THE RETENTION OF SYNTHETIC PHENOUC ANTIOXlDANTS IN ' . MODEL FREEZE-DRE!) F000 SYSTEMS Dissertation for the Degree of Ph. D. mum STATE UNWERSITY ALLERé W. KJRLEIS 1975 This is to certify that the thesis entitled The Retention of Synthetic Phenolic Antioxidants in Model Freeze-Dried Food Systems presented by Allen W. Kirleis has been accepted towards fulfillment of the requirements for Ph . D; degree in .Eoo.d_S.c.i.en.ce If MMSM Major professor WW5 0-7639 "" ‘fi 3” amounts av llUAli & 'SllNS' BOOK llllllliRY lllfi. uemm amazes srntncropr. mm; *- ..—-—--w 3 [min ill] [Willi lllllll 3 1293 01088 8067 We"; w“ 100 0165 - ‘ Ia ""- é; ’ 1‘35 firth-r a 6 if Jon- 2" ,x ABSTRACT THE RETENTION OF SYNTHETIC PHENOLIC ANTIOXIDANTS IN MODEL FREEZE-DRIED FOOD SYSTEMS BY Allen W. Kirleis The retention of three synthetic antioxidants: mono- tertiary-butylhydroquinone (TBHQ), butylated hydroxyanisole (BHA) and butylated hydroxytoluene (BHT) was investigated in model freeze-dried systems designed to simulate a high lipid food. The system was composed of 15 percent pregelatinized wheat starch, five percent lipid (soybean or coconut oil) and 80 percent water. The variables studied were antioxidant level, freezing rate, lipid type and drying time. Data were interpreted by current volatile retention theories. Antioxidant determinations were carried out by reversed phase high-resolution liquid chromatography on antioxidant- lipid extracts of the freeze-dried systems. The extraction and chromatographic techniques used yielded recoveries of 96 to 100 percent antioxidant at the 0.01 percent anti- oxidant level. Among the antioxidants investigated, TBHQ was retained to the greatest extent under all processing parameters, and BHA to a greater degree than BHT. The control samples initially containing 400 ug antioxidant per g soybean oil, Allen W. Kirleis slowly frozen and freeze-dried for 12.5 hr at a chamber pressure of 5 x 10-3 torr on the average retained 32, 27 and 16 percent TBHQ, ERA and BHT, respectively. The effects of freezing rate on antioxidant retention in model systems containing soybean oil was determined by freeze-drying samples subjected to either a slow or a rapid freezing treatment. The slow freezing treatment (control samples) improved antioxidant retention by about 1.2x. Significant differences in retention between freezing rates were found for systems containing TBHQ but not for systems containing BHA or BHT. The influence of oil type on antioxidant retention was investigated in a saturated coconut oil system and in an unsaturated soybean oil system. Antioxidant retention was greater in the saturated coconut oil systems than in the unsaturated soybean oil systems (control samples). Retention in the saturated oil samples was improved about 1.2x for TBHQ and 1.3x for BHA or BHT. The effects of extended freeze-drying of soybean oil samples (control samples) on antioxidant retention was determined. It was found that BHA and BHT were completely removed when the normal drying time was doubled and that TBHQ content was reduced by an additional 19 percent over this same period. Allen W. Kirleis At the two initial antioxidants concentrations studied (200 and 400 ug Per g soybean oil) BHA, BHT and TBHQ re- tention expressed as percentage of initial concentration showed little difference. Although the retention per unit weight of oil was greater in all cases for samples con- taining the greatest initial amount of antioxidant. It was found that the retention of BHA and BHT was directly related to the final moisture content of the freeze-dried system in the water concentration range be- tween one to five percent. TBHQ retention however, was found to be independent of the final moisture content of the system over the same water concentration range. The retention of prOpyl gallate was not investigated because, it was found to be absorbed to the wheat starch component of the model system. This made it impossible to quantitatively extract prOpyl gallate from the freeze- dried system by conventional methods. THE RETENTION OF SYNTHETIC PHENOLIC ANTIOXIDANTS IN MODEL FREEZE-DRIED FOOD SYSTEMS BY IMO" Allen WLIKirleis A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Food Science and Human Nutrition 1975 Dedicated To My Father Eugene G. Kirleis ii ACKNOWLEDGEMENTS The author extends his sincere appreciation to Dr. C. M. Stine for his guidance and support throughout the course of this graduate program. Appreciation and thanks are extended to members of the guidance committee, Drs. P. Markakis, Department of Food Science and Human Nutrition, L. L. Bieber and H. A. Lillevik, Department of Biochemistry for their advice and effort in reading this manuscript. The author is also indebted to Dr. J. R. Brunner for the use of his liquid chromatography apparatus, the absence of which would have made this study improbable. The technical assistance of Mr. Lyman C. Aldrich Jr. is also gratefully appreciated. Genuine appreciation is also offered to the Department of Food Science and Human Nutrition for the financial assistance and for the use of its fine research facilities. The author expresses his grateful appreciation to his parents, for their understanding and encouragement. TABLE OF CONTENTS DEDICATION . . . . . . . . . . . . . . . . ACKNOWLEDGEMENTS . . . . . . . . . . . . . LIST OF TABLES . . . . . . . . . . . . . . LIST OF FIGURES . . . . . . . . . . . . . INTRODUCT ION . C O O O I O O O O O O O O 0 LITERATURE REVIEW . . . . . . . . . . . . . . . . . Oxidative Deterioration of Food Lipids . . . . Mechanism of Autoxidation of Food Lipids . Lipid Oxidation in Freeze-Dried Foods . . AntiOXidants O O O O O O O O O O O O O O O O 0 Types of Antioxidants . . . . . . . . . . Function of Fat and Oil Antioxidants . . . Antioxidants Used in Foods . . . . . . . . Antioxidant Effectiveness in Low Moisture FOOdS O O O O O I O O O O 0' O I O O O Antioxidant Volatility . . . . . . . . . . Retention of Volatile Organic Compounds During Freeze-Drying . . . . . . . . . . . Basic Observations . .,. . . . . . . . . . Volatile Retention Mechanism . . . . . . . EXPERIMENTAL PROCEDURES . . . . . . . . . . . . . . Preparation of Model System Ingredients . . . . Defatting of Wheat Starch . . . . . . . . Preparation of Pregelatinized Wheat StarCh O O O I O O O O O O O O O 0 0 Storage of Soybean and Coconut Oil . . . . iv Page ii iii vii ix cocooxwwwI-I 10 12 15 18 20 21 26 38 38 38 38 39 Preparation of Antioxidant Spiked Soybean and Coconut Oil . . . . . . . . . . Preparation and Freeze-Drying of Model systems 0 I O O O O O O O O O I O I O 0 Model System Composition . . . . . . . . Model System Preparation . . . . . . . . Freezing of Model Systems . . . . . . . Freeze-Drying of Model Systems . . . . . Experimental Variables . . . . . . . . . . . Antioxidant Level . . . . . . . . . . . Freezing Rate . . . . . . . . . . . . . Oil Type . . . . . . . . . . . . . . . . ANALYTICAL PROCEDURES 0 C O O O O O O O C O O O 0 Determination of Isomeric Purity of BHA . . . Standard Preparation . . . . . . . . . . Sample Preparation . . . . . . . . . . . Gas Chromatography . . . . . . . . . . . Calculations . . . . . . . . . . . . . . Determination of Fatty Acid Composition of Soybean and Coconut Oil . . . . . . . . Methyl Ester Preparation . . . . . . . . Gas Chromatography . . . . . . . . . . . calculations 0 0 O O O 0 O O I O O O 0 Differential Thermal Analysis of Model systems 0 O O O O O O O O O O O O O O O Peroxide Value Determination of Freeze-Dried MOdel system I O O O O O O O O O O O O O Quantitation of Antioxidants in Model systems 0 O O O O O O O O O O O O O O O Solvent Purification and Testing . . . . Chromatographic Apparatus . . . . . . . Chromatographic Conditions . . . . . . . Extraction of BHA, BHT and TBHQ From Dry Model Systems . . . . . . . . . Quantitation of BHA, BHT and TBHQ in Dry Model Systems . . . . . . . . . 40 41 41 41 42 42 43 44 44 44 45 45 45 45 46 46 47 47 48 48 49 49 51 51 51 53 54 56 Recovery of Added Antioxidants From Freeze- Dried Model Systems . . . . . . . . . 58 Statistical Interpretation . . . . . . . . . . 58 RESULTS AND DISCUSSION . . . . . . . . . . . . . . 59 Analyses of Model System Components . . . . . . 59 Soybean and Coconut Oil . . . . . . . . . 59 BHA Isomeric Purity . . . . . . . . . . . 61 Determination of Fat in Wheat Starch . . . 61 Extraction and Determination of Antioxidants. . 62 Reliability of LC System . . . . . . . . . 62 Reliability of Extraction Methods . . . . 66 Antioxidant Calibration Graphs . . . . . . 68 PrOpyl Gallate Experiments . . . . . . . . . . 69 Model System PrOperties . . . . . . . . . . . . 75 Model System Emulsions . . . . . . . . . . 76 Model System Freezing and Freeze Drying. . 76 Oxidative Changes in Oils During Model System Preparation and Freeze- DrYing O O I O O O O O O O O O O O O 79 Retention of Antioxidants in Freeze Dried Model Systems . . . . . . . . . . . . . . 80 Mechanism of Antioxidant Retention . . . . 93 Effects of Initial Antioxidant Concentration . . . . . . . . . . . . 98 Effects of Freezing Rate . . . . . . . . . 99 Effects of Oil Type . . . . . . . . . . . 101 Extension to Other Work . . . . . . . . . 102 SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . 104 APPENDIX A Detection of Peroxides in Organic SOlvents O O I I O O O I O O O O O O O 107 APPENDIX B LC Antioxidant Calibration Values . . . 108 APPENDIX C Statistical Analysis of Antioxidant Retention Data . . . . . . . . . . . . 111 LIST OF REFERENCES . . . . . . . . . . . . . . . . . 114 vi Table 1. 10. LIST OF TABLES Common synthetic antioxidants used in fOOdS O O O 0 O O O O O O O O O O O 0 O 0 Quantity of diBHA added to antioxidant extracts of freeze-dried model systems and sample sized used for LC analysis of antioxidants . . . . . . . . . . . . . . . . Antioxidant and internal standard weights used to prepare calibration graphs . . . . . Characteristics and fatty acid composition of soybean and coconut oil . . . . . . . . . Reproducibility of chromatographic system 0 O O C O I O O O O O O O O I O O O O Recoveries of antioxidants added to freeze- dried model systems containing soybean and coconut oil . . . . . . . . . . . . . . . . Percentage of oil found in freeze-dried model systems containing soybean or coconut oil by various extraction procedures . . . . . . . Composition of samples used in PG recovery 8 tudy O O O O O O O O O O O O O O O I O O O Peroxide value of soybean and coconut oil used to prepare model systems and of the same oil after extraction from freeze-dried model systems . . . . . . . . . . . . . . . Page 13 55 57 60 65 68 69 75 80 Retention of BHA by freeze-dried model systems containing soybean oil placed at various platen locations . . . . . . . . . . . . . . vii 82 Table Page 11. Retention of BHA by freeze-dried model systems subjected to different treatments . . . . . . 86 12. Retention of BHT by freeze-dried model systems subjected to different treatments . . . . . . 87 13. Retention of TBHQ by freeze-dried model systems subjected to different treatments . . 88 14. Correlation coefficients of antioxidant retention to final moisture content in freeze- dried model systems . . . . . . . . . . . . . 