04.. o DETECTION AND [DENTIIFICATDON 0F FEED FLAVOR COMPONENTS JEN RAW AND PROCESSED WLK BY VAPOR PHASE CHROMATOGRAPHY Thesis for the Dogma of M. S. MICHIGAN STATE UNIVERSlTY John D» Wynn 1959 0". 07-. I ‘O'I'C'O‘OOQ-HWU'CVQNN IHESIS nun} .__; Z‘QI‘AI‘C Umvcrslty DETECTION AND IDENTIFICATION OF FEED FLAVOR COMPONENTS IN RAW AND PROCESSED MILK BY VAPOR PHASE CHROMATOGRAPHY By John Dee Wynn AN’ABSTRACT Submitted to the College of Agriculture of Michigan State'University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Dairy 1959 l" J . .. - ' -' 4.;1 -'-' :' I I . .,1' "'i Appro ued ' :' ' ' a. ' ' .-.' " ‘ '": 4' ".r".,"‘ “"=__ 1- 1 A L A I i L‘ - >I- - “L t ABSTRACT JOHN DEE WYNN Objectives of this study were (1) to obtain gas chro- matograms of normal flavored milk and various types of feed flavored milk and (2) to compare gas chromatographic sensi- tivity with organoleptic sensitivity as a means for evalu- ating the effectiveness of Vacéfieat pasteurization treat- ment in reducing feed flavor. The sources of volatile material studied were (1) milk during Vac-Heat pasteurization treatment, (2) raw milk prior to VacAHeat pasteurization treatment, (3) processed milk following Vac4Heat pasteurization treatment, and (4) milk from known controlled feeding trials. Three trials were conducted in which a single cow was fed fresh beet tops, alfalfa silage and fresh onion tops in quantities sufficient to produce typical feed flavors. Procedures employed were (1) reduced pressure collec- tion system to collect volatile flavor components from lab- oratory samples of milk and from.Vac4Heat pasteurizer, (2) a Perkin-Elmer, Model 154-B, Vapor Fractometer, with three columns, filled with different immobile phases to fraction- ate the collected volatiles and a matching recorder, (3) a Perkin-Elmer, Model 21, double-beam infrared spectropho- tometer equipped with a NaCl prism for analysis of indi- vidual components, (4) hydrozone formation from gaseous components passing through 2,4-dinitrophenylhydrazine traps, (5) collecting samples of milk during each volatile ABSTRACT JOHN DEE WYNN collection period to be given a flavor score and criticism by three Judges. Results show that acetaldehyde, acetone, methyl sul— fide and an unidentified aldehyde or ketone, in increasing order of concentration, were removed from milk by VacAHeat pasteurization treatment. Volatile materials collected from milk prior to Vacefleat pasteurization treatment were identical to those removed by the VacaHeat pasteurization treatment. These volatile components were somewhat lower in concentration than the volatiles from the VacAHeat. This is attributed to the small volume of milk used for laboratory collection compared to the large volume passing through the VacAHeat pasteurizer at the time of collection. The amount of volatile material found in milk after Vac-Heat pasteurization treatment was practically nil. The milk prior to Vac4Heat treatment was given a flavor score twice of 57, 57.5 and was criticized for having a feed flavor. A flavor score twice of 39.0 and 39.5 was given to this milk after the removal of volatiles by Vac-Heat pasteur- ization treatment. Aoetaldehyde, acetone, methyl sulfide, and an unidenti- fied aldehyde or ketone were found in the milk samples from each of the feeding trials and were present in the same order of concentration. The final trial for each of the alfalfa silage and onion feeding experiments yielded milk ABSTRACT JOHN DEE NYNN with a definite characteristic flavor. A volatile compound was isolated from each, which was specific for the feed flavor from which it came. These components were not identi- fied, due to their low level of concentration. No chromat- ographically recognizable volatile off-flavors were found to be associated with feeding as high as eighty pounds of fresh beet tops per ration. An average flavor score of 37 was given to this milk, which is considered good in flavor quality. DETECTION AND IDENTIFICATION OF FEED FLAVOR COMPONENTS IN RAW AND PROCESSED MILK BY VAPOR PHASE CHROMATOGRAPHY By John Dee Wynn A THESIS Submitted to the College of Agriculture of Michigan State'University of Agriculture and Applied Science in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Dairy 1959 ACKNOWLEDGMENTS The author is sincerely grateful to Dr. J. R. Brunner, Associate Professor of Dairying, for his willing advice and encouragement during the entire graduate study and for his assistance in the preparation of this manuscript. Sincere thanks are expressed to the entire staff of the Department of Dairy for their words of advice and as- sistance. Acknowledgement is also due the dairy plant per- sonnel for their cooperation during the sample collection periods. The writer sincerely appreciates the financial support of the Dairy Industries Supply Association and the funds and facilities provided by Michigan State University. The author is most grateful to his wife, Jane, for her assistance in preparing the manuscript and her encouragement during the course of study. TABLE OF CONTENTS INTRODUCTION . . . . . . . . . . . . . . . . . . . . . T mlla; OF LITEPUZLTUIIE o o o e o o c o e o o o o o o o o [‘0 EXPERIHZE AL RRCCEDURE . . . . . . . . . . . . . . . . 9 Sources of Volatile Flavor material . . . . . . . 9 Known feed flavors . . . . . . . . . . . . . 9 VacAHeat pasteurizer . . . . . . . . . . . . l0 Analytical Methods . . . . . . . . . . . . . . . 10 Collection of volatile material . . . . . . 10 Collection from laboratory samples . . . . 10 Collection from the Vac-Heat pasteurizer . 12 Analysis of volatile material . . . . . . . 13 Gas chromatography of volatile material . 13 Infrared spectrophgaetry . . . . . . . . . 16 Hydrozone formation . . . . . . . . . . . 17 Sensory evaluation . . . . . . . . . . . . 17 RESULTS . . . . . . . . . . . . . . . . . . . . . . . 22 Gas Chromatograms of Authentic Compounds . . . . 22 Volatiles Collected from the Vac-Heat Vacuum Chamber . . . . . . . . . . . . . . . . . . . 22 Volatiles Collected from milk Before and Following Vac-Heat Treatment . . . . . . . . . . . . . . 23 Volatiles of Milk from Controlled Feeding Programs . . . . . . . . . . . . . . . . . . . 24 ii TABLE OF CONTENTS (continued) Page DISCUSSION . . . . . . . . . . . . . . . . . . . . . . 47 Gas Chromatograms of Authentic Compounds . . . . 47 Volatiles Collected from Vac-Heat Vacuum Chamber 47 Volatiles Collected from Milk Before and Following Vac-Heat Treatment . . . . . . . . . 49 Volatiles of Milk from Controlled Feeding Programs . . . . . . . . . . . . . . . . . . . 50 SUMMARY AND CONCLUSIONS . . . . . . . . . . . . . . . 53 LITERATURE CITED . . . . . . . . . . . . . . . . . . . 56 iii LIST OF FIGURES FIGURE 1. Diagram of laboratory and Vac-Heat collection apparatus 0 O O O O O O O O O I O O O O I O O O 2. Flow schematic of the vapor fractometer . . . . 