Anal. Chem. 1999, 71, 1692-1698
Articles
Determination of Oxidative Stability of Oils and Fats Kang Tian and Purnendu K. Dasgupta*
Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-1061
In a new approach to evaluating the oxidative stability of oils and fats, the consumption of oxygen by a sample confined in a reactor of adjustable temperature is monitored with a gas-phase flow injection analysis (FIA) system. Temperature-dependent data are collected in a low-oxygen-content atmosphere. For a variety of samples, log(oxygen consumption) is linearly related to the reciprocal of the absolute temperature (minimum linear r2 > 0.99). This makes it possible to extrapolate the temperature-dependent data to predict the stability of the samples at other temperatures, e.g., typical ambient storage temperatures at which the direct determination of oxidative stability would be too slow for most samples. The proposed method is instrumentally simple and is easily automated. The sample throughput rate is an order of magnitude faster relative to current alternatives; temperature-dependent stability characterization for a sample (three temperatures, triplicate measurement at each temperature) requires e 2 h. The reproducibility of the results is excellent. For a cottonseed-oil sample studied over 3 days, the slope and intercept of the log(O2 consumption) vs 1/T linear plot (for all the 45 measurements made) exhibited uncertainties of 2.1% and 2.0% for the slope and the intercept, respectively, with a linear r2 value of 0.9929. In a high-temperature (160 °C) oxidation experiment with various oils, the oxygen consumption was wellcorrelated (linear r2 0.9692) with the concomitant decrease in iodine absorption number (IAN). In contrast, it was poorly and negatively correlated with an increase in the peroxide value.
It is well-known that when left exposed to an oxygen-containing atmosphere, oils and fats undergo oxidative degradation. This process mainly involves the oxidation of unsaturated fatty acids or their derivatives present in oils and fats. The oxidation of saturated fat is generally much slower.1 The mechanism of lipid * Corresponding author. Tel.: (806) 742-3064. Fax: (806) 742-1289. E-mail:
[email protected]. (1) Brodnitz, M. H. J. Agr. Food Chem. 1968, 16, 994-999.
1692 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999
oxidation has been well-characterized for some time.2 The reactions of unsaturated fatty acids and oxygen to produce peroxides require relatively high activation energy.3 This suggests that the direct attack of oxygen on unsaturated fatty acids is improbable. Lipid autoxidation is generally believed to involve a free-radical chain mechanism: (1) initiation steps that lead to free radicals (R•), (2) propagation of the free radicals (R• + O2 f ROO•, ROO• + RH f ROOH + R•), and (3) termination steps (R• + R• f R-R, R• + ROO• f ROOR, ROO• + ROO• f O2 + ROOR (or alcohol and carbonyl compound)). The oxidation of lipids thus results in peroxides as the primary oxidation products. In most cases, an induction period is observed before the onset of the significant production of peroxides. However, even during this period, oxygen is consumed in a zeroorder process,4 apparently leading to poorly characterized initial intermediates prior to the formation of the peroxides.5 The peroxides, in turn, degrade further to products such as aldehydes, ketones, and carboxylic acids,6 leading to rancidity, loss of nutritional value, and increase in toxicity,7 which, in turn, contribute to diseases and accelerate the aging process.8,9 The degree of unsaturation, the presence of indigenous or added antioxidants, prooxidants, and thermal and illumination conditions of storage all affect lipid oxidative stability. Even the same sample type may exhibit different stabilities. The types and degrees of unsaturation present in porcine lard have been shown, for instance, to be diet-dependent.10 The stability also depends (2) See for examples: Bloomfield, G. F.; Sundralingham, A.; Stutton, D. A. Trans. Faraday Soc. 1942, 38, 348-355. Bolland, J. Q. Rev. London 1949, 3, 1-21; Farmer, E. H.; Bateman, L. Q. Rev. London 1954, 8, 147-167; Swern, D. In Autoxidation and Antioxidants, Vol. 1; Lundberg, W. O., Ed.; Interscience: New York, 1961; pp 1-54; Hawkins, E. G. E. Organic Peroxides; Van Nostrand: New York, 1961; pp 355-409. (3) Labuza, T. P. Crit. Rev. Food Technol. 