90 15. The effect of extended freeze-drying on the 1088 Of BHA, BHT, Or TBHQ o o o o o o o o o o 92 Bl. Relative calibration values obtained for BHA added to freeze-dried model systems containing soybean and coconut oil . . . . . . . . . . . 108 82. Relative calibration values obtained for BHT added to freeze-dried model systems containing soybean and coconut oil . . . . . . . . . . . 109 BB. Relative calibration values obtained for TBHQ added to freeze-dried model systems containing soybean and coconut oil . . . . . . . . . . . 110 C1. Analysis of variance of BHA and BHT percentage retention per one percent moisture in freeze- dried model systems subjected to various treatments (antioxidant level, freezing rate, oil type) . . . . . . . . . . . . . . . . . . 111 CZ. Main factor means for BHA and BHT percentage retention per one percent moisture in freeze- dried model systems subjected to various treatments (antioxodant level, freezing rate, oil type) . . . . . . . . . . . . . . . . . . 112 C3. Analysis of variance of TBHQ retention in freeze-dried model systems subjected to various treatments (antioxidant level, freezing rate, oil type) . . . . . . . . . . 113 C4. Main factor means for TBHQ retention in freeze-dried model systems subjected to various treatments (antioxidant level, freezing rate, oil type) . . . . . . . . . . 113 viii Figure l. 2. LIST OF FIGURES Page Vapor pressure of BHA, BHT and TBHQ at various temperatures . . . . . . . . . . . . . 19 Schematic of liquid chromatographic SYSteIn O O O O O O I O O O O .0 O O I O O O O 52 Typical LC separation of antioxidants and diBHA in SOYbean Oil 0 O O O O O I O I O O O O 63 Relative calibration graphs for BHA added to freeze-dried model systems containing soybean and coconut oil . . . . . . . . . . . . . . . 71 Relative calibration graphs for BHT added to freeze-dried model systems containing soybean and coconut Oil 0 O O O O O O O O O O O O O O 7 2 Relative calibration graphs for TBHQ added to freeze-dried model systems containing soybean and coconut oil . . . . . . . . . . . . . . . 73 Typical temperature profiles of freeze-drying model systems subjected to a slow or rapid freeze treatment . . . . . . . . . . . . . . . 77 Effect of final moisture content on BHA retention in freeze-dried model systems containing soybean oil. Initial BHA content 400 ug per 9 oil . . . . . . . . . . . . . . . 83 Effects of final moisture content on BHA retention in freeze-dried model systems containing soybean oil. Initial BHA content 200 ug per 9 oil . . . . . . . . . . . . . . . 84 ix INTRODUCTION In recent times, there has been an increase in the production of many fat containing freeze-dried foods e.g. eggs, dairy products, soup ingredients, and many "backpack rations". One of the major barriers to large- scale applications of freeze-drying has been problems in the storage stability of foods that are susceptible to oxidation. In order for such foods to have reasonable shelf life and acceptable flavor characteristics, protec- tive additives to prevent oxidation, such as antioxidants, are often added before dehydration. Based on studies by Radanovics (1969) and Labuza 35 31. (1969) it can be concluded that the effectiveness of synthetic phenolic antioxidants for controlling oxi- dation in freeze-dried foods is limited. This ineffec- tiveness may be in part due to their volatization and escape from the freeze-dried food during dehydration; since, these compounds themselves have sufficiently high vapor pressure that they can volatize at room temperatures. However, very little information on the retention of syn- thetic antioxidants during freeze-drying has been reported. This project was undertaken to investigate the effect of processing variables and lipid type on the retention of three commonly used synthetic phenolic antioxidants in model freeze-dried systems designed to simulate a high lipid food. The simplified model system approach to the problem offers the possibility of developing information applicable to the control of processing factors which mini- mize antioxidant retention in freeze-dried foods of high fat content. LITERATURE REVIEW Oxidative Deterioration of Food Lipids The oxidative deterioration of food lipids involves, primarily, autoxidative reactions, which are accompanied by various secondary reactions having oxidative or non- oxidative character. In general, autoxidation may be de- fined as the reaction of any material with molecular oxy- gen. Several mechanisms have been discovered or prOposed for the autoxidation of various organic materials, and have been reviewed by Horner (1961) and Labuza (1971). The principle mechanisms that are involved in the primary autoxidation of fatty materials have been elucidated fairly well. However, a number of secondary reactions, some of which are virtually unstudied, are involved in the oxidative deterioration of lipid materials, and information concerning the mechanisms involved in these secondary reactions is, in some cases, limited. Mechanism of Autoxidation of Food Lipids From the standpoint of food oxidation, the important lipids are the unsaturated fatty acid moieties, particu- larly oleate, linoleate, and linolenate, the predominant ones in foods (Labuza, 1971). The susceptibility and rate of oxidation of these fatty acid constituents increases in a somewhat geometric fashion in relation to their degree of unsaturation. Oxidation of the unsaturated fatty acids has been described in depth in a number of articles (Uri, 1961; Lundberg, 1962; Scott, 1965; Emanuel and LyaskovsKaya, 1967; Labuza, 1971) and is almost universally accepted as a free radical chain mechanism taking place in three stages as described below. Initiation - This probably corresponds to the oxida- tion induction period of a fat or oil, and during this stage lipid molecules convert to unstable fatty free radicals which can catalyze further free radical formation in the substrate. This may be depicted as: RH J-R- + H- (fat molecule) (fatty free radical) Various agents, such as light(especially in the UV region), heat and heavy metals (particularly c0pper and iron) are principal initiators of autoxidation. PrOpagation - Fatty free radicals which have formed can combine with molecular (atmospheric) oxygen to form peroxide free radicals which can react with the substrate to form more fatty free radicals and hydroperoxides. The hydroperoxides are formed very slowly in the beginning of storage, but they gradually accumulate. The period of time when changes in the peroxide value are nearly imperceptible is called the induction period. The end of the induction period is the stage of autoxidation when the peroxide value of fat suddenly begins to increase. The propagation chain reaction may be depicted as: R- + 02 :h— ROO- (peroxide free radical) ROO- + RH ~—’- ROOH + R- (hydroperoxide) ROOH ~4t- R0- + ~OH -OH + RH 1’. R- + H20 etc. During propagation, especially in the presence of catalytic agents, decomposition of the hydrOperoxide leads to forma- tion of a wide variety of aldehydes, ketones, acids, etc., responsible for the rancid off flavors produced which re- duce shelf life as well as nutritional value of food fats and oils. What makes this reaction so important, eSpe- cially in its early stages, is that many of the small molecular weight compounds formed have very low odor thresh- hold values. Thus, if they are objectionable in odor, only a few ppm or ppb are needed to give food an unacceptable odor. Termination - Termination of the oxidation chain reaction, in the absence of antioxidants, occurs when the free radicals (autocatalysts) are deactivated or destroyed. This may occur in various ways, such as: R-+R- ih—RR R- + ROO- ab— ROOR etc. Several termination reactions are also possible, producing ketones or alcohols. Obviously, the mechanism of autoxidation in fatty acids is a much more complex reaction than the direct addition of oxygen to the substrate. Details of the reaction mechanism including the singlet state oxygen mechanism of initiation have been thoroughly reviewed by Labuza (1971), and should be consulted if more detail is desired. Lipid Oxidation in Freeze-Dried Foods The oxidation of lipids is particularly prominent for freeze-dried foods because of the very large internal sur- face areas of freeze-dried materials and because of the very low moisture contents which are usually achieved during freeze-drying (Karel, 1963). Lipid oxidation leads to a number of undesirable changes in freeze-dried foods in- cluding production of off-flavor, off-odor, and changes in color, texture and rehydratability (Goldblith and Tannenbaum, 1966). These changes also lower the nutritional value and limit the shelf life of the food. In several research studies (Karel, 1963: Goldblith, 25 ‘31. 1963; Goldblith and Tannenbaum, 1966) in which the rate of oxidation of linoleic acid was measured as a func— tion of oxygen partial pressure it was found that the sur- face area exposed to the oxygen is an important variable. In their studies oxidation rates for linoleic acid exposed to oxygen in bulk or in well-agitated vessels decreased with decreasing oxygen partial pressure, but when linoleic acid was diapersed on powdered cellulose the rate of oxidation remained high and was independent of oxygen partial pressure down to partial pressures less than 0.005 to 0.01 atm. The lipid distribution in freeze-dried foods is expected to approach this colloidally disPersed case, especially if there has been melting of the lipids during drying (King, 1970a). These results show the need for added antioxidants along with very complete removal of oxygen before packaging and for the use of relatively oxygen impermeable packaging materials to insure long shelf life to many foodstuffs. Lea (1958a) has shown that the amount of water adsorbed on food systems has a very significant effect on the rate of lipid oxidation. He presented evidence showing that increasing concentrations of water slows the oxidation of lipids. It has been postulated that the calculated mono- layer value of moisture adsorption defines the region of highest stability for freeze-dried food items (Salwin,l962). The importance of adsorbed water in model systems undergoing lipid oxidation has been studied by Maloney st ‘21. (1966); Labuza gt_gl. (1966) and Karel 22 El; (1967). In their work it was shown that the water exhibited two effects: hydration of some metals that catalyze lipid oxidation, rendering them inactive; and formation of hydro— gen bonds with hydrogen-peroxides produced, thereby pre- venting them entering into initiation reactions. Lipid oxidation products are also known to participate in reactions with nonlipid constituents in foods. At extremely low water contents proteins may be degraded due to reactions with free radicals originating from oxidation of lipids (Zirlin and Karel, 1969). In the same work it was observed that at higher humidities the protein was crosslinked into higher molecular weight units. Both effects can lead to undesirable textural changes and loss of biological value in foods. Oxidation products of fatty acids may also