3. Part A - Reduced pressure volatile collection apparatus from Vac-Heat pasteurizer. Part B - Gas sample inlet valve of the vapor fractome- ter with liquid-air trap attached . . . . . . . 4. Perkin-Elmer, iodel 154-B vapor fractometer with its matching recorder , , , . . . . . . . 5. Gas chromatograms of a mixture of known com- pounds. Peak 1 - acetaldehyde, Peak 2 - methyl sulfide, Peak 3 - acetone. A,B,C,D represent successive samples of the same volatilized mix— ture O O O O O I O 0 O o O O O O O I O O O O O 6. Gas chromatograms of a mixture of known com- pounds. Peak 1 - acetaldehyde, Peak 2 - methyl sulfide and acetone. A,B,C,D represent success- ive samples from the same volatilized mixture . 7. Gas chromatograms of a mixture of known com- pounds. Peak 1 - acetaldehyde and methyl sul- fide, Peak 2 - acetone. A,B,C,D represent successive samples from the same volatilized mixture . . . . . . . . . . . . . . . . . . . . 8. Gas chromatogram of volatile materials col- lected from raw milk prior to the Vac-Heat treatnlent o o o o o o o o o o o o o O o o o O 9. Gas chromatogram of volatile materials col- lected from processed milk following Vac- Heat treatment . . . . . . . . . . . . . . . 10. Gas chromatogram of volatile materials col- lected during VacaHeat treatment of milk and resolved on a dioctylphthalate column , , . . . ll. Gas chromatogram of volatile materials col— lected during Vac-Heat treatment of milk and resolved on a didecylphthalate column . . . . . iv N ()1! LIST OF FIGURES (continued) 12. Gas chromatogram of volatile materials col- lected during Vac-Heat treatment of milk and resolved on a "Carbowax 400" column . . . . . . 33 13. Gas chromatogram of volatile materials col- lected from the milk of a cow fed alfalfa silage two hours prior to milking and re- solved on a dioctylphthalate column. 2x denotes a two fold increase in sensitivity . . 35 14. Gas chromatOgram of volatile materials col- lected from the milk of a cow fed green onion tsps two hours prior to milking and resolved on a dioctylphthalate column , . . . . 36 15. Gas chromatogram of volatile materials col- lected from the milk of a cow fed green onion tops two hours prior to milking and resolved on a dioctylphthalate column . . . . . 57 16. Gas chromatogram of volatile materials col- lected from the milk of a cow fed green beet tops two hours prior to milking and resolved on a dioctylphthalate column . . . . . . . . . 38 17. Gas chromatogram of volatile materials col- lected from the milk of a cow fed alfalfa silage two hours prior to milking and re- solved on a didecylphthalate column . . . . . . 39 18. Gas chromatogram of volatile materials col- lected from the milk of a cow fed green onion tOps two hours prior to milking and resolved on a didecylphthalate column . . . . . 40 19. Gas chromatogram of volatile materials col- lected from the milk of a cow fed green beet tops two hours prior to milking and resolved on a didecylphthalate column . . . . . , . . . 41 20. Gas chromatogram of volatile materials col- lected from the milk of a cow fed alfalfa silage two hours prior to milking and re- solved on a "Carbowax 400" column . . . . . . . 42 LIST OF FIGURES (continued) Page Gas chromatogram of volatile materials col- lected from the milk of a cow fed green onion tops two hours prior to milking and resolved on a "Carbowax 400" column , , , , , , ;3 Gas chromatogram of volatile materials col- lected from the milk of a cow fed green beet tops two hours prior to milking and resolved on a "Carbowax 400" column . . . . . . . . . . 45 Infrared spectrum of peak 8 in Figure 15 . . . 46 vi E t“ N F" o ('23 t1] LIST OF TABLES Present status of flavor components in milk . Retention volumes of known volatile compounds Retention volumes and concentration of vola- tile materials collected from milk before and following the Vac-Heat treatment and resolved on the dioctylphthalate column . Retention volumes and concentration of vola- tile materials collected during Vac-Heat treatment of milk and resolved on different chromatographic columns . . . . . Chromatographic characteristics of volatile material collected from milk of cows fed Specific flavor imparting feeds . vii 34 44 INTRODUCTION The objectives of this study were (1) to obtain gas chromatograms of normal flavored milk and various types of feed flavored milk and (2) to compare gas chromatographic sensitivity with organoleptic sensitivity as a means for evaluating the effectiveness of Vac-Heat pasteurization treatment in reducing feed flavors. It has been Just recently that volatile materials con— tributing to the various off-flavors of milk have been char- acterized. To this date there have been no studies in lit- erature regarding the identity of compounds responsible for feed flavors. This flavor is the most common off-flavor defect in a milk supply. Since most consumers base the quality of milk upon its flavor, a knowledge of compounds responsible for off-flavors are needed. Small amounts of compounds of chemical origin are not usually detected in milk by ordinary chemical methods. Gas chromatography has made it possible to more easily classify small quantities of these volatile compounds. This study has been an attempt to isolate and identify volatile materials responsible for specific feed flavors. The procedures used to accomplish these objectives are ex- plained in detail in the body of this thesis. REVIEW OF LITERATURE The techniques utilized in the chemical characterization of the volatile constituents of food materials may be divided into four basic steps: isolation, concentration, purifi- cation, and identification. Various methods for accomplish- ing these steps have been used in working with the volatiles of milk and are reviewed herein. Townley and Gould (1945) reported that when milk is subjected to high heat treatment volatile sulfides are pro- duced, largely H28, imparting a cooked flavor. The volatile sulfides were measured quantitatively by bubbling nitrogen through the milk during heat treatment and for thirty min- utes thereafter. The liberated sulfides were collected in alkaline zinc acetate and converted to methylene blue by the addition of a hydrochloric acid solution of p-aminodimethyl- aniline in the presence of ferric chloride. The color in- tensity, an indicator of the sulfide concentration, was measured photometrically. Gould (1951) and Patton (1955) published excellent reviews on compounds formed during high heat treatment of milk. A considerable amount of study has been devoted to the problem.of sunlight flavor in milk. Patton (1954) presented evidence that paper chromatograms containing methionine when sprayed with ninhydrin reagent emitted a broth like odor similar to that produced in milk by sunlight. The