1971, 2, 365-368. (4) Labuza, T. P.; Bergquist, S. J. Food Sci. 1983, 48, 712-715. (5) Privett, O. S.; Blank, M. L. J. Am. Oil Chem. Soc. 1962, 39, 465-469. (6) Shahidi, F. In Natural Antioxidants. Chemistry, Health Effects and Applications; Shahidi, F., Ed.; AOCS Press: Champaign, IL, 1997; pp 1-11. (7) See for examples: Kaneda, T.; Ishii, S. Biochem. (Tokyo), 1954, 41, 327335; Rackis, J. J.; Sessa, D. J.; Honjg, D. H. J. Am. Oil Chem. Soc. 1979, 56, 262-271; Moll, C.; Biermann, U.; Grosch, W. J. Agric. Food Chem. 1979, 27, 239-243. (8) Steinberg, D.; Parthasarathy, S.; Carew, T. E.; Choo, J. C.; Witztum, J. L. New Engl. J. Med. 1989, 320, 915-924. (9) Harman, D. In Free Radicals in Biology; Pryor, W. A., Ed.; Academic Press: New York, 1982; Vol. 5, pp 255-275. (10) Stine, W. R.; Faut, O. D.; Rees, M.; Stockham, E. B. In Applied Chemistry, 2nd ed.; Allyn and Bacon, Inc.: Boston, 1981; pp 380-383. 10.1021/ac981365t CCC: $18.00
© 1999 American Chemical Society Published on Web 03/31/1999
on the exact source of the lipidsthe stability of piscine lipids from the skin is different, for example, from that of lipids from other tissues.11 Evaluation of the oxidative stability of lipids is an old and complex topic. Still, a generally applicable, fully satisfactory method is yet to emerge. For oil and fat samples, it is important to know both the current oxidative status and the relative potential to undergo oxidative degradation. A single measurement of the peroxide content (peroxide value, PV) can be used as an index of current oxidation status only if the peroxides formed are stable enough so that they do not decompose after formation. For very easily oxidized lipids, at least, this is not true. While the activation energy for the peroxide formation from unsaturated acids is about 146-272 kJ/mol,3 the activation energy for the decomposition of several lipid peroxides has been reported to be in the range of approximately 84-184.5 kJ/mol,12 suggesting that the peroxides are less stable than the lipids themselves. It is also well-known that PV and the onset of rancidity are not always well-correlated; it is possible for a rancid sample to have a low PV.13,14 Ironically, oxidative stability assessment based on a single PV measurement is often used to estimate the shelf life of a product (time for the onset of rancidity, with an adequate safety margin). Even if PV may provide an indication of the current oxidative status, it can give no information on the relative potential of a sample to oxidize. Broadly, there are two approaches to determining potential oxidative stability. The first is to evaluate the induction period observed before significant production of peroxides (or secondary products) begins. The induction period is very dependent on the conditions of the oxidation experiment. As such, the oxidation experiment must be carried out according to some set protocol. The most commonly used method is called the active oxygen method (AOM), promulgated by the American Oil Chemists Society.15 This method has been largely unchanged in the past five decades. Several 20-mL aliquots of fat samples are taken, aerated at 2.33 mL/s at 98 °C, and periodically analyzed for PV by an iodometric procedure. The time to reach a PV of 100 mequiv/kg is taken to be an index of the oxidative stability (the actual endpoint in industrial practice can be as little as 20 mequiv/kg, depending on the type of fat16). A large amount of sample, numerous analyses, and a critical control of the air flow rate are needed. A typical AOM determination requires 20-40 h. For samples that form relatively unstable peroxides, a PV of 100 mequiv/kg may never be reached, and an AOM measurement can have little meaning. The induction period can also be monitored by monitoring the consumption of oxygen. In the