participate in nonenzymatic browning if protein is present in the system. LeRoux and Tannenbaum (1969) studied this interac- tion in a system of casein-methyl linoleate at several humidities. Their results indicate that at low humidities little browning pigments were produced; however, when their systems were transferred to higher humidities the pigments were produced rapidly. Antioxidants Types of Antioxidants Modern methods of food processing and handling require the addition of certain chemicals in order to insure high quality and a shelf life equal to, if not greater than, the normal distribution and marketing time of foods. Anti- oxidants, as a class of chemicals, are ubiquitous in their presence by addition to foods including cereals, bakery products, snack foods, animal feeds, intermediate moisture, and dehydrated foods, as well as incorporation into pack- aging materials. They do not improve the quality of the product, but, rather, maintain it by preventing oxidation of labile lipid components. They also are effective at very low concentrations, as would be expected for inhibition of a free radical chain reaction. To prevent autoxidation from occurring, or at least to minimize it, several types of antioxidants can be utilized. Scott (1965) has reviewed the whole area of antioxidant mechanisms and classifies them into three types: Type I - Free radical chain stoppers, of which BHA (butylated hydroxyanisole or 2- and 3- isomers of tertiary butyl-4-hydroxyanisole), BHT (butylated hydroxytoluene or 2,6 ditertiary butyl-4-methyl phenol), PG [propyl gallate or grpropyl ester of 3,4,5-trihydroxygenzoic acid (gallic acid)}, TBHQ (mono-tertiary-butylhydroquinone), tocopherol, and gum guaiac are examples, in foods. These are primarily phenolic type compounds which donate a hydrogen to a radical. Type II - Inhibitors of free radical production in foods, such as the chelating agents EDTA (ethylenediamine- tetraacetic acid), citric acid, and various forms of ascorbic acid. They act mainly by tying up metal catalysts. 10 Type III - Environmental factors, such as lowering of oxygen partial pressure in the package or holding at a critical moisture content for a dehydrated food. Since the antioxidants used in this study all fit into the Type I classification the remaining discussion of antioxidants will deal mainly with Type I antioxidants used for preventing fat oxidation. Function of Fat and Oil Antioxidants Since the autoxidation of fats and oils is initiated and pr0pagated by the formation of free radicals, Type I antioxidants are substances that interfere with the oxida- tion by inactivating the free radicals that propagate the chain reaction and thereby extend the shelf life of the substrate until autoxidation takes place. However, con- flicting evidence appears in the literature with regard to the basic mechanism of antioxidant action, which will not be discussed here. It is generally agreed that anti- oxidants work primarily be breaking the free radical chain reaction through either the removal of the alkylperoxy or alkyl radicals from the chain step according to the reac- tions below: ROO- + AH :F- ROOH + A- R- + AH 4" RH + A- and. . . A- + X :>- non-radical where ROO-, R-, AH, A-, and X are alkylperoxy free radical, 11 alkyl free radical, antioxidant, antioxidant free radical, and some other moiety, respectively (Uri, 1961; Lundberg, 1962; Stuckey, 1962; Scott, 1965; Labuza, 1971). It is essential that the antioxidant free radical, thus formed, not have capabilities of initiating or prOpagating the oxidation reaction. This is the case with phenolic com- pounds, since the resultant phenoxy radical is a "stable resonance hybrid" (Altwicker, 1967). Scott (1965) has reviewed the rules of substitution of phenolic antioxidants at the ortho, meta and para positions in terms of anti- oxidant effectiveness. Basically the more effective the resonance forms of the antioxidant free radical, the better the antioxidant. Therefore, substitution at the ortho and para positions are much more effective than at the meta position because of the greater number of resonance forms. Another important factor contributing to phenolic anti- oxidant effectiveness, as pointed out by Sherwin (1972), is the size of the substituting group. A bulky attachment helps protect the antioxidant radical and yields added stability towards further reaction, but also makes it more difficult to react with the peroxy radical. The concentration of an antioxidant in a food fat or oil influences its stability. Labuza, _£__1, (1969) found that as antioxidant concentration is increased beyond a certain limit oxidation is actually enhanced and protection no longer occurs. Pakorny (1971) reported that at 12 relatively low antioxidant concentrations, e.g. between 0.001 and 0.02 percent, the protection factor or induction time of a fat is directly proportional to antioxidant concentration. Thus, a prOper balance is sought between the quantity which provides maximum stabilization and that which may participate in the chain reaction and thereby intensify oxidation. Another factor limiting the effectiveness of antioxi- dants in food fats and oils is the time of addition. Since, phenolic antioxidants function as free radical inhibitors, it is essential that they be added to a food fat or oil before the peroxide level is too high (Sherwin, 1972). There are reports in the literature (Scott, 1965 and Mabrouk and Dugan, 1961) where antioxidants were added to an oxidized olefin system with no effect occurring. As in cases of food quality, once the product is rancid, addition of a compound cannot make it better again. Pakorny (1971) also has evidence that phenolic antioxidants offer very little protection if fat or oil metal concentrations are too high. Antioxidants Used in Foods Examples of synthetic antioxidants commonly used in human foods, which comply with safety regulations of the Food and Drug Administration (FDA) of the United States, are illustrated in Table l. l3 QAWNVOH ofimmva ova Ema oom Omma mcocfldoouomnamusm Sunfluuoulocoz IO IO :0 IO mimmvom mimmcom miom-ov.aomafi CNN Ema oom 8mm ocosaoumxouomm webmasbsm «scam mimmvoa miomuo..aomcn omH Ema oom dmm oaomwcmmxouvhm emumasusm nice :o n «has .sammeoo emcee amaummm moma .mcmmfiou xmoox cmEummmm a mimmcm.a mimuc.aomce almNme.o man and com 0m oumaamw ahmoum IO IO 01 :omxomxoo-.o. o .mU00m CH new: mnemowxoflucm oeuonucmm COEEOU HAO ammnsom Hfio useoooo umumz 500v ..H0m w Numaensaom semen: Hmasomaoz icoom mo memes name vases mom aonmxm oEmz £08800 .H magma 14 In the United States, regulations issued by the FDA permit the use of a single antioxidant, or combinations of more than one antioxidant, at concentrations up to 0.02 percent based on the weight of the fat or oil or the fat content of a food product (Code of Federal Regulations, 1974a, 1974b). These levels have been arrived at from tests of both acute and chronic toricity in a variety of animals, (Johnson, 1971). The amount found to cause patho- logical symptoms or death in test animals is, as a general rule, at least 100 times greater than the amount which is permitted in food fats. Most antioxidants, shortly after ingestion, are excreated as glucuronides and/or ethereal sulfate in the urine (Dacre, 1960; Hathway, 1966). Nu- merous workers (Riemenschneider, 1955; Stuckey, 1968; Johnson, 1971) have concluded that at levels of use fixed by present day legislation there is no evidence of any danger of toxicity for humans arising out of the use of antioxidants in foods. At present, synthetic antioxidants commonly used in foods are phenols, either polyphenols such as PG, monoalkyl substituted hydroquinoes such as TBHQ or hindered phenols such as BHA and BHT. The polyphenolic and substituted hydroquinone inhibitors are somewhat soluble in water while the hindered phenols are insoluble in water,as shown in Table 1. Because the hindered phenols are insoluble in water, they are not easily extracted into the water phase 15 during the storage or processing of food fats and are more suitable for the stabilization of water-containing fatty foods (Pakorny, 1971). Evidence presented by several authors indicates that TBHQ usually affords the best pro- tection to vegetable and other polyunsaturated oils (Thompson and Sherwin, 1966; Scherwin and Luckadoo, 1970; Chahine and Macneill, 1974), where as BHA, BHT and PG are effective in hydrogenated vegetable oils and animal fats (Lea, 1958b; Chipault, 1962). Lea (1958b) also states that PG can provide considerably greater protection to dry fats than BHA or BHT but tends to be less stable towards heat and alkali, a prOperty which often makes it less effective in complex foods and in baked and fried products. Antioxidant Effectiveness in Low Moisture Foods A useful measure of stabilization by the use of anti- oxidants has been obtained with a number of dehydrated products, however, there are also reports of added anti- oxidants having been ineffective in preventing oxidation. Reimenschneider (1955) stated that the degree of protection obtainable in complex foods is usually well below that produced by direct addition of the same antioxidants to an extracted fat. Drazga st 21. (1964) found that nitrogen packaging was as good or better than the addition of BHA, BHT, or toc0pherols for potato flakes. Talburt and Smith (1967) 16 stated, after reviewing the use of antioxidants in dehy- drated potato products, that antioxidants are more effec- tive against preventing oxidative odor than against oxi- dative flavor. They felt that the most important needs of the potato granule industry, at that time, was for more improved and less expensive methods for the control of oxidative deterioration in dehydrated potato products. Gooding (1962), in reviewing the stability of dehydrated foods, felt that the only practical method to prevent oxi- dation was nitrogen or vacuum packing. With sweet potato flakes, Deobald and McLemore (1964) found that antioxidants were only effective when the oxygen level was at six and ten percent whereas, below these values, little extra pro- tection was evidenced. Abbot and Waite (1962, 1965) found little protection offered by tocopherols for dry whole milk powder while PG was somewhat effective, BHA did little to improve keeping quality. BHT was shown to be effective in delaying oxidative deterioration in shredded oat cereal by Shmigelsky 22 31. (1964). Radanovics (1969) studied the effect of PG, BHA and BHT in retarding lipid oxidation in freeze-dried sour cream. He found that releasing the vacuum on the freeze- dryer with nitrogen gas after drying, was as effective in retarding lipid oxidation as the use of any antioxidants. Bishov 33 21. (1960) investigated the protective action of BHT in a freeze-dried model system containing, carboxy- methyl cellulose, refined soy oil and soy protein. These 17 workers reported that BHT had only a slight effect on decreasing the rate of fat oxidation. In a study by Karel 22 31. (1966) PG was effective in retarding oxidation in a freeze-dried microcrystalline cellulose-methyl linoleate model system. In a more recent work Labuza gt a1. (1969), using the same microcrystalline cellulose-methyl linoleate freeze-dried model system, confirmed the effectiveness of PG but found that BHT was not effective in controlling oxi- dation in the dried system. These workers postulated that the poor protection by BHT may have been due to its vola- tilization during freeze-drying thereby, reducing its final concentration to an ineffective level in the dry system. In a series of recent studies, lipid oxidative stability of humidified freeze-dried model and food systems was inves- tigated (Labuza gt_gl. 1971; Labuza 25 21. 