compound was identified, by means of infrared spectral data, as a - 3 - product of the sunlight catalyzed reaction between methio- nine and riboflavin. Forss, Font, and Stark (1954) used an elaborate pro- cedure for the isolation of compounds responsible for oxi— dized flavor in skimmilk. Distillates, which were combined and redistilled, were obtained from several batches of milk by steam distillation under a reduced pressure. The vola- tile components were extracted from the aqueous steam distil- late with light petroleum. After removing the solvent, by passing through a fractionating column packed with Fenske helices, a 100 milligram portion of oil, possessing a strong cardboard odor remained, and was called the crude 'extract'. The fractions of this extract which contained cardboard fla- vor were converted to 2,4-dinitrophenylhydrozones. The Girard reaction was used for the separation of the carbonyl compounds from the non-carbonyl compounds. This established the fact that the cardboard compounds were aldehydes and/or ketones. The hydrozones were identified by means of paper chromatography and light absorption at 351.5 millimicrons. A number of compounds were identified in a series of unsat- urated aldehydes. The normal alkyl 2 - enals containing 5 - 11 carbons as well as several dienals were implicated in the flavor. 2 - octenal and 2 - nonenal were found to exhibit the flavor most typically. These impart to milk, at a di- lution of one part in 107 to 109, an oxidized type of flavor resembling cardboard flavor. These compounds were believed to arise through autoxidation of the milk lipids. Knodt, Shaw, and White (1942) detected acetone bodies in milk from normal cows. Morgan, Forss, and Patton (1957) employed techniques similar to those of Forss gg_gl (1954) for characterizing volatile carbonyl compounds from high- heat treated skimmilk. They found relatively large amounts of acetone present in the distillate of raw skimmilk. Fur- fural and acetaldehyde were found to be the principal vola- tile carbonyl compounds generated in skimmilk by high-heat treatment. Harper and Huber (1956) used a rather extensive pro- cedure to isolate and identify the various carbonyl compounds in raw milk. The milk was deproteinized by adding 5 milli- liters of 10 per cent sodium tungstate, agitating and then acidifying to pH 4 with 0.67N sulfuric acid. The precipi- tated material was removed by centrifugation. The super- natant was filtered through a Whatman No. 1 filter paper, dampened and dusted with washed silicic acid. The sediment- ed layer was added on the filter and washed several times with 20 - 50 milliliters of distilled water at 25° C. The combined filtrate was reacted with 2,4 DNPH for one hour at 25° C. The phenylhydrazine derivatives were extracted with ether and chromatographed for acidic components. The chro- matographed acids were identified by a combination of Rf values and light absorption in the visible range. Acetone and acetaldehyde were identified by chromatographic analysis of the neutral hydrozones. Patton (1956) postulated that a compound with a struc- ture and properties similar to acetone might be involved in the normal flavor of milk. The component was isolated from the exhaust gases of an air-agitated thousand-gallon, cold- wall tank of unpasteurized whole milk and concentrated by passing through various trapping solutions and finally over activated coconut charcoal. The volatile substance was identified by comparing the characteristics of vapor phase chromatograms of the unknown and authentic methyl sulfide. It was confirmed further by infrared and mass spectral analysis. Day, Forss, and Patton (1957) used a low-temperature, reduced-pressure distillation technique for the recovery of volatiles from irradiated skimmilk. The procedure consisted of distillating the irradiated skimmilk for four hours at a pressure of 20 to 30 millimeters of mercury at a temperature of 40° C. Cold traps were employed to fractionate the vapors. The carbonyls collected in the wet ice traps were studied in the form of DNPH derivatives which were identified by paper chromatography, melting points, and spectral characteristics. The volatiles collected in the liquid nitrogen traps of the system were concentrated into a U-tube by manipulation of temperature and pressure. The volatiles were resolved by gas chromatography and collected into traps as they emerged from the column. Methyl sulfide, acetaldehyde, acetone, butanone and ethyl alcohol were identified in skimmilk ir- radiated at both 2x106 and 5x106 rep. Of these methyl sul- fide and acetaldehyde were the only materials of off-flavor and odor significance. The primary off-flavors found in these studies resulted from low molecular weight, reduced sulfur compounds. Only acetone was revealed in control samples of skimmilk. - Keeney and Doan (1951) reported that the predominant odor compounds from oxidized milk fat were ketones, a large proportion of them being unsaturated. Presumptive evidence was presented which suggested that lactones were present in the volatile material and these were correlated with the fruity and coconut-like odor of certain fractions of the distillate. Jackson and Morgan (1954) reported that a malty aroma elaborated by a certain species of Streptococcus was due to the conversion of leucine and isoleucine to isovaleraldehyde (3-methylbutanal). Taste observers revealed that as little as 0.5 p.p.m. of 5-methylbutanal added to milk simulates the characteristic aroma. A summary of components found contributing to various flavors of milk and the techniques employed from the litera- ture reviewed are presented in Table l. OBOQOQ NHOOO .s.o.o HtHoo.o efioQoQ moo 0800.00.» HOOImOOO cowueapceocoo havesoueoeam meme e commune“ .hcamamopesopco mam unawaMOpmsonno women a menace denounce e anomameumsonno homes a madden concave“ a heamawopmsoano Lemma coneamca w knamamopmsenno women a assaoo sHHeoaaooaoooea vehoaosm monogamooe ”as; aoeaoaaa Herpes mocouex use Amamcoao .mamcolm Human Hesaocv mouzneuae ooznmoamamamboma Hecoaspos monopoma .mecopmx Ammm haemamav aoeaoaea oaaoaaop venom mpcecoasoo sane gap seas swam eaaaenxo ease eoao>mflm seams sass eeao>aaa pamaaaam one mafia soaaoaxo game zap seamen» pea: Boost Havana: Aomaav ceased Amman mm mm mated Aammav comma: one momxomw Remedy copped AHmmHV seen one hocemx 335 3:8 one edge. if 82¢ l)? I I («Egg xawz :H monocoasoo uo>mah no mspmpm pcemoam H mqmde hemmemopmsoaso menace a momma oeaaooaaoooa Humans.“ escapee xaaaaaaa sea Ea: mm mm cameo: .s.o.o «00.0 cmummomos .1932. .a.o.o «Ho.o oeaaaaa assume .a.a.a Hmc.o oeaoflaaae assume 0 E 0 Q o Q om OCBQQDD .sdd 00m Epespnpeoom mews om 2303 .l I. .séd m4 Pamnwouwsonno new mgnmoampmom 33.5me 6333.5.