Sylvester test,17 the sample is heated to 100 °C in a closed vessel, and the pressure decrease, due to the consumption of oxygen, is monitored. An automated embodiment of this is the Oxidograph17 in which the sample is heated in a reaction vessel under 100% oxygen, and the induction period is determined from a sudden pressure decrease. Another official method that measures the induction period is the oil stability index (OSI) method.18 A stream of purified air is passed through a hot lipid sample, and the effluent air is bubbled through deionized water. The conductivity of the water is continuously monitored. As the final oxidation products, volatile carboxylic acids (mostly formic acid) are formed in the oil or fat sample. These are flushed out by the flowing air and are collected in the water, resulting in an increase in the conductivity. The time of the onset of the major conductivity rise of the water is obtained by tangential extrapolation. The OSI values are generally wellcorrelated to the corresponding AOM values if the PV is 100 or greater.19 The method is automated and thus easier to use than the AOM. However, it is time-consuming as well and suffers from the common problem that large errors can be incurred from small variations in the air flow rate.20 The element of deeper concern is that carboxylic acids are further down the degradation chain and are only one of the possible products of hydroperoxide decomposition. The OSI is thus based on an event which occurs well after the process of interest and may not bear a linear correspondence with the latter. Moreover, several volatile acids other than formic acid are known to form in different ratios in different oil/fat samples.18 Further, with formic acid, itself, significant losses from the collector solution may occur if collection temperatures exceed 20 °C.18 The other principal option for assessing potential oxidative stability is monitoring the rate of peroxide formation or oxygen consumption. Versions of the AOM that attempt to measure the oxidative stability in a shorter period involve a single measurement of the PV after 4 or 20 h of aeration. In oxygen-consumption methods, the sample is sealed under air or oxygen and stored at a constant temperature. The oxygen concentration in the headspace is monitored by periodically withdrawing small samples through multiply sealed septa affixed to the container and analyzing them by gas chromatography. Several different types of samples have been analyzed.21,22 The fundamental problem with all of these methods is that lipid stability is determined at a fixed temperature, usually far above the ambient, because it takes an unacceptably long time otherwise to make meaningful measurements. The result must be extrapolated to make a decision about the comparative stabilities of two different types of samples under ambient storage
(11) Yamada, J. Bull. Tokai Reg. Res. Fish. Lab. 1979, 99, 23-29; Yamaguchi, K.; Toyomizu, M.; Nakamura, T. Bull Jpn. Soc. Sci. Fish. 1984, 50, 12451249. (12) Swern, D. In Organic Peroxides; Wiley-Interscience: New York, 1970; Vol. 1, pp 115-128. (13) Helrich, K., Ed. Official Methods of Analysis of the Association of Official Analytical Chemists, 15th ed.; The Official Methods of Analysis of Official Analytical Chemists, Arlington VA, 1990; pp 955-956. (14) Coppen, P. P. In Rancidity in Foods, 2nd ed.; Allen, J. C., Hamilton, R. J., Eds.; Elsevier Applied Science: London, 1989; pp 83-104. (15) American Oil Chemist’s Society Official Method Cd-12-57; American Oil Chemists Society: Champaign, IL, 1983. (16) Swern, D., Ed. Bailey’s Industrial Oil and Fat Products, Wiley: New York, 1996; Vol. 1.
(17) Wewala, A. R. In Natural Antioxidants. Chemistry, Health Effects and Applications; Shahidi, F., Ed.; AOCS Press: Champaign, IL, 1997; pp 331345. (18) American Oil Chemist’s Society Official Method Cd-12b-92; American Oil Chemists Society: Champaign, IL, 1993. (19) La¨ubli, M. W.; Bruttel, P. A. J. Am. Oil Chem. Soc. 1986, 63, 792-795; deMan, J. M.; Tie, F.; deMan, L. J. Am. Oil Chem. Soc. 1987, 64, 993-996. (20) Hill, S. E.; Perkins, E. G. J. Am. Oil Chem. Soc. 1995, 72, 741-743. (21) See for examples: Bunick, F. J. Lipid Autoxidation in Human Blood Cell Membrane. Ph.D. Dissertation, University of Massachusetts at Amherst, Massachussetts, 1984; Chen, Z. Y.; Nawar, W. W. J. Am. Oil Chem. Soc. 1991, 68, 47-50; Ohshima, T.; Fujita, Y.; Koizumi, C. J. Am. Oil Chem. Soc. 1993, 70, 269-276. (22) Chen, Z. Y.; Ratnayake, W. M. N.; Cunnane, S. C. J. Am. Oil Chem. Soc. 1994, 71, 629-632.
