package and oxygen consumption by an oxidizable product. The predicted oxygen partial pressure within the package, based on the simulation model developed, showed a fair agreement with experimentally determined values obtained from actual storage test data. The model developed simulates the mechanism of oxygen con- sumption within the package reasonably well for a first approximation. However, improved agreement may be obtained in future studies by improvement of analytical methodology and/or refinement in the simulation model. Based on the present studies and results obtained, the following future studies are proposed. 1. The rate of oxygen consumption as a function of extent of oxidation in the food product should be determined prior to the shelf-life studies. 2. The change in quality index should be monitored by following the oxygen partial pressure changes, as well as an independent chemical analysis of the product. Correlation should be established between the oxygen uptake and chemical analysis. Such a correlation 46 47 establishes and/or validates the assumption that the loss of quality index due to oxidation is related directly to the extent of oxygen uptake by the product. The influence of other independent variables (i.e. relative humidity and temperature) which were kept constant in the present studies should be studied. After completion of the model food study, an actual food system should be evaluated. APPENDICES APPENDIX A TABLES FOR COMPOSITION AND DISTRIBUTION OF MODEL LIPID FOOD SYSTEM Table 11. Composition of Model Lipid Food System. Component Relative Percent Soybean oil 65.5 CMC 6.9 Tween 20 3.5 Buffer salt 19.2 Water 4.9 Table 12. Lipid Distribution in Model Lipid Food System. Weight of Weight of Sample Sample (g) Oil (g) [B]/[A] x 100 Number [A] 41B] (%) 1 3.90 2.41 61.79 2 3.59 2.29 63.71 3 3.27 2.07 63.22 4 4.28 2.62 61.33 48 APPENDIX B HEADSPACE VOLUME DETERMINATION Control studies were carried out to determine the effect of sorption of carbon dioxide by the food system. Two FLEX-CAN Pouches containing the model lipid system were prepared as previously described for the oxidation rate studies. Pure carbon dioxide (1 ml) was injected into the headspace of each FLEX-CAN through a sampling septum. The concentration of carbon dioxide within the FLEX-CAN was determined by gas chromatography after 10 minutes, 1 hour and 3 hours. As shown in Table 13, no sorption of carbon dioxide was detected between 10 minutes to 3 hours after pure carbon dioxide injection. The validity of the relationship, shown in Equation (11), was established by the following preliminary study. Using FLEX-CAN Pouches of known headspace volume [i.e. 50, 100 and 140 ml (STP)], headspace volume determi— nations were carried out by a serial dilution technique and the calculated and actual headspace volume values compared. As shown in Table 14, the calculated results agreed well with the actual volume. 49 Table 13. Change in the Concentration of the Carbon Dioxide within the FLEX-CAN by Sorption as a Function of Time. Time After Pure Concentration of Carbon Dioxide Carbon Dioxide Injection within the FLEX-CAN (%) 10 minutes 1.48 1 hour 1.47 3 hours 1.50 Table 14. Comparison of the Determined Headspace Volume to the Actual Volume. Experimentally Actual Headspace Determined Headspace Error Volume [ml (STP)] Volume [m1 (STP)] (%) 50 50 0 50 50 0 100 104 4.0 100 100 0 140 139 2. 140 131 6.4 Average % Error 2.2 50 I l . .Ix Tm oommmomom flHEo 095mmoam mmooefih oumm Mom .HL\HEV coHuQESmcou oommmomom onEwm ocu manna: oesao> Hwfluhmm :oflpmesmzou opmm somxxo mo map cflcpflz %o cowoapflz oowmmomo: flBoomxxo :omxxo :oflumESmcoo oesfio> comxxo mo flmooasao> mo oESHo> cowxxo oEsHo> .H onEmm z