“ CmmC He am has ocean escapee e095 agnooampmoe some» manooamfiuom compo. ofipmomopmom (we 00.0 engaged momma owuoomamxo {me am 00.8.55 e oflmpsamopmxnm woman hcommepmseso momma oacwooomdmxo in“... tea 3mm: .895 one amemm coapenpcoosoo cohoaase mega—Boos page.“ 35:09:00 6033» 323m: huflosfié 'I" Aeoaaaoaoov H mamas P EXPERIMENTAL PROCEDURE The immediate problem was the collection of volatile flavor components from the Vac-Heat pasteurizer and from laboratory samples of milk from animals on controlled feed- ing programs. A reduced pressure collection system, similar to the set-up used by Day gt_glp(1957) was employed to trap the volatiles. These materials were introduced to a vapor fractometer for separation and identification of the compo- nents. Vapor fractometers are especially useful in this respect although their sensitivity to small concentrations of volatile materials leaves something to be desired. Con- trol samples of milk were obtained each volatile collection period for flavor evaluation and correlation of removed vola- tile components with flavor scores. Sources of Volatile Flavor material Egggg_fggg_flavors. Controlled feeding trials were performed to introduce known "feed flavor" defects. Three trials were conducted in which a single cow was fed fresh beet tops, alfalfa silage, and fresh onion tops as required for the experiment. Fifty pounds of alfalfa silage was fed at various intervals approximately two hours before the evening milking. Onions were fed in the order of one and one-half pounds of chopped bulbs to ten and thirty pounds of fresh tops. Fresh beet tops were fed in the order of 50, 80, and 100 pounds. The milk was taken directly from the milking - 9 - 10 machine for collection of volatile material and flavor scoring. Vac-Heat pasteurizer. Preceding the vacuum treatment in the pasteurization process, volatile material was col- lected from milk obtained from the bulk storage tank by using the laboratory procedure discussed below. During the processing of the milk, volatile material was collected at various intervals from the vacuum unit of the Vac-Heat pas- teurizer. Milk was obtained after treatment and subjected to the laboratory collection procedure to determine the ex- tent of removal of volatile materials. The collection pro- cedure was empirical to assure that the results would be relative. Analytical methods Collection of Volatile Material Collection £32m_laboratory samples. The procedure used for recovering flavor volatiles from fresh raw milk was ac- complished by low-temperature, reduced-pressure distillation. The scheme is illustrated in Figure l. The procedure con- sisted of distillation at a vacuum of one inch of mercury and at a temperature of 48 - 50° C. The volatiles were fractionated by means of wet~ice, ethanol-dry ice, and liquid- air traps. The liquid-air traps were filled three-fourths full of glass beads to provide more surface area to insure efficient trapping of the volatile material. 11 The same procedure was followed, for the recovery of volatiles, from all milk in the laboratory. The wet-ice, ethanol-dry ice and liquid-air baths were filled and the traps were connected. Ten liters of milk were placed in a lZ-liter round bottom flask along with 40 p.p.m. of Dow- Corning antifoam A.F. emulsion. The flask was immersed in a warm water bath and attached to the trapping system. The vacuum pump was turned on and the vacuum adjusted to one inch of mercury. Nitrogen gas bubbled through the milk, providing agitation and a nitrogen atmosphere in the trap- ping system. The temperature of the water bath was adjusted to 48 - 50° c. A distillation period of three hours was sufficient for the collection of a working quantity of volatile components. Approximately 100 milliliters of total distillate were col- lected during the distillation period, most of which was condensed in the wet-ice trap as water. During the first experimental trials the individual traps were observed for odor. The ethanol-dry ice traps suggested the presence of volatile material while the first liquid-air trap (shown in Figure 1) contained the greatest concentration. At the end of the distillation period, the volatile material present in the ethanol-dry ice traps was concentrated into the first liquid-air trap by placing the ethanol-dry ice traps in warm water and decreasing the vacuum to one-half inch for 45 min- utes to insure sufficient time to remove the volatile material 12 from the ethanol-dry ice traps. The liquid-air traps were clamped off and disconnected from the ethanol-dry ice traps and while still cooled by a liquid-air bath were connected to the gas sample inlet of the gas chromatograph. Collection figm’ghg’Vac-Heat pasteurizer. The recove- ry of flavor volatiles from the vacuum pasteurization unit was accomplished with the same trapping assembly as pre- viously described. The collection assembly was attached to a three-eighths inch stainless steel tube, welded into the line preceding the vapor condenser located just beyond the second stage vacuum treatment (Figures 1 and 5A). Collec- tions were made for a period of 50 minutes to an hour. A second wet-ice trap was required to remove the large amount of water being removed from the system. The volatiles were collected and concentrated by the procedure already described. The traps constituted approximately 500 milliliters of con- densate, with most of this being in the first wet-ice trap. The presence of water soluble components was deter- mined according to the procedure of Jackson and Hussong (1958). The distillate was adjusted to a phenolphthalin end point (pH 8.4) with sodium carbonate and saturated with sodium chloride. This was extracted twice with 100 milli- liter aliquots of peroxide-free di-ethyl ether. The ether was evaporated slowly, at room temperature, to a very small volume in the presence of anhydrous calcium sulfate. A two- hundredth milliliter sample of the concentrate was examined 13 in the gas chromatographic unit for volatile components. Standard chromatograms for known materials were obtained at the same operating conditions to assist in the identification of the unknown volatiles when present. Analysis of Volatile Material §g§,chromatography gg volatile material. A Perkin- Elmer, Model 154-B, vapor fractometer was used for frac- tionating the collected volatiles. The apparatus is illus- trated in Figure 4. Figure 2 illustrates a flow schematic of the vapor fractometer. The sample inlet system was slightly modified, so that a liquid-air trap containing material for fractionation could be connected and the vola- tile material conveniently introduced into the system shown in Figure 3B. The gas sample inlet system is constructed to permit the carrier gas to pass through or bypass the sample-containing liquid air trap. Three columns filled with different immobile phases were employed in this study; namely, dioctylphthalate, di- decylphthalate, and "carbowax 400" (a polyethylene glyco M. N. 400). Reports in the experimental literature suggested that these immobile phases were applicable to the fraction- ation of volatile material present in milk (Day g£_gl 1957, Miller 1956, Perkin-Elmer 1956). The immobile phases were supported on 30 to 60 mesh acid washed celite and/or fire- brick. 14 The "carbowax 400" column was prepared in the following way. Thirty-five grams of crushed C23 firebrick 50-60 mesh were heated at 3000 C., washed with concentrated hydrochloric acid, then washed free of acid and dried at 1500 0. Fifteen grams of "carbowax 400" were mixed with the crushed fire- brick and then ethyl ether was added to insure that every particle would be uniformly coated with "carbowax 400". The ethyl ether was removed on a steam bath, leaving an al- most dry mixture. It was dried further in an oven at 2300 C. Twenty grams of the packing material were poured in a six foot, one-quarter inch (O.D.) cOpper tubing, and packed while constantly vibrating the tubing with an electric vi- brator.- After plugging the ends with glass wool the column was bent into a‘fl shape. The dioctylphthalate column was prepared by using twenty grams of prepared packing material from Burrell Corporation (30-60 mesh celite). This six foot, one- quarter inch (O.D.) copper tubing was packed and bent into a fl_shape in the same manner. The didecylphthalate column was six foot, one-quarter inch (O.D.) stainless steel tube packed by Perkin-Elmer Corporation. As a precautionary measure, all columns were flushed with helium for seven hours at a temperature of 1000 C. to remove any contaminants. The relative time intervals at which components emerge from the column are determined primarily by the column char- acteristics. The pressure and carrier gas flow used for each 15 column was determined by running known mixtures at different operating conditions. An operating temperature of 750 C. was selected for the oven of the vapor frectometer. The liquid-air trap containing the volatiles to be fractionated was attached to the gas sampling device. During this time helium (carrier gas) was passing through the column, but was bypassing the liquid-air trap. The liquid-air trap was opened to the system while still being bathed in liquid- air and helium was diverted through it for ten minutes to flush out a large amount of air. Then, helium flow was di- verted, thus bypassing the trap. While the trap was still attached to the device, the liquid-air bath was removed and in rapid succession the trap was quickly heated by immersion in hot Dow-Corning 550 011 (100° 0.). After three minutes had been allowed for heating the trap, the helium was again diverted through the trap to flush the volatiles into the column. Then the liquid-air trap was again bypassed by immediately returning the valve of the gas sampling device to its original position. The components observed on the recording chart were detected by Keane of a thermal conduc- tivity cell. Identities for fractionated components of the volatile mixture were obtained by comparing their retention volume with those of known compounds. Retention volume of a com- ponent is the volume of carrier gas passed through the column, from the time the component is injected until the 16 component reaches the column exhaust. Both the flow rate and the temperature determine the rapidity with which sample components emerge. The relative time intervals at which the components emerge, however, are primarily determined by the column characteristics. The same operating conditions and columns were used for the authentic compounds as used for the unknowns. These compounds were placed in a trap, at- tached to the sampling device, warmed up and flushed through the system as before. Infrared spectrophgéetry. A Perkin-Elmer, medal 21 double-beam infrared spectrophotometer equipped with a NaCl prism was used for analysis of individual components. These individual components were collected in liquid-air traps attached to the column sample collection valve. The liquid- air trap attached to the gas sample inlet valve was flushed several times in order to collect sufficient amount of the individual component. As the component would begin to emerge, the switch of the column sample collection valve was turned to divert the emerging component through the liquid-air collection trap. This was done a number of times to provide for a sufficient amount of volatile material for infrared analysis. The gas cell was evacuated and the liquid-air trap containing the component was attached, then warmed up to volatilize the component and the cell filled by opening the valves. The cell was closed and placed along side the reference cell in the infrared for analysis. The sample 17 beam window was set to approximately 100 per cent trans- mission and a scanning range of 2-14.5 microns was used. Hydrozone formation. Gaseous components were passed through 2,4-dinitrophenylhydrazine traps attached to the top of the sample collection valve. Ketone and aldehyde- type materials were converted to their 2,4-dinitrophenyl- hydrozones. These hydrozones were converted to crystals according to the procedure of Shriner, Fuson, and Curtin (1956). Hydrozones were filtered, washed four times with acidified water, and dissolved in warm 95 per cent ethyl alcohol. Water was added for crystal formation and let stand for 12 hours. The material was filtered and dried for subsequent melting point determinations. Sensory evaluation. Odor observations were made of each trap after the material was volatilized. Individual components were observed for odor at the exhaust of the column and by collecting in a liquid-air trap attached to the column sample collection valve. Control samples of milk were obtained for each volatile collection period. The milk from controlled feeding trials was given a flavor score and flavor criticism by three judges approximately twenty hours after it was drawn. Samples of milk collected before and following Vac4Heat treatment were scored approxi- mately four hours after they were collected. The difference in time of scoring was due to the fact that milk of con- trolled feeding trials was obtained at the evening milking. 18 .mspduemme ”3.300.300 psomuomb was hnoueuonua no acumedn .H onsmdh mahdcgmd zo_hou....oo bdwzuo<> was: mo. hm} .53.. .9; 2.4 345 azu 33:». pm. 213400 0... Oh 2:82; \ o... x / 53323 o» / ..ma<5 «.4 9:03 245 no. 535053”? .,_H.“_..m_s.. 83¢ a». / ‘34..er a. we a 355.34 2233.60 5.95693 fiIJ . .w , xm44u _ as... we. 53 20.5.3.5... K\ x, i 2305.: \ 2230 n / [J]. \i/ _'. 18 .mspuuwmnw "3.300.300 awomuoab Una macawnonoa no Samoa .H onzmfim m3h<¢ was: mo. hm} £13.”. 64> 024 2131.00 0... O... NOS—h OZN undhm hm _ .2505... \ 2. s / «352.8 o» / .22: a: 9.5... 