Analytical Chemistry, Vol. 71, No. 9, May 1, 1999
1693
conditions. This extrapolation makes the tenuous assumption that the activation energies for the lipid oxidations of the two samples are the same. It has been shown that the rate of lipid oxidation can be dependent on whether the sample is ground or whole and whether the lipid is isolated from the sample and then examined by itself.22 The assumption that two altogether different samples have identical activation energies is obviously difficult to justify. We propose here a new system to determine the stability of oils and fats. An automated, stopped-flow gas-phase flow injection analysis system using a programmable temperature reactor coupled with an electrochemical oxygen sensor is used to measure the oxygen consumption of the same sample at various temperatures. The temperature dependence of all samples examined to date by this technique exhibits Arrhenius behavior (log(oxygen consumption) varies linearly with reciprocal absolute temperature), allowing the extrapolation of the temperature-dependent oxygen consumption to any storage temperature. EXPERIMENTAL SECTION Peanut oil, vegetable oil, corn oil, olive oil, cottonseed oil, safflower oil, and lard were procured from local supermarkets. Poultry fat and fish oil samples were supplied by Novus International (St. Louis, MO). Unless otherwise stated, a carrier gas containing 0.1% O2 (balance N2) was used (TRIGAS Industrial Gases, Lubbock, TX). Air was used in some experiments and was purified by passing through sequential columns packed with activated carbon, silica gel, and soda-lime. Other concentrations of oxygen were generated by in-line mixing of purified air with cylinder nitrogen. Mass flow controllers (model FC-280, Tylan General, Los Angeles, CA) were used for all flow measurements. All other chemicals and reagents used were reagent-grade or better and used without further purification. Peroxide values were determined by the standard iodometric procedure.23 The iodine absorption number (IAN) for all samples was determined according to the official method.13 Experimental Arrangement. The gas-phase FIA system is shown in Figure 1a. Carrier gas flows through a mass flow controller (MFC) and through an electropneumatically operated six-port stainless-steel HPLC type valve V (model 7000, Rheodyne, Cotati, CA). Further detail of the oxidation reactor (OR) incorporated in the loop of valve V is shown in Figure 1b. It is composed of a glass tube (150 × 16 mm o.d.) modified with an air inlet and outlet and a sample-injection tube (each 45 × 4 mm o.d.). Liquid oil/fat samples are inserted into the reactor through a poly(tetrafluoroethylene) (PTFE) tube connected to the inject port, equipped with a female Luer fitting at the other end. The air inlet and outlet tubes were cemented to 1.5-mm-o.d., 1-mm-i.d. Ni tubing with high-temperature epoxy (RP 4026, Ciba-Geigy). A heating tape (128 × 50 mm, 50 W at 110 V, Thermolyne, IA) was secured around the OR with glass-wool thread. A platinum resistance thermometric detector (RTD) was placed between the heating tape and the glass exterior of the OR. The reactor temperature was programmed with a temperature controller (CN8500, Omega Engineering, Stamford, CT). In Figure 1a, a cold trap is imposed between the valve V and the oxygen sensor (model 2550, Systech Instruments, IL). In most (23) American Oil Chemist’s Society Official Method Cd-8-53; American Oil Chemists Society: Champaign, IL, 1996.
1694 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999
Figure 1. (a) The gas-phase FIA system for the determination of oxygen consumption of oils and fats. (b) Detailed view of the oxidation reactor.