1972; Chou and Labuza, 1974). In general, their results showed BHA to be an effective antioxidant for delaying lipid oxidation under the experimental conditions used. In many cases, the effectiveness of present antioxi- dants has been measured in foods by the active-oxygen method, peroxide number or the 2-thiobarbituric acid test. These methods present difficulties in interpretation of stability in comparison to an actual shelf life study (Chipault, 1962; Scherwin, 1968). That the methods by which the effectiveness of antioxidant compounds have been proven and tested in foods are not consistent with the 18 present knowledge of oxidative kinetics was also recently pointed out by Labuza (1971). He feels that an oxygen- adsorption measurement procedure gives a direct measure of the extent of rancidity develOpment in foods and should be used more for testing antioxidant effectiveness. In summation, results in dried food systems have shown variable results as to antioxidant effectiveness and in some cases the methodology used for testing effectiveness is subject to question. Since, antioxidants are being continuously proposed for addition to dehydrated foods there is also a need to determine if adequate antioxidant levels remain in these foods after processing to reduce oxidative deterioration. Antioxidant Volatility It has been shown that antioxidants, particularly those with phenolic structures are quite volatile. They are lost rapidly, not only through steam distillation from a frying fat (Stuckey, 1962), but by vaporization at room temperatures (Hannah, 1962; Caldwell 22 31., 1964). Sherwin (1972) states that phenolic antioxidant losses can also occur under processing conditions such as in spray- or freeze-drying of food products. A plot of the vapor pressure of BHA, BHT and TBHQ at various temperatures is shown in Figure l (Sherwin, 1974). 19 I0.0 ’ / I I 5? LOO E E J :=,; g o IOO 0.. Q) '5 8 13 < 0.0IO 0.00| ‘ l ‘ O 50 IOO l50 Temperature, °C Fig. l Vapor pressure of BHA, BHT and TBHQ at various temperatures. 20 It can be seen that, at room temperature and below, the order of volatility ranging from least to the most volatile is TBHQ, BHA and BHT. Retention of Volatile Organic Compounds During Freeze - Drying Freeze-drying is a relatively recent commercial de- hydration process for the preservation of foods. It con- sists of a two-step dehydration process by which the food is first frozen and then moisture is simultaneously removed as ice from the frozen product by sublimation and as water from the concentrated matrix between the ice pores by evap- oration. Freeze-dried foods are generally superior to those dehydrated by other drying methods. This advantage arises largely from the structural rigidity afforded by the frozen material at the surface where sublimation occurs. This rigidity to a large extent prevents collapse of the solid matrix remaining after drying. This results in a porous, non-shrunken structure in the dried product which facilitates retention of volatile flavor and aroma compo- nents and allows rapid and nearly complete rehydration when water is added to the substance at a later time. Much attention has been given to the behavior of volatiles during freeze-drying. Recently, reviews have been made by Flink (1970), King (1970a), and Karel (1973). 21 Basic Observations Among different drying methods, retention of volatiles is considered to be highest during freeze—drying. Lee ggflgl. (1966a, 1966b, 1967), in comparing the retention of volatiles in apricots, peaches and apples dried by "hot air dehydration", ”dehydro-freezing" and "freeze-dehydration”; found the latter method yielded the highest retention. Reineccius and Coulter (1969) found similar results when studying retention of diacetyl added to skim milk during roller-, spray-, and freeze-drying. These conclusions were re-affirmed by the more elaborate studies of Chandrasekaran and King (1971) on liquid-, slush-, and freeze-dried apple juices. Berry and Froschner (1969) stated that distinction has to be made as to the kind of volatile compounds studied. They observed oil soluble compounds to be retained to a higher extent by foam-mat drying than by freeze-drying, whereas the water soluble volatiles were almost entirely lost when foam-mat drying was used. The first quantitative analysis of the retention of volatiles was made by Rey and Bastein (1962). They found, in contrast to the behavior during distillation, a highly volatile compound such as acetone, in an Earle-salt solu- tion, also containing glucose and glycine, to be retained to an important extent. It was found that the retention was influenced by the presence of solids in the mixture. 22 Acetone retention increased with increasing solids con- centration. Saravacos and Moyer (1968) studied the retention of ethyl acetate, ethyl butyrate, methyl anthranilate and acetic acid in freeze-dried model food gels such as pectin, starch, gelatin, and cellulose. They also conducted stud- ies by adding these volatiles to apple slices. These authors found the more volatile the compound, the smaller the retention. They also reported that retention in the gel systems varied according to the type of solid compound and that adding glucose or sucrose to the gel further in- creased retention. The rate of volatile loss was highest during the first period of drying and the fractional loss of ethyl acetate and ethyl butyrate was higher than the fractional loss of water during this drying period. Re- tention increases were found as the drying slab temperature increased. Chandrasekaran and King (1971) found the effects of solids concentration also important in real food systems such as apple juices. They, however, contradict the results of Saravocas and Moyer (1968) when they found the retention of different volatiles to be independent of their relative volatility. Sauvageot g; 31. (1969) investigated the influence of different freeze-drying parameters on the retention of the volatiles of orange and raspberry juices. These real 23 systems gave complex results, which in several cases were conflicting. However, some general trends were observed i.e., more rapid freezing rates and increasing frozen layer thicknesses tended to decrease retention while an increase in solids content increased retention. Due to the complex nature of real food systems most researchers have conducted their investigations using model systems. An extensive study on volatile retention in model systems was accomplished by Flink (1969). His results are presented in a series of papers (Flink and Karel, 1970a, 1970b, 1972; Flink and Labuza, 1972). These authors studied the retention of acetone, methyl acetate, 2-pr0panol, t- butanol, and n-alcohols (C1 to C C7) in solutions con- 5: taining either glucose, maltose, sucrose, lactose or dex- tran-lO as the solids fraction. Retention was found to be independent of the final moisture content. In experiments using 2-pr0panol as the volatile and maltose for the solids compound, it was shown that the volatile was homogeneously retained throughout the sample. Retention decreased with increasing sample thickness for glucose-acetone and dex- tran-lO-Z-prOpanol solutions and increasing freezing rate using n-butanol-maltose solutions. When the dissolved glucose concentration was fixed at ten percent, relative retention of acetone was constant in the low volatile con- centration range while retention decreased in the high con- centration range. A different relationship was found when 24 the amount of volatile retained was calculated on an abso- lute weight basis. Volatile retention was also found to depend on the type of solute used. Within a homologous series of n-alcohols, the lower compounds were retained to a smaller extent than the higher ones. Depending on the kind of solutes, compounds with a long chain-length showed a decrease in retention. The optimum was found to be ethanol in glucose, prOpanol in maltose, and probably hexanol in dextran-lO solutions. Retention increased with increasing platen temperatures with an upper limit being set by the collapse temperature, which was more thoroughly studied by MacKenzie (1965) and Bellows and King (1972, 1973), of the solids present. Bellows and King (1973) have shown a correlation between volatile retention and the ex- tent of sample collapse, using ethyl acetate in sucrose solutions. Another extensive study was carried out by Rulkens and Thijssen (1972). They investigated the retention of acetone, methanol, n-prOpanol, and n-pentanol in solutions of malto-dextrin. When all other variables were kept con- stant, they found retention to increase with increasing solids content up to a maximum above which it remained con- stant. The retention of volatiles increased with inr creasing molecular size of the volatiles, but was inde- pendent of the relative volatility. The volatile retention was found to be independent of final moisture content in 25 the range between one and 15 percent moisture. Decreasing the freezing rate increased volatile retention. A similar result was obtained when the ice front temperature was lowered. Voiley §E_§l. (1973) followed the retention of n- alcohols (C2 to C6), 3-methy1-l-butanol and cycloPentanol in aqueous solutions containing sugars (glucose, sucrose, and fructose), citric acid, calcium chloride, pectins, and albu- mins. Retention was independent of final moisture content, decreased with increasing freezing rate, was directly re- lated to the increasing chain—length of the alcohols and differed according to the type of solids used. In contrast with the findings of Flink and Karel (1969) percent reten— tion increased with increasing initial volatile content. Ofcarcik (1972) and Ofcarcik and Burnes (1974) studied retention of pyruvic acid in model systems containing either glucose, sucrose, or lactose or a combination of them; and the retention of volatile carbonyls, expressed as pyruvate, in onion juices. Retention increased as carbo- hydrate concentration increased, but varied among the carbo- hydrate type in the model solution. However, in onion juices containing different types of sugars, no change in retention was found. Volatile retention decreased with increasing freezing rate and increasing sample thickness. Among other quality parameters Ettrup Peterson g£_gl. (1973) studied the retention of volatiles in freeze-dried 26 coffee. These workers also observed an increase of volatile retention for slowly frozen samples versus rapidly frozen ones. Chamber pressure exerted little influence over the range of 0.3 to 0.7 torr whereas, a large decrease in re- tention occurred at a chamber pressure of 0.8 torr. At this pressure, the ice front temperature had exceeded the col— lapse temperature of the solution. Volatile Retention Mechanisms From the reported results it is clear that volatile retention during freeze-drying is greatly influenced by the presence of solutes in the drying liquid. Although the exact retention mechanism is not yet known three theories have evolved to explain volatiles retention. Adsorption - Rey and Bastien (1962), as mentioned before, found that acetone retention was strongly influenced by the initial glucose content of the solution being freeze- dried, increasing from five percent acetone retention when the initial solution contained five percent glucose to 45 percent retention when the initial solution contained 25 percent glucose. Furthermore, the acetone retention of the specimen was not reduced during more than 30 hr in a vacuum chamber at 40 C and less than 10—3 torr (absolute pressure), subsequent to freeze-drying. This surprisingly good reten- tion was credited, by Ray and Bastien (1962), to an adsorp- tion phenomenon. However, it is not clear how acetone 27 could be held onto the adsorption sites more strongly than water, which is highly polar and therefore strongly ad- sorbed. Issenberg 32 El: (1968) and Boskovic and Issenberg (1969) measured sorption of small quantities of ethanol, hexane, and acetone on cellulose. Although these authors did not correlate their data with possible volatile reten- tion, it is clear from their sorption isotherms that only small quantities of volatiles can be adsorbed. The adsorption theory has also been discredited by Flink and Karel (1970b). In one experiment, maltose so- lutions were frozen in layers, some of which were volatile- containing others were volatile-free. After freeze-drying, the layers were separated and analyzed separately for the volatile (2-pr0panol). They found that, after freeze- drying, retention occurred only in those layers which initially contained the volatile. In another experiment these authors (1969) found that small amounts of 2-pr0panol present in maltose solutions before freeze-drying are held much more tenaciously after drying than is 2-propanol that is adsorbed onto maltose after freeze-drying. Recently Chirife gp‘al. (1973) and Chirife and Karel (1973a) reported research on the retention of n-propanol in freeze-dried polyvinylpyrrolidone (PVP) solutions. Chirife and Karel (1973a) presented evidence that a small but signi- ficant amount of the retained alcohol was due to surface 28 adsorption, but that most of the retained alcohol in the PVP system was held by microregion entrapment. The adsorbed fraction of alcohol increased as the PVP system freezing rate was increased. Selective Volatile Retention - Saravocos and Moyer (1968) stated that the volatile retention was caused by "locking in" the volatiles by sugars. From experimental data presented by Flink and Karel (1969), Rulkens and Thijssen (1972) and Voiley gE_al. (1973) it can be seen that volatiles are lost together with the water vapor during the first period of drying. As shown by Saravacos and Moyer (1968) some volatiles are lost to a large extent. At a certain critical moisture content, volatile loss ceases even as the moisture content continues to decrease. Two main mechanisms have been prOposed to describe the phenomenon. Selective Diffusion - When aqueous solutions are air dried, a film is formed at the surface of these solutions. This film has been shown to be impermeable to volatiles (Menting and Hoogstad, 1967a; 1967b). The permeability of this membrane was shown to be controlled by its moisture content. This behavior has been explained by Menting and Hoogstad (1967a; 1967b) and Thijssen and Rulkens (1968) by a change in diffusion coefficients. By measuring the diffusion coefficients of acetone and of water in malto- dextrin at different moisture contents they were able to 29 solve both water and volatile flux equations and thereby predict volatile losses (Thijssen and Rulkens, 1968; 1969). Their findings demonstrate that below a critical water con- centration the system becomes selectively permeable to water and completely impermeable to volatile molecules. The critical moisture content is dependent on the nature of the dissolved solids, and increases with increasing size of the volatile molecule and decreasing temperature (Thijssen, 1971). These findings have been applied to volatile retention during freeze-drying by Thijssen and Rulkens (1968, 1969) and Thijssen (1971). In their analysis the authors as- sumed the ice and the solutes layers to be clearly sepa- rated within the frozen product, the ice being present in the form of equally spaced, circular, tubular crystals of uniform diameter, leaving after sublimation a honeycomb of pores within a matrix of eutectic solids or concentrated solutes. The volatiles were assumed to be initially homo- geneously distributed within the solute phase. It was as- sumed that the temperature at the sublimation front and the drying rate does not change very much during freeze-drying. As a consequence, the degree of volatile loss should be independent of the location in the sample. This assumption has been experimentally proven correct by Flink and Karel (1970b) in their study of 2-propanol retention in sectioned freeze-dried maltose solutions. 30 The loss of water and of volatiles out of this matrix is then treated as a diffusion process. As the sublimations front retreats, the residual water will diffuse from the concentrated solute matrix into the gas pores. When sec- ondary drying proceeds, the moisture content decreases with increasing distance from the ice front. Since, ice crystals are impermeable to volatiles, except for a thin layer just under the ice front, volatiles can only be lost from a distinct zone above the sublimation front. This zone will follow the ice front as it proceeds through the sample. Because of the low temperature of the remaining matrix near the sublimation front, and the low water concentration the diffusion coefficient of the volatile molecules in the ma- trix is much smaller than that of water. This means that the volatile escapes more slowly from the drying matrix than does the water and is no longer lost after a critical moisture value is reached. From this theory, Thijssen and Rulkens (1969) predicted that volatile retention should increase with increasing the thickness of the matrix i.e., slower freezing rates and with a decrease in water concentration in the matrix i.e., higher solids content. These predictions were confirmed by the data of Rulkens and Thijssen (1972). Following and extending somewhat the approach taken by Thijssen and Rulkens (1968, 1969) King (1970a, 1970b) used the diffusion concept to predict qualitative effects of 31 processing parameters upon volatile retention in the freeze- drying of homogenous solutions. King (1970b) states that the extent of volatile loss should decrease with decreasing values of the dimensionless Fourier group Dt/Lz, where D is the volatile effective diffusivity, this is the time during which diffusion can occur, and L is the thickness of the region from which volatile loss occurs. King (1970b) and Thijssen (1971) present evidence to show that this analysis of the selective diffusion concept correctly explains the experimentally observed beneficial effects upon volatile retention caused by high initial dissolved solids content, slower rates of freezing, and smaller piece sizes (Sauvageot g£_gl. 1969; Flink and Karel, 1970a; Rulkens and Thijssen, 1972). Chandrasekaran and King (1972) worked with a ternary diffusion model, considering water, dissolved solids and the volatile species as separate components to make a quantitative prediction of retention and concentration pro- files of aroma compounds within the medium being freeze- dried. In general their results support qualitative con- clusions for volatile retention inferred by Thijssen and Rulkens (1969). Microregion Entrapment - Based on measurements of volatile retention in carbohydrate solutions Flink and Karel (1969, 1970b, 1972) presented a theory for volatile 32 retention during freeze-drying based on the formation during freezing and drying of volatile containing micro- regions in the dried cake. They postulate that the volatile components are trapped into microregions of carbohydrates at sufficiently low water contents. In these microregions the carbohydrate orients during freezing and freeze-drying in such a way as to surround the volatile molecules and prevent their escape. This carbohydrate structure for en- trapping volatile molecules is due to an inter-hydrogen bonding which increases as water is removed and at a certain critical low moisture content the amount of these bonds will be at a maximum and volatile loss will be stOpped. Thus, volatile retention will depend on the degree of inter- carbohydrate association and on the degree and perfection of the matrix structure formed. The possibility of volatile entrapment in frozen aqueous solutions has been demonstrated by Flink and Gejl-Hansen (1972) and by Lambert 93.3l. (1973a, 1973b). Although the exact mechanism is not elucidated, it was shown that volatile retention depends on freezing rate, equilibrium freezing temperature and the place where the volatiles are immobilized during freezing. Evidence to show that volatiles are entrapped in micro- regions at the place they were initially present has been presented by Flink and Karel (1970b) with their layered system experiments described above. These authors in 33 another experiment using the same 2-pr0panol maltose system showed that the size of the microregion within the dry amor- phous matrix is very small since grinding and evacuation of the freeze-dried material did not release any additional alcohol. Flink and Karel (1972) demonstrated the importance of the hydrOgen bonded amorphous matrix for maintaining vola- tiles in the freeze-dried cake. Freeze-dried samples were placed in an air atmosphere at constant temperature and constant relative humidity, in order to determine the effect or re-humidification on volatile retention. Volatiles were lost, depending on the extent of hydrOgen bond disruption. At each new relative humidity, volatiles were lost down to a new, lower retention level. When the samples reached higher moisture contents complete volatile loss was noted and accompanied by a visible structural change in the dry cake. They attributed the volatile losses upon humidifi- cation to be a reflection of the action of sorbed water entering the carbohydrate matrix and disrupting the hydro- gen bonds holding the carbohydrate molecules into the cage structure which had trapped the volatile molecules. To support this explanation Flink and Karel (1972) showed that additions of 2-pr0panol to dextran-lo gave the same shape isotherm as dextran-lO. From this experiment they concluded that volatiles do not show any competition 34 with water vapor for the sorption sites. Instead, volatiles may have a different mode of interaction with the carbo- hydrate. Recently Chirife §£_ai. (1973) and Chirife and Karel (1973a, 1973b, 1974), who studied volatile retention in different model systems, presented evidence to support the microregion theory. Retention of n-propanol in PVP (Chirife $5.31. 1973; Chirife and Karel, 1973a), in cellu- lose and starch suspensions (Chirife and Karel, 1973b), in protein solutions (Chirife and Karel, 1974) were studied. Special attention was given to the microregion disruption mechanism. All the dry cakes were sensitive to water sorp- tion. As was the case with carbohydrates at a specific relative humidity, volatiles were lost, down to a new lower retention level in all systems. In this series of experi- ments, it was found that the maximum amount of volatile retained depended strongly on the type of solute, cellulose having the lowest retention capacity. The low volatile retention in cellulose suspensions was explained on the basis of the low mobility of the cellulose chains as com- pared with other polymeric materials (Chirife and Karel, 1973b). The influence of processing parameters on the reten- tion of organic volatiles in freeze-dried carbohydrate solutions was studied by Flink and Karel (1970b). Effect of initial dissolved solids content, freezing rate, drying 35 rate, and sample thickness on volatile retention was ex- plained on the basis of the way each variable changed microregion structural integrity. The interpretation, based on microregion entrapment of volatiles predicts the same outcome for these processing variables as does the selective diffusion mechanism. In fact, as pointed out by King (1970a), the selective diffusion and the microregion entrapment mechanism for the retention of organic volatiles are quite similar, and are probably describing the same basic phenomena from two different approaches, namely mathematical or macrosc0pic vs morphological or micro- sc0pic viewpoints. Droplets Formation - Massaldi and King (1974a) have suggested that during freezing of aqueous solutions con- taining organic volatiles, not all of these volatiles will be homogenously distributed throughout the solute matrix. Consequently, during freeze-drying, the retention charac- teristics of the volatiles will vary with its location after freezing. Some support for the theory that volatiles are heter- ogenously distributed throughout the solute matrix after freezing is given by the work of Flink g£_§l, (1973). These workers studied the retention behavior of hexanol and butanol in aqueous malto-dextrin solutions on a freeze- drying microsc0pe stage. They observed movement of drop- lets of hexanol and butanol, relative to their initial 36 location, into the solute matrix. Flink and Gejl-Hansen (1972), who made microsc0pic observations of aqueous malto- dextrin solutions containing various volatiles, correlated drOplet formation with the solubility of the volatile in water. They found that highly soluble compounds (ethanol, prOpanol, and acetone) did not form drOplets, whereas less soluble ones (butanol, diacetyl, isovaleraldehyde, and hexanol) did, methyl ethyl ketone was an exception. To determine and interpret the retention behavior of volatile organic substances which have limited solubility and form drOplets in aqueous solutions Massaldi and King (1974a) studied the retention of d-limonene in freeze-dried sucrose solutions. These authors predict three locations for the volatiles in the frozen slab: l) drOplets of vola- tiles adjacent to the ice-crystal interface; 2) droplets of volatiles in the solids region; and 3)homogenously dis- solved volatiles in the solids region. Based on this model of volatile location they then predict the retention char- acteristics of volatiles by their location as: 1) com- pletely lost through vaporization after the sublimation of ice; 2) lost only through a relatively slow diffusion pro- cess; and 3) lost in the same way as substances below their solubility limit i.e., substances not forming dr0p1ets. The overall retention of volatile is predicted to be the summation of the different retention characteristics by the authors. In the same work this model was applied to their 37 data of d-limonene retention. They found the percent d- 1imonene retention to be constant until its initial con- centration exceeded its solubility limit in the aqueous sucrose solution (corresponding to drOplet formation). Above this level, the percent d-limonene retention de- creased and corresponded with the retention predicted by their model. The loss mechanism depicted by Massaldi and King (1974a) may also be an explanation for the lower retentions found for less soluble components in the studies of freeze- drying model solutions reported by Flink and Karel (1970b) and Chirife SE 3;. (1973). While the latter authors ex- plained this behavior by accepting the microregions to be oversaturated at high concentration levels of the volatile, the former authors attribute the volatile loss to the formation of droplets in the emulsion. Massaldi and King (1974b) also studied d-limonene re- tention in freeze-dried orange juices and found, with some modifications, retention behavior to support their model developed for synthetic emulsions. EXPERIMENTAL PROCEDURES Preparation of Model System Ingredients Defatting of Wheat Starch Wheat starch, used in the model system, was obtained from Far. Mar. Co., Inc., Hutchinson, Kansas. The starch was defatted according to the method of Schoch (1964) in a Soxhlet extractor with 95 percent ethanol for 36 hr. Following extraction the alcohol-wet cake was transferred to a beaker of distilled water, mixed with a magnetic stirr for 15 min, then suction-filtered, washed with distilled water, and finally dried overnight at 40°C. The fat con- tent of the starch was analyzed before and after alcohol extraction by a procedure reported by Schoch (1964) using 1009 starch samples for each analysis. Preparation of Pregelatinized Wheat Starch The defatted wheat starch was gelatinized, on a batch process, by weighing 70g of starch (as is moisture basis) into a 1000 m1 Erlenmeyer flask containing 375 ml of dis- tilled water. The flask was covered with a rubber stOpper and heated, with continuous agitation on a magnetic stirr, in a 601°C water bath for 30 min. The gelatinized starch was poured into an 11-1/4 x 7-1/2 x l-l/2 in aluminum pan 38 39 and frozen at -20°C. Pans containing the frozen starch slurry were transferred to a Virtis Repp model 42 freeze dryer and freeze-dried for 20 hr at a platen temperature of 52°C and a chamber pressure below 5 x 10.3 torr. After freeze-drying the gelatinized starch was allowed to stand eight to 12 hr at room temperature, to attain moisture equilibrium. The dried gelatinized starch was then sized by grinding first on a Waring blender followed by grinding in a Wiley mill equipped with a stainless steel U.S. no. 60 sieve. The sized, defatted, pre-gelatinized wheat starch was then analyzed for moisture content by the Karl Fisher method (Smith, 1964) and stored at room temperature in sealed brown glass jars until used as a model system ingredient. Storage of Sgybean and Coconut Oil Refined soybean oil and coconut oil were obtained from PVO International, Inc., Boonton, New Jersey. The lipid samples were analyzed for refractive index, iodine number, peroxide value and fatty acid composition with an Abbe refractometer (AOCS, 1974a), by the Hanus method (AOAC, 1975a), by the American Oil Chemists' Society technique (AOCS, 1974b), and by gas chromatography (see analytical procedures), respectively. Following analysis samples were placed in 4 oz clear glass bottles. The filled bottles were evacuated in a vacuum chamber at 10.7 torr with a 40 water aspirator and nitrogen gas was introduced into the vacuum chamber until the inside pressure reached equi- librium with the outside atmosphere. This treatment was repeated two times. After the second treatment the vacuum chamber was Opened, the bottles were sealed and placed in vacuum desiccators. These desiccators were then evacuated with a water aspirator, gassed with nitrogen, and stored in a dark place at -20°C until antioxidants were added to the lipid samples. The storage conditions were designed to minimize lipid oxidation thereby, maintaining oil quality throughout the study. Preparation of Antioxidant Spiked Sgybean and Coconut Oil Soybean oil samples were removed from storage approxi- mately 12 hr before addition of antioxidants. Antioxidant spiked soybean oils were prepared, just prior to their incorporation into model systems, by weighing an appro- priate amount of BHA, BHT or TBHQ (Eastman Chemical Products, Inc., Kingsport, Tennessee) and soybean oil into glass jars to give an antioxidant concentration of 200 or 400 pg of antioxidant per g of soybean oil. The jars con- taining the antioxidant-soybean oil mixtures were blanketed with nitrogen gas and sealed. The mixtures were then agi— tated with a magnetic stirrer at room temperature for about one hr to insure uniform distribution and true solution. 41 Coconut oil spiked with BHA, BHT, or TBHQ was pre- pared by the same method used for soybean oil-antioxidant mixtures, except that a temperature of 30°C was employed to maintain the coconut oil in a liquid state. Preparation and Freeze-Drying of Model Systems Model System Composition The model system used to study phenolic antioxidant retention consisted of 15 percent pregelatinized wheat starch, five percent lipid material (antioxidant spiked soybean or coconut oil) and 80 percent distilled and de- ionized water on a dry weight basis. Model System Preparation Model systems were prepared by incorporating the de- sired amount of pregelatinized wheat starch into water (30°C) and then adding the correct amount of antioxidant Spiked soybean or coconut oil being studied. One min was allowed for blending each component in a Waring blender. Emulsions were then homogenized on a Logeman laboratory homogenizer model C-8 (Chase-Logeman Corp., Brooklyn, New York) at a pressure of 2500 psi. Normally 95g portions of the homogenized emulsion were placed into two stainless steel pans measuring ten by ten by four cm and subjected 42 to a slow or a rapid freezing treatment; exceptions to this are noted at apprOpriate sections in the Results and Dis- cussion. In some cases thermocouples (iron-constantan), used for measuring sample temperature during freezing and freeze-drying, were positioned at different levels above the bottom of the pan before the addition of the sample. Freezing of Model Systems Samples subjected to slow freezing were frozen in the freeze-dryer on a platen at -22°C. The samples were held under these conditions for three hr before freeze-drying. Rapid freezing was accomplished by immersing the sample in a liquid nitrogen bath (-l96°C) for five min. The frozen samples were then transferred to the freeze-dryer and placed on a -22°C platen for one hr, before freeze-drying. Freeze-Drying of Model Systems Each freeze-drying run was conducted, with two samples at the same left and right position, on the tOp platen in a Virtis Repp model 42 freeze-dryer. The freeze-dryer was equipped with automatic controls for constant recording of absolute pressure and