2:: 3. 5550355 .L.H.H..,..w.n...,u.u. 333 Maw / ,.. . 34.519 «. ......._.. x r... \ 1 353.3% 298330 53383.. an] ’ xmcau _ .25 mo. 5; 20:54.53 K\ I, * zuooctz \ 2230 h / /I {470 J _.- 19 u>4<> @Zjac‘dm mdw w>J<> oz _J¢24aa3m m..<> 20.83.50 5flwgom 3:3 2:38 223400 zw>o > 2w .— 20 Figure 3. Part A - Reduced pressure volatile collec- tion apparatus from Vac-Heat pasteurizer. Part B - Gas sample inlet valve of the vapor fractometer with liquid-air trap attached. 21 Figure 4. Perkin-Elmer, Model 154-B Vapor Fractometer with its matching recorder. RESULTS Gas Chromatograms of Authentic Compounds Three columns were used to obtain gas chromatograms from a mixture of acetaldehyde, methyl sulfide, and acetone, shown in Figures 5, 6, and 7. The chromatograms in these figures are reduced five-eighths of their actual size. Table 2 shows the retention volumes of these authentic com- pounds. A, B, C, and D of each figure represents successive samples of the same volatilized mixture. The didecylphtha- late column was not capable of separating methyl sulfide and acetone, whereas the "carbowax 400" was not capable of sep- arating acetaldehyde and methyl sulfide. Volatiles Collected from the Vac-Heat Vacuum Chamber Gas chromatograms of volatile materials collected dur- ing Vac-Heat treatment of milk are shown in Figures 10, 11, and 12. Table 4 shows retention volumes and mole per cent concentration of these volatile compounds resolved on dif- ferent chromatographic columns. The mole per cent concen- tration of the original size of each.peak,was calculated by 'multiplying the width at half-height by the tetal height and dividing each area by the sum of all areas. All chromato- grams of unknown volatile material were reduced to one-half of the actual size. The dioctylphthalate column gave the best resolution, resolving the mixture into five components. - 22 - The retention volume of peaks 1, 2, and 5 is approximately that of acetaldehyde, methyl sulfide, and acetone, respec- tively. The elution of peaks 3 and 4 into 2,4-dinitr0phenyl- hydrazine indicated that they contained a carbonyl group. Peak 2 gave a negative test with 2,4-dinitr0phenylhydrazine. Odor observations of each peak posessed a different, dis- agreeable odor. Peak 5 which appeared on all chromatograms of the unknown material is due to water condensing from the air in the traps. The didecylphthalate and the "carbowax 400" columns each resolved the volatile mixture into three components. Volatiles Collected from.Milk Before and Following Vacsfleat Treatment Figure 8 shows a gas chromatogram of volatile material collected from milk before VacAHeat treatment. All peaks of this chromatogram have approximately the same retention volumes as the authentic compounds and components collected during Vac-Heat treatment. Peaks 2, 5, and 4 gave the same results with 2,4-dinitrophenylhydrazine as those from Vac- Heat treatment. Odor observations of these components ap- peared to be the same as components from VacAHeat treatment. A gas chromatogram of the volatile material collected from the milk following Vac-Heat treatment is shown in Figure 9. The dioctylphthalate column resolved this material into two components. The amount of volatile material present 24 in milk after Vac-Heat treatment is shown to be practically nil. Table 3 presents retention volumes and mole per cent concentration of this material. Flavor scores and flavor criticisms are also indicated in this table from milk before and following VacAHeat treatment. Volatiles of Milk from Controlled Feeding Programs Figures 15 through 22 show gas chromatograms of vola- tile materials collected from the milk of cows fed specific flavor imparting feeds. Alfalfa silage flavored milk shows a higher number of components when resolved on the dioctyl- phthalate and didecylphthalate columns than did the other collections throughout the experiment. Table 5 contains the flavor score given to this milk, which was slightly lower than scores from milk of the other two feeds, with the ex- ception on one trial with onions. The retention volumes and concentrations for components of each specific feed flavor milk are also recorded in this table. Figure 23 shows the infrared spectrum of Peak 3 in Figure 15, with an absorption maxima at 4.5 and 14.5 microns. No conclusions were made on the basis of this observation as to the nature of the compo- nents. l ” 1 Column ------- Digctylphthalate Temperature-- 75 C. j ‘ Helium Flow-- 44 ml./min. f Sensitivity-- 64 r m l m ‘3- a 2 3 3 f i 9' E4 53 co __ kale) L_ ._ L L) fig 10 5 C 10 5 o I" F ‘ f I 3 i C D i _. m ‘_ I: H d «1:; 2 l 1 A {i 1 1: I . . f i ' I 3 i F P E 3-P ‘f‘ «2 g f. 1 g; E l (E); a, ‘ ES if ' E3 f m 1' L C’) __L) ~¥—> #14 j 1' J 10 5 O 10 5 5— Figure 5. Gas chromatograms of a mixture of known com- pounds. Peak 1 - acetaldehyde, Peak 2 - methyl sulfide, Peak 3 - acetone. A,B,C,D represent successive samples of the' same volatilized mixture. l I W Column ------- Didecylphthalate . ' Temperature-- 75° C. Helium Flow-- 48 ml./min. Sensitivity-- 64 2 l a a f T TAIR :h'd lAIB 26 § § l v u] I __g_ [J L 10 5 o 10 5 O 2 2 i .9 2 l l l )1 J l ___l_i 1 j A v —‘ _' 10 5 0 10 5 Figure 6. Gas chromatograms of a mixture of known compounds. Peak 1 - acetaldehyde, Peak 2 - methyl sul- fide and acetone. A,B,C,D represent successive samples from the same volatilized mixture. C“r‘ l i [ Column ------- "Carbowax 400" Temperature-- 75 C. m » Helium Flow-- 75 ml./min. :1 Sensitivity-- 64 1 r A E. l 2 2 54 F" En; St: 5 53 a) ¥ 1 l lfi J l_l 1 10 5 0 lo 5 N D: H a l C .. 2 2 1 i 2 ii 6* E4 a: N DISCUSSION Gas Chromatograms of Authentic Compounds Figures 5, 6, and 7 Show a different retention volume for each authentic compound. The operating conditions for each column were the Same except for flow rate. The dioc- tylphthalate and didecylphthalate columns were operated at a range difference of four milliliters in flow rate. They gave retention volumes of 156 and 80 milliliters, respec— tively for acetaldehyde. Such a range in retention volume of these columns is due to the difference in column char- acteristics. Part A of each figure shows the most volatile compound being resolved first and having the highest con- centration. Successive chromatograms in each figure show that the least volatile compound --highest molecular weight-- increase in concentration while the most volatile compounds decrease with each successive chromatographic sample. Volatiles Collected From VacéHeat Vacuum Chamber The results Show that acetaldehyde, acetone, methyl sulfide, and an unidentified aldehyde or ketone, in increas- ing order of concentration, were removed from milk by Vac~ Heat treatment. The odor from each liquid air trap was characteristically cowy, which was similar to a mixture of acetone and