cases, the trap consisted of a 600-mL beaker filled with ice-water in which a coil of Ni tubing (1.6 mm o.d., 1.0 mm i.d., 400 mm long, coiled into a 32-mm-diameter coil) was immersed, followed by a “buffer” glass tube (8.6 mm i.d., 10.2 mm o.d., 180 mm long). The cold trap allows any oil vapor in the air stream to condense. Condensation of the oil in the sensor, itself, can greatly reduce useful sensor lifetime. The sensor output was acquired by a personal computer (Gateway 2000, P-75) equipped with a DAS1601 data acquisition board (Keithley-Metrabyte, Taunton, MA). All tubing used in the gas flow system are 1-mm-i.d. Ni tubes. A home-built digital timer was used to control the switching of V and, hence, the residence time of the oxygen-bearing carrier gas in the OR. To perform a stability determination, 12 mL of the sample is loaded into the reactor, the reactor is flushed by the carrier gas, heated to the intended temperature, and then switched to the isolated loop position. The loop is switched back in-line to measure the amount of oxygen consumed after a desired interval of time, as programmed by the valve timer. After a desired number of replicates (typically three measurements at a given temperature), the temperature is increased to the next desired value, and the process is repeated. The entire process is repeated for as many temperatures as desired. RESULTS AND DISCUSSION Performance of the Present Experimental System. System Output. A typical output of the present system is shown in Figure 2 for cottonseed oil as a function of temperature. As the sample is heated to higher temperatures, greater amounts of oxygen are consumed, and the magnitudes of the negative peaks increase. The reproducibility of measurements at each temperature is obviously very good. If long reactor residence times are used, oxygen intrusion from the outside through polymeric connecting conduits and fittings can have an effect on the results. Silicone oil (that cannot be oxidized at our experimental temperatures) is therefore used as a reference sample and oxygen consumption computed relative to this control in such cases. It should be noted that while Figure 2 shows data for triplicate determinations at six different temperatures, these data were
Figure 3. Oxygen-consumption signal as a function of oil sample taken. Figure 2. A typical readout of the gas-phase FIA system. Residence time of the oil sample in oxidation reactor: 5 min. Oxidation temperature: (1) 70, (2) 80, (3) 90, (4) 100, (5) 108, and (6) 120 °C.
obtained with the same sample, loaded once into the reactor. Each sample was analyzed, the reactor was flushed with new carrier gas, and the sample reanalyzed, for a total of three replicates. The temperature was then increased to the next desired value, and the process was repeated. Since so little of the sample is actually oxidized during an individual determination (vide infra), it is not necessary to introduce fresh sample aliquots. Dependence on Adjustable Parameters. Sample Volume. The present system is basically a surface reactor where only the exposed surface is available for oxidation. It is expected that as the reactor is filled, the exposed surface will become nearly constant, and hence, the oxygen-consumption signal will essentially become constant. This was confirmed using safflower oil as a sample. Different amounts were loaded into the oxidation reactor (100 °C, 40 mL/min carrier-gas flow, sample-residence time 9 min). The signal response reached a maximum and remained essentially constant when the sample volume exceeded ∼11 mL (Figure 3). There are several advantages of a surfacereactor arrangement (as opposed to a bubbled or vigorously stirred reactor). These include the ability to reuse the sample because of the small extent of oxidation and relative independence from sample size. Further, availability of oxygen during the course of the reaction does not become so limited as to affect the observed rate. In addition, in an agitated reactor, not only one must provide some means of agitation but also the conditions must be exactly reproduced. Carrier Gas Flow Rate. The experimental results for the effect of the carrier-gas flow rate (10-130 mL/min) on the signal peak height and peak area are shown in Figure 4. Other conditions are the same as in the previous paragraph. Decreasing the carriergas flow increases the sweeping time through the reactor but this variation is relatively small compared with the fixed reaction time of 9 min. The second effect is the greater dispersion induced by the faster flowing gas stream. The peak areas in units of mV‚s
Figure 4. Reciprocal of the flow rate vs oxygen consumption signals (peak height and peak area). Safflower oil: volume, 10 mL; 100 °C; residence time 9 min.
increase linearly with the reciprocal of the flow rate with a negligible y intercept. Therefore, if the area data were plotted in terms of V‚mL, the area counts would remain essentially constant. This suggests that dispersion is the dominant factor in the observed change in the signal heights. The net effect is that a lower carrier gas flow rate results in higher peak heights, but the peaks span a greater width in time. To achieve a compromise between analysis time and sensitivity, we chose 12-mL samples and a 30 mL/min carrier gas flow rate for further experiments. Determination of Oxygen Consumed at Different Residence Times. Figure 5 shows the relationship between the oxygen consumption and the residence time of the carrier gas in the oxidation reactor for some oil samples at 125 °C. The oxygen detector produces a signal that is linearly related to the amount of oxygen consumed. It is readily apparent that the ease with which different oils consume oxygen is quite different, fish oil being the most easily oxidized. Liquid-phase diffusion is slow, and Analytical Chemistry, Vol. 71, No. 9, May 1, 1999
1695
Figure 5. The oxygen consumption at different residence times for some samples (temperature, 125 °C; carrier gas flow rate, 35 mