HUGO A. MASSALDI’ and C. JUDSON KING Dept. of Chemical Engineering, University of California, Berkeley, CA 94720 DETERMINATION OF VOLATILES BY VAPOR HEADSPACE ANALYSIS IN A MULTI-PHASE SYSTEM: d-LIMONENE IN ORANGE JUICE INTRODUCTION VAPOR HEADSPACE analysis by gas- liquid chromatography is commonly used in the food field as a means of identifying and determining relative concentration of volatile compounds in foods and model solutions (Wolford et al., 1963; Teranishi et al., 1963, 1966; Schultz et al., 1964, 1967; Chandrasekaran and King, 1971; etc.). The technique consists of equili- brating the vapor in contact with the liquid phase where the volatiles are present, at a fixed temperature in a closed vessel. An aliquot of the vapor is removed and injected in a chromatograph, now usually equipped with a flame-ionization detector. The area obtained for a given peak is assumed to be directly propor- tional to the liquid concentration of that substance. Proper calibration with a solu- tion of known concentration further allows determination of absolute concen- trations (Buttery et al., 1969; Massaldi and King, 1973). A headspace analysis technique has also been developed (Massaldi and King, 1973) to determine solubilities of scarce- ly soluble volatile compounds in water and aqueous solutions. When one is deal- ing with very insoluble substances or with substances of moderate solubility which are present at relatively high concentra- tion levels, the possibility exists that the solubility limit is exceeded in the liquid phase. In this case, the vapor concentra- tion is independent of the liquid concen- tration, and the peak area in a vapor sample is no longer sensitive to the con- centration in the liquid. The aroma components in a liquid food are often substances of low water solubility, al- though for most of these compounds the natural concentration is much lower than the solubility. However, important excep- tions can be pointed out. In the case of coffee, recovered aroma species can be dissolved, in coffee oil and added back to coffee extract to enhance the flavor, the result being a distribution of the aroma components between the aqueous and oil phases. In orange juice d-limonene, which ’ Present address: Departamento de Tec- nolo& Quimica, Facultad de Ciencias Exactas, UNLP. 47 Y 115, La Plata, Rep&blica Argentina accounts for 90-95% of orange oil, is known to be part of the water-insoluble fraction of the aroma. Further, it has been suggested (Scott et al., 1969; Miz- rahi and Berk, 1970). that orange oil combines somehow with the lipid frac- tion of the “cloud” in orange juice. It was found in the present work that d- limonene actually dissolves in a third phase, very probably the lipid fraction, in a way consistent with these observations. Under the conditions in orange juice, not only does the essence of oil separate out as another liquid phase, but also this liquid phase is in equilibrium with the oil dissolved in both the aqueous and lipid phases. Scott and Veldhuis (1966) developed a chemical method to estimate recoverable oil in orange juice, reported as d- limonene. The method is based on extrac- tion and distillation of the oil, with fur- ther titration of d-limonene. If all the oil is extracted from the juice, the procedure yields a direct estimation of the total oil content, characterized as d-limonene. In the present paper, an alternative tech- nique, based on vapor headspace analysis, is presented for the determination of d- limonene in synthetic emulsions and in orange juice. The method is general, how- ever, in the sense that it allows determina- tion of any volatile components distrib- uted between phases. It is also felt that it can provide a better understanding of the problem of phase equilibria in a liquid food. BASIS FOR THE METHOD When Nsi moles of a solute are allowed to partition between condensed phases L and F in equilibrium with a vapor phase V, the following mass balance holds, Nsi = yN, + XLNL + XF~‘F (1) where NV, NL and NF are the total num- ber of moles in phases V, L and F respe& tively, and y, XL and xF are the mole fractions of the substance in those phases. In Equation (l), it is assumed that a pure solute phase is not present, e.g., that the equilibrium partial pressure of the solute is sufficiently less than its vapor pressure. By defining an equilibrium ratio K as, and a distribution coefficient KF as, KF=z (3) and substituting Equations (2) and (3) in Equation (1) one obtains, Y Y Nsi = yN, + ~NL +KKFNF (4) which, solved for the vapor mole fraction, yields, Nsi ’ = NV + NJK + KFNF/K (5) When phase F is not present, Equation (5) reduces to, Nsi ‘=Nv+NL/li (6) Synthetic emulsions of d-limonene in aqueous solutions In this case, two liquid phases, a dis- persed, pure d-limonene phase and a continuous, aqueous phase are present. If an aliquot of this emulsion is mixed with a volume of water sufficient to dissolve all the d-limonene, and the new, diluted solution is equilibrated with a vapor phase in a closed flask, then Equation (6) holds, where subscript L stands for the aqueous phase. Under these conditions, for fixed values of NV and NL and assum- ing that K does not change with d- limonene concentration, y is directly proportional to Nsi, where Nsi is the number of moles of d-limonene originally present in the aliquot. If the dilution with water creates a final solute content below 5%, the value of K in Equation (6) can be taken as that for d-limonene in pure water. At 25°C K = 1.392 x lo3 in pure water (Massaldi and King, 1973). K is obtained as K = yP;/P, where y = activity coefficient of d-limonene, Pt = vapor pressure of d-limonene and P = total pres- sure. y is reported by Massaldi and King (1973) to be constant for yP/P,” < 0.8. In order for the d-limonene content of the original emulsion to be determined, Nsl must be calculated. A calibration may be made by headspace analysis of a simi- lar flask with the same NV and NL and with a known amount of d-limonene, N so, added. Equation (6) applied to the 434 -JOURNAL OF FOOD SCIENCE- Volume 39 (1974)