methyl sulfide. Patton (1956) stated that a cowy odor was characteristic of methyl sulfide. All the - 47 - 48 volatile material of component 4 in Figure 10 was converted to their 2,4-dinitrophenylhydrozones. A sufficient quantity of crystals was not obtained for a melting point determin- ation. The same volatile material collected at a different period was resolved into three components by the didecyl- phthalate column (shown in Figure 11). In this chromatOgram it was presumed that acetaldehyde was masked by the air peak. Peak 1, of this same figure, represented a combination of methyl sulfide and acetone. An unsuccessful attempt for in- frared analysis was made on all the volatile material con- stituting peak 2. This peak was the same component as peak 4 in Figure 10. The "carbowax 400" column was not as re- solving for the volatile materials as the other two columns. MOrgan gt_gl (1957) found relatively large amounts of acetone present in the distillate of raw skimmilk. Day 22 31 (1957) also found acetone in substantial quantities (1.0- 10.0 p.p.m.) from the distillation of raw skimmilk. Acet- aldehyde was found in a trace amount. These were studied in the form of 2,4-dinitrophenylhydrozones. In this study no water soluble material was detected in the water distillate from the wet ice and ethanol-dry ice traps after concentrat- ing the volatile material into the liquid air trap. Using the same operating conditions the peaks of the chromatograms showed no difference. If any water soluble material was present, it was in such a small concentration that the sen- sitivity of the vapor fractometer was not sufficiently Sharp 49 for detection. Another likely explanation could be that only a very small sample of the material was introduced into the column. Larger volumes of volatile-containing water distillate would have saturated the column. Volatiles Collected From Milk Before and Following Vac4Heat Treatment Volatile materials collected from milk prior to Vac- Heat treatment were identical to those removed by the Vac- Heat treatment. These volatile components are somewhat lower in concentration than the volatiles from the Vac-Heat. This is attributed to a small volume of milk used for lab- oratory collection, compared to the large volume passing through Vac-Heat at the time of collection. Harper and Huber (1956) found formaldehyde, acetaldehyde, and acetone in trace amounts as occuring in raw milk. Patton (1956) isolated methyl sulfide from the exhaust gases of raw milk. The Slight difference in retention volume of the resolved peaks in the sample from the Vac-Heat unit was due to the difficulty in duplicating the exact air flow and temperature in the chromatographic column. It is very necessary to con- trol the column temperature as precisely as possible for accurate comparison of the patterns from analysis to analy- sis. Phillips (1956) stated, "Retention volumes are very dependent upon the column temperature; a 10 C. variation of temperature giving rise to about 5 per cent change in retention volume." The milk prior to Vac-Heat treatment was given a flavor score twice of 57, 57.5 and was criticized for having a "feed" flavor. The removal of volatiles during Vac-Heat treatment is effective in removing off-flavors. Patton (1956) sug- gested that methyl sulfide and other regularly occuring volatile material contribute significantly to the character- istic flavor of milk. Table 5 Shows the retention time of the very small amount of volatile material remaining after Vac-Heat treat- ment. This material apparently has a higher molecular weight and is not as volatile as the material that was com- pletely taken out during the Vac-Heat treatment. A flavor score twice of 59.0 and 59.5 was given to this milk after the removal of these volatiles. The quality of this milk was considered satisfactory. A nonvolatile cooked flavor develops in milk as a result of HTST and Vac-Heat treatment. Patton (1955) reported this to be associated with browning of milk. AS these off-flavor volatiles are removed from milk. The flavor score is improved in relation to the a- mount of volatile material removed. Volatiles of Milk From Controlled Feeding Programs Alfalfa Silage flavored milk yielded one more volatile component than the normal, mixed herd milk. This component was not identified due to its low concentration and the 51 methods of identification used. Infrared sensitivity was not high enough to allow identification of such a small concentration. Corresponding retention volumes with au- thentic compounds could not be obtained for this compound. The odor of the liquid air traps was essentially the same aroma as previously noted. Three judges criticized this milk as having a strong feed flavor and scored it at 56. It appears as the flavor score becomes considerably lower, the concentration of volatiles increase. The didecylphtha- late column resolved volatile material from another trial of alfalfa Silage flavored milk into the same components. The volatile material collected from the milk resulting from the first two feeding trials on green onion tops yield- ed a typical cowy odor in the collection traps. The vola- tiles collected from the milk of the third trial, which was increased from 10 to 50 pounds of green onion tops, smelled very distinctly of onion odor. The flavor score and flavor criticism indicated the presence of a definite onion flavor. Milk from the first two trials was given a higher score and criticized for having a "feed" flavor. The volatile mater- ial from the last trial yielded an additional component on the dioctylphthalate column, Figure 14, peak 1. The peak size indicated that this component was present in low con- centration. The odor of this component was faintly remines- cent of onion odor as it emerged from the column outlet. Stahl (1957) reports that propionaldehyde, methyl alcohol, and propyl mercaptan are very abundant in the volatiles of onions. The component separated on the vapor fractometer could not be identified due to its low concentration and the low sensitivity in the infrared Spectrum. Peak 5 in Figure 15 was Similar to the trailing component resolved in other chromatograms throughout this study. This peak consistently gave a putrid odor as it was eluted from the column The high concentration of the component in this run suggested the possibility of collecting it for an infrared analysis. The infrared spectrum (Figure 25) revealed an absorption maxima at 4.5 microns and 14.5 microns, indicative of the C=O of an aldehyde or ketone. The chromatographed volatiles of beet top flavored milk indicated the presence of fewer volatile components. The odor of the liquid air traps was suggestive of a slight cowy aroma. An unsuccessful attempt for infrared analysis was made on the volatile material of peak 