thermocouple connections connected to a Honeywell Electronic 15 multipoint strip chart recorder for temperature measurements. The drying took place with the heating platen set for 52°C and the condenser tempera- ture set at -60°C. Sample drying time was 12.5 hr and 43 26.5 hr for slow and rapid freezing, respectively. The 3 torr (McLeod chamber pressure registered below 5 x 10— Gauge) for most of the drying cycle. In some experiments, the effect of extending the drying time of slow frozen samples was studied, in these cases the altered drying times are noted at appr0priate sections in the Results and Discussion. In order to minimize lipid oxidation, leading also to antioxidant degradation in the dried systems, the chamber vacuum was always released with nitrogen gas. After freeze-drying the dried system was transferred to a one pint clear glass jar, designed to fit an Osterizer blender, and pulverized. The pulverized powder was passed through a U.S. no. 12 stainless steel sieve, placed in a glass jar, sealed, and immediately analyzed for antioxidant concentration (see analytical procedures). The sized, freeze-dried system was also analyzed for moisture content by a vacuum oven method (AOAC, 1975b) and for fat content by 16 hr ether extraction on a Soxhlet extractor (AOAC, 1975c). Experimental Variables The effect of antioxidant level, freezing rate, and oil type on antioxidant retention in freeze-dried model systems was studied. The design of the experiments used to 44 study these variables, consisted of three-three by two factorial experiments as described below: Antioxidant Level Three antioxidants (BHA, BHT and TBHQ) were studied at two concentration levels (200 and 400 ug per g soybean oil) in the model system. Freezing Rate Three antioxidants (BHA, BHT and TBHQ) using two freezing rates (rapid and slow) were studied in the model system using 400 ug of antioxidant per g of soybean oil. Oil Type Three antioxidants (BHA, BHT and TBHQ) were studied at one concentration level (400 ug Per 9 of oil) using soybean or coconut oil as the lipid portion of the model system. In all experiments at least two freeze-drying runs of each treatment were made. Each freeze-drying run consisted of drying two samples of a single treatment. 45 ANALYTICAL PROCEDURES Determination of Isomeric Purity of BHA The isomeric purity of BHA was determined by modifying the method of Mahn pp pl. (1970) as described below. Standard Preparation About 400 mg 3,S-di-tert-butyl-4-hydroxyanisole (diBHA) (Universal Oil Products Co., East Rutherford, New Jersey), used as an internal standard, was accurately weighed into a 200 m1 volumetric flask. The internal stan- dard was dissolved and diluted to volume with carbon disul- fide. A working standard solution was prepared by accurately weighing about 90 mg of reference 3-tert-butyl-4-hydroxy- anisole and ten mg of reference 2-tert-butyl-4-hydroxy- anisole (Food Chemicals Codex) into a 25 ml volumetric flask and then dissolving and diluting to volume with the internal standard solution. Sample Preparation About 200 mg of BHA (Eastman Chemical Products, Inc.) was accurately weighed into a 50 ml volumetric flask. The sample was dissolved and diluted to volume with internal standard solution. 46 The working standard solution and sample solution mixture were chromatographed in triplicate as described under Gas Chromatography. Gas Chromatography Gas chromatographic analyses were performed using a F & M 810 chromatograph, equipped with a flame ionization detector. A 1/8 in o.d., five ft stainless steel column was packed with ten percent XE-60 on acid washed, 80 - 100 mesh Chromosorb W. The column oven temperature was 155°C, the injection port was maintained at 225°C, and the detec- tor at 250°C. The nitrogen carrier gas flow was 30 ml per min. Flow rates of hydroqen and air were 30 ml per min and 300 ml per min, respectively. The chromatograph sensi- tivity was set at 4 x 10-11 amps per mV for the 2-isomer and 32 x 10-11 amps per mV for the 3-isomer and diBHA. Samples of one to two ul were injected, and the emerging peak areas were calculated by multiplying peak height times peak width at half-height. Calculations A detector reaponse (K) value was determined for each isomer by chromatographing known concentrations of the iso- mers with internal standard (diBHA) and equation (1). 47 Equation (1): K = total_peak area for isomeereak area of internal standard weight of isomer/weight of internal standard Peak areas observed for the isomers of BHA were then used to determine the percent isomer (W/W) by using equa- tion (2) and established K values. Equation (2): % Isomer (W/W) = (total peak area for isomer)(wgt of internal standard)5000 (peak area of internal standard)(K value) mg sample Determination of Fatty Acid Composition of Soybean and Coconut Oil Methyl Ester Preparation About ten mg of soybean or coconut oil was weighed into a ten by 125mm centrifuge tube provided with a Teflon- 1ined screw cap. The one ml of boron fluoride-methanol (1409 BF per liter of methanol): benzene: methanol (25:20: 3 50 V/V) was added under nitrogen to the lipid. The tubes were sealed and heated for 30 min at 100°C, as suggested by Morrison and Smith (1964). After cooling the samples to room temperature, two ml of pentane and one ml of water were added to each tube. The contents were mixed and the samples were centrifuged until both layers were clear. The upper pentane layer was removed and used for gas chroma- tographic analysis of fatty acid methyl esters. 48 Gas Chromatggraphy Gas chromatographic analyses were performed using a Beckman GC-S dual column chromatograph, equipped with flame ionization detectors and stainless steel columns (six ft x 1/8 in o.d.) packed with 15 percent high efficiency EDGS on acid washed, 80 - 100 mesh Chromosorb W. The column oven temperature was held isothermally at 180°C or pro- grammed from 80°C to 180°C at 11° per min for soybean or coconut oil, respectively. The nitrogen flow rate through the columns was 23 ml per min. Flow rates of hydrogen and air were 18 ml per min and 100 ml per min, respectively. Samples of two to four ul were injected, and the emerging peaks were identified by comparing retention times to those of standard mixtures of known fatty acid methyl esters. Calculations Peak areas were calculated by multiplying peak height times peak width at half-height. Weight percentage com- positions were calculated by applying a correction factor of unity over the range C8 - C20. 49 Differential Thermal Analysis of Model Systems Differential thermal analysis (DTA) was employed for studying the thermal behavior of the mode systems. The instrument used for this investigation was the DuPont 900 Differential Thermal Analyzer. Model systems were slurried with glass beads and approximately 30 mg of sample was placed in a four by 25 mm glass tube. Glass beads were placed in a similar tube to a depth of seven mm which served as the reference standard. Thermocouples were placed into the center of the tubes and the samples were inserted into position. The T scale was set at 20° C per in and the AT scale at 10°C per in. The systems were fro- zen and lowered to approximately —60°C by means of the liquid nitrogen cooling unit. The heating rate was 16°C per min. Peroxide Value Determination on Freeze-Dried Model System The fat was directly extracted from freeze-dried model systems so that peroxide value determination (AOCS, 1974b) could be made. Fat extractions were accomplished by accu- rately weighing 209 of the sized freeze-dried model system into a 250 m1 Fleaker (Pyrex brand - Reg TM Corning Glass 50 Works) and adding 100 ml of pentane (see Solvent Purifi— cation and Testing). The Fleaker was cbvered with a watch glass and the contents were mixed with a magnetic stirrer for one hr. After mixing, the sample was filtered under vacuum through a 150 ml Buchner type funnel, with a fine fritted disc, and collected in a 250 ml Erlenmeyer flask fitted with a standard taper stOpper (ST 27). The Fleaker was rinsed with two 30 ml portions of pentane and each of which was added to the filter cake with the vacuum off. Each rinsing was allowed to pass into the filter cake be- fore reapplying the vacuum. After filtration the flask was transferred to a rotary evaporator and the pentane was re- moved under vacuum (water aspirator). During evaporation the flask was partially immersed in a water bath maintained at 24i2°C. Evaporation was continued to an oily residue and then the vacuum on the evaporator was released with nitrogen gas. The oil in the flask was used for the per- oxide value determination. A portion of the freeze-dried sample taken for per- oxide value determination was also analyzed for moisture (AOAC, 1975b) and fat (AOAC, 1975c) content so that the amount of fat used in the peroxide value determination could be calculated. Quantitation of Antioxidants in Model Systems Solvent Purification and Testing Technical grade pentane (Fisher Scientific Company) was re-distilled in an all glass apparatus prior to use. Pentane, hexane and ethyl ether (all obtained from Burdick and Jackson Laboratories, Inc., Muskegon, Michigan) were routinely tested for the presence of peroxides with vanadium-sulfuric acid (Appendix A). Chromatggraphic Apparatus A schematic of the high-resolution liquid chromatog- raphy (LC) system used is shown in Figure 2. The unit consisted of a C903 liquid chromatograph pumping system (Waters Associates,Inc., Milford, Massachusetts), capable of a maximum flow rate of 2.7 ml per min and pressure capa- bility up to 3000 psi. The solvent reservoir (1000 ml capacity) was placed about seven in above the pump since a slightly positive pressure was beneficial for operation. The solvent was degassed by slowly stirring with a magnetic stirr and was heated externally, slightly above room temper- ature. Downstream from the pump, all components in the system were connected with 0.009 in, i.d., x 0.0625 in, o.d. stainless steel tubing to insure a minimal dead volume in the system. 51 52 [Jr—- To Waste "—_'1' Solvent Reservoir Monitor ll 1 ‘_ """J 17 Recorder Cl Magnetic Stirrer Temperature Control Block 8 Columns LL Pump 1 Q—g Injection Part Six Port Valve Loop Fig. 2 Schematic of liquid chromatographic system. A six-port, rotary valve (Waters Associates, Inc.), positioned just before the injection port, was used to deliver a flushing solvent through the system to wash oil off the column after each injection. The loop connected to the valve ports was constructed of 1/8 in, o.d., stain- less steel tubing and had a volume of about 20 ml. Injections were made on-column through a septum in- jector (Waters Associates, Inc.) fitted with WSR septum discs (Waters Associates, Inc.). A Hamilton 702N, 25 ul syringe (Hamilton Co., Whittier, Calif.) was used for making injections. The two 1/8 in, o.d. x two ft stainless steel columns used in the system were fitted into an aluminum temperature control block (waters Associates, Inc.) which was connected to.a water bath equipped with a circulating pump. 53 The detector employed in the system was a dual beam UV Monitor model 1200 (Laboratory Data Control, Riviera Beach, Florida) with a one cm pathlength flowcells of eight ul holdup volume and a maximum sensitivity of 0.01 absorbence units (full scale) measured at 280 nm. The reference cell was filled with mobile phase. The detector output was monitored by a Linear Instruments, Corp. (Costa Mesa, California) Model 281 ten mV recorder. Chromatographic Conditions Antioxidants (BHA, BHT, or TBHQ) and internal standard (diBHA) were separated from oil extracts, without prior steps of sample preparation, by reversed phase liquid/solid LC as outlined by Dark (1973). In this method two columns (1/8 in o.d. x two ft) packed with BondapakTM C18/Corasil, obtained from Waters Associates, Inc., were the stationary phase. The mobile phase, consisting of distilled water : acetonitrile (Burdick and Jackson Laboratories, Inc.) : n-butanol (Burdick and Jackson Laboratories, Inc.) (5:3:2 V/V), had a flow rate of 0.6 ml per min. The appropriate size sample was injected (Table 2) in duplicate and the detector sensitivity was adjusted to keep all peaks on scale. After the antioxidant-internal standard separation was made, about 20 m1 n-butanol, introduced into the system through the valve-100p accessory, was used to wash the oil off the column. Another injection was not made until the column 54 had again established equilibrium with the mobile phase. The column was operated at ambient temperature when soybean oil samples were analyzed and at 30°C for coconut oil samples. Peak height was used as a basis for quantitation. The emerging peaks were identified by comparing re- tention times to those of standard antioxidant-internal standard mixtures and by comparing the UV absorption spec- trum of the column effluent, corresponding to a specific peak, to the spectrum of antioxidant standards. Extraction of BHA, BHT and TBHQ From Dry Model Systems BHA 5 BHT - BHA or BHT was extracted from sized, well- mixed, freeze-dried model systems with pentane. About five 9 (weighed to the nearest mg) of the dry system was weighed into a 150 ml beaker and 60 m1 pentane was added. The beaker was covered with a watch glass and mixed with a magnetic stirrer for 15 min. After mixing the walls of the beaker were washed with a few ml of pentane, and the beaker was covered with a watch glass and allowed to stand over- night in the dark at room temperature. The sample was brought back to its initial volume with pentane, mixed with a magnetic stirrer for 15 min, suction-filtered through a 150 ml Buchner type funnel, with a fine fritted disc and collected in a 100 ml boiling flask equipped with a 24/40 standard taper joint. The beaker was rinsed with two 20 ml 55 portions of pentane and each time added to the filter cake with the vacuum off. Each rinsing was allowed to pass into the filter cake before reapplying the vacuum. The appro- priate amount of diBHA (Table 2) was added to the flask and pentane was removed under vacuum (water aspirator) on a rotary evaporator. During evaporation the flask was partially immersed in a water bath maintained at 24i2°C. Evaporation was continued to an oily residue. The oil was removed from the flask with a Pasteur disposable pipet, filtered through a 5.5 cm Whatman no. one filter paper, and collected in a glass-stoppered one m1 volumetric tube. The recovered oil samples were analyzed for BHA or BHT content by LC as described under chromatographic conditions. Table 2 Quantity of diBHA added to antioxidant extracts of freeze-dried model systems and sample size used for LC analysis of antioxidants. diBHA Solutiona Sample Size Antioxidant Conc. (Hg/m1) ml Added Injected (ALL. BHA 100 2 20 BHT 50 l 20 TBHQ 100 3 15 a . Hexane solution TBHQ - TBHQ was extracted from sized, well-mixed freeze-dried model systems by exactly the same procedure 56 used for BHA and BHT except ethyl ether was used in place of pentane. anntitation of BHA, BHT and TBHggin Dry Model Systems An internal standard method was used to quantitate antioxidants. DiBHA was chosen as the internal standard because of its structural similarity to the antioxidants being quantitated. Calibration curves were obtained for each antioxidant (BHA, BHT and TBHQ) in model systems con- taining soybean and coconut oil by adding varying amounts of antioxidants (Table 3) to blank model systems and subjecting them to an extraction and chromatographic treatment iden- tical to that used for the unknown samples. Calibrations graphs were prepared by plotting the peak height ratio for each antioxidant to that of diBHA against the ratio of the weight of the antioxidant to that of diBHA. To determine the concentration of an antioxidant in an unknown sample, the peak height ratio was measured and from the calibration graph the weight ratio of the antioxidant to diBHA was read. Since the amount of diBHA added, the sample weight, and the percent fat in the sample were known, the concentration of antioxidant could be calculated using equation (3). Equation (3) ug Antioxidant 9 oil (wt. ratio fromggrgph) (ug diBHA added) (% fat in moael system)(sample wt. 9) .oumocmum Hmcuouch 57 mm.o com com oo.o com omH om.o com oma mm.o com ooa useoooo came mo.o com oca om.o com omH mm.o com ooH oa.o com on emmasom came ov.~ om owe oo.~ om cod ov.a cm on oo.H om om useoooo amm oo.¢ om cow oo.m om oma oo.~ om ooH oo.H om om cmmnsom 9mm oo.H com com me.o oo~ oma ascoooo om.o cow ooa no mm.o ecu om cmmnsom «mm dmmflo\uamowx0wusm Amnvunmflo3ammmwo emnvunmaoz Eoumww H6602 usacfionusd ofiumm erases unmenxoeuea as ago no «use .mnmaum soMumnnHHmo anemone on can: munowmz ouaosmum Haauoucw can usaofixoaund .m manna 58 Recovery of Added Antioxidants From Freeze-Dried Model Systems Recoveries of antioxidants (BHA, BHT or TBHQ) added to freeze-dried model systems containing soybean and coconut oil were determined at the 100 ug antioxidant per 9 oil level. Recoveries were determined by subjecting blank and antioxidant spiked systems to an extraction and chroma- tographic treatment identical to that used for unknown samples except the the blank system oily residue was spiked with 100 ug of the appropriate antioxidant prior to LC injection. The samples spiked with antioxidant prior to LC injection were assigned a recovery value of 100 percent. Statistical Interpretation The degree of antioxidant and treatment differences were ascertained by the analysis of variance (Snedecor, 1956). A fixed effects model was used for calculating F- ratios. Differences between treatment means were tested for significance utilizing Dunnett's procedure (Kirk, 1968). The relationship of some variables was studied utilizing linear correlation (Snedecor, 1956). 59 RESULTS AND DISCUSSION Analyses of Model System Components Soybean and Coconut Oil The iodine number, refractive index, and fatty acid composition found for soybean and coconut oil are recorded in Table 4. These results are similar to typical values given by Swern (1964) for these oils. The degree of unsatu- ration between these oils is indicated by the measured iodine number and fatty acid composition. Coconut oil, a lauric acid group oil, which consists of about 93 percent saturated fatty acids had an iodine number of 8.10. Soy- bean oil, a.member of the linolenic acid group, is made up predominately of unsaturated fatty acids and had an iodine number of 126.14. These oils were selected because of their dissimilar fatty acid composition to determine if the degree of unsaturation in an oil influenced antioxidant retention during freeze-drying. The adequacy of the storage conditions used to hold the oils until used in the research trials is supported by the insignificant change in peroxide value of oils stored for one year (Table 4). 60 Table 4. Characteristics and fatty acid composition of soybean and coconut oil. Analysis Coconut Oil Soybean Oil Characteristic Iodine Number (Hanus) 8.10 126.14 Refractive Index at 40°C a 1.449 1.467 Peroxide Value (No storage) a 0.10 0.72 (1 yr storage) 0.15 0.95 Fatty Acids, wt. % Common Name Common GC Designation Caprylic 8:0 9.0 - Capric 10:0 6.9 - Lauric 12:0 47.6 - Myristic 14:0 18.1 - Palmitic 16:0 8.8 12.0 Stearic 18:0 2.5 4.0 Total Saturated 92.9 16.0 Oleic 18:1 5.6 24.4 Linoleic 18:2 - 51.4 Linolenic 18:3 1.5 8.2 Total Unsaturated 7.1 84.0 aSee Experimental Procedures for storage conditions. 61 BHA Isomeric Purity Takahashi (1967) and Mahn, gt 31. (1970) reported that commercial BHA usually consists of about 90-95 percent 3-isomer and 5-10 percent 2-isomer. Gas chromatographic analyses, indicate that the BHA (Eastman Chemicals Products, Inc.) used in this study contained 90.5 percent 3-isomer and 8.1 percent 2-isomer. This so-called isomeric purity of the BHA used is within the expected range of values for a commercial product and conforms with the WHO definition (WHO, 1962), which states that commercial BHA should con- tain not less than 98.5 percent material with the empirical formula of C11Hl602° Determination of Pat in Wheat Starch All cereal starches are known to contain small amounts of lipids. In addition, industrial starch products some- times contain salt, boric acid, sulfonated oil, etc. (Schoch, 1964). Since the wheat starch used for preparing the model systems was an industrial product, a preliminary experiment was made to determine if such adjuncts were removed from the freeze-dried starch emulsion by the antioxidant ex- traction procedure. High-resolution liquid chromatography (LC) of the extract revealed that indeed a small amount of extraneous material, having a retention time similar to 62 that of BHA, was extracted from the wheat starch. However, it was subsequently observed that an ethanolic extraction, used to remove fatty material from starches, removed these extraneous substances, which interferred with BHA analyses, from the wheat starch. No further attempt was made to identify the interferring substance(s). The percent fat found in the parent and the alcohol extracted wheat starch was 0.4 and 0.1 percent,respectively. These values are typical of those reported in the litera- ture for wheat starch (Schoch, 1942, 1964). Extraction and Determination of Antioxidants Before investigating the parameters controlling the retention of antioxidants during freeze-drying, it was necessary to have a reliable extraction and analytical procedure for the determination of the antioxidants used in this study. Reliability of LC System The resolving power of the LC system was determined by chromatographying samples of soybean and coconut oil con- taining single antioxidants (BHA, BHT or TBHQ) and diBHA. Typical chromatograms, shown in Figure 3, demonstrate that each antioxidant can be separated from diBHA in soybean oil 63 .HHo cmmn»0m CH «mmflo paw muampflxoflucw mo coaumummmm on Hmowmme m .mfim .Hfio m umm «mans mnom .amm mnooa in “HA0 6 men «mafia maoom .«mm @100H An “HHO 0 non 4mmflp maoom .Omma mnooa Am "GOHUMHUGGOGOU mamfimm “anon A0 “Hzom an “Hana Am “muflm mHmEmm “E: omm .umumsouonm >D "Houomump uucwfinfim "musumummEmu ucfiE\HE 0.0 “mums 30am uA>\>.~\m\mv Hocmusmlc\waHuuflcoumUM\Hmum3 ”mmmnm mafinoe “ca m\a x um v uncoflmcmEHp “Hamwuoo\ HO zaxmmmpcom ”GEDHOU 2.2 . mzrh m g (w N_ m g. 0 j i _. .oo_c_ .oo_c_