2 in Figure 16. This milk was given a high flavor score for raw milk. The inten- sity of the flavor sweetness increased as the amount of beet tops was increased. Trout and Taylor (1955) reported the off-flavor of fresh beet tops as being "neutralizer", "bak- ing soda", and "sweetened". No volatile off-flavor appeared from three trials of feeding fresh beet tops as determined by vapor phase chromatography. SUMMARY AKD CONCLUSIONS The primary objectives of this study were (1) to obtain gas chromatograms of normal flavored milk and various types of feed flavored milk and (2) to compare gas chromatographic sensitivity with organoleptic sensitivity as a means for evaluating the effectiveness of Vac-Heat pasteurization treatment in reducing feed flavors in milk. The sources of volatile material studied included: (1) milk during Vac- Beat pasteurization treatment, (2) raw milk prior to Vac- Heat pasteurization treatment, (5) processed milk following VacAHeat pasteurization treatment, and (4) milk from known controlled feeding trials. The unknown volatile material was tentatively identi- fied by comparing the retention volumes of its peaks arising in gas chromatograms to those of authentic compounds on gas chromatographic columns containing three different immobile phases. .Infrared analysis and hydrozone formation were con- fined to the volatile compound appearing on the chromato- gram in the highest concentration. Acetaldehyde, acetone, methyl sulfide, and an unidenti- fied aldehyde or ketone in increasing order of concentra- tion were detected as the volatile compounds being removed from milk by VacAHeat treatment. Volatile materials collect- ed from milk prior to Vac-Heat treatment were identical to those removed from milk during Vac-Heat treatment. The quantity of volatile material collected from milk following - 53 - Vac-Heat treatment was practically nil. Flavor scores on this milk indicated that it was of good flavor quality, and superior to the same milk prior to Vac-Heat treatment. Acetaldehyde, acetone, methyl sulfide, and an unident- ified aldehyde or ketone were found in the milk Samples from each of the feeding trials and were present in the same order of concentration. The final trial for each of alfalfa si- lage and onion feeding experiments yielded milk with 8 def- inite, characteristic flavor. A volatile compound was iso- lated from each, which was Specific for the feed flavor from which it came. These components were not identified by retention volume or by other means, due to their low level of concentration. No chromatographically reCOgnizable vola- tile off—flavor components were found to be associated with feeding as high as 80 pounds of fresh beet tops per ration. An average flavor score of 57 was given to this milk by three judges, which is considered good in flavor quality. The compounds isolated in this study were believed to contribute Significantly to "cowy" and "feed type" flavored milk. Removal of these by the Vac-Heat treatment resulted in a uniform and a good flavor quality milk. Perhaps other volatile materials were present, but at too low a concen- tration to be detected by the vapor fractometer. Even‘ though these minor constituents may not have been detected by vapor fractometry, conceivably, they could contribute to the normal flavor of milk. Although the technique of gas chromatography possesses limitations along with its many advantages, it is believed that compounds responSible for Specific feed flavors can be isolated, separated, and positively identified providing very sensitive equipment such as mass spectrometry is used in conjunction with the vapor fractometer. (2) (5) (4) (5) (6) ('7) (8) Day, E. A., 1957a. l9"3'7"""'b. ' LITERATURE CITE‘ Forss, D. A., and Patton, 8. Flavor and odor defects of gamma-irradi- ated skimmilk. I. Preliminary observa- tions and the role of volatile carbonyl compounds. J. Dairy Sci., 40:922-951. 9 FIavor and odor defects of gamma-irradi- ated skimmilk. II. Identification of volatile components by gas chromatography and mass spectrometry. J. Dairy Sci., g9: 932-941. Forss, D. A., Pont, E. G., and Stark, N. 1955. Could, I. A. 1951. The volatile compounds associated with oxidized flavour in skimmilk. J. Dairy Research, 2§;9l-102. Compounds formed by heat-treatment of milk. Nfllk Plant Monthly, 49(2):44-49, 68. Harper, V. J., and Huber, R. M. 1956. Jackson, H. 1958. Jackson, H. 1954. Keeney, H., 1951. Some carbonyl compounds in raw milk. J. Dairy Sci., 5251609. J., and Hussong, R. V. Secondary alcohols in blue cheese and their relation to methyl ketones. J. Dairy Sci., 41:920-924. 0V0, and LIOI‘gan, M. E. Identity and origin of the malty aroma substance from milk cultures of Strep- tococcus Lactis var. Malti enes. J. Dairy Sci., 52:1516-1524. and Doan, F. J. Studies on oxidized milk fat. III. Chem- ical and organoleptic properties of vola- tile material obtained by fractionation with various solvents and Girard's reagent. J. Dairy Sci., 54:728-754. - 56 _ (9) (10) (ll) (12) (13) (14) (13) (16) (17) 57 Knodt, C. B., Shaw, J. C., and White, G. C. 1942. Studies on ketosis in dairy cattle. II. Blood and urinary acetone bodies of dairy cattle in relation to parturition, lacta- tion, gestation, and breed. J. Dairy Sci., 25:851-860. Morgan, N. E., Forss, D. A., and Patton, S. 1957. Volatile carbonyl compounds produced in skimmilk by high-heat treatment. J. Dairy Sci.,‘ggz57l-578. Miller, c. T 1956. Gas chromatography: Automatic analyzer for the chemist. Res. and Engin. February. 499. Patton, S. 1954. The mechanism of sunlight flavor formation in milk with special reference to methio- nine and riboflavin. J. Dairy Sci., 51: 446-452. Patton, S. 1955. Browning and associated changes in milk and its products: A review. J. Dairy Patton, S. 1956. Methyl Sulfide and the flavor of milk. Perkin-Elmer Corporation 1956. Instruction manual for Model 154-B. Vapor Fractometer. 24pp. (multilithed) Norwalk, Conn. Phillips, C. 1956. Gas Chromato raoh . 1st Ed. Butterworths ScIentifIc FublIcationS, London. 105pp. Shriner, R. L., Fuson, R. C., and Curtin, D. Y. 1958. Thg Systematic Identification of Organic Cgmpgunds. 4thmEd. John NiIey_and Sons, New York. 426pp. (18) (19) (20) Stahl, W. H. 1957. Gas chromatography and mass spectrometry in the study of flavor. Chemistry of natural food flavors. A symposium. pp. 58-76. Quartermaster Food and Container Institute for the Armed Forces, 1819 N. Pershing Road, Chicago 9, Illinois. Townley, R. C., and Gould, I. A. 1945. A quantitative study of heat labile sul- fides of milk. I. Method of determina- tion and the influence of temperature and time. J. Dairy Sci., 26:689-705. Trout, G. M., and Taylor, G. E. 1955. The effect of beet tops on the flavor and odor of milk. Mich. Agr. Expt. Sta. Quart. Bu11., 1§(1):57-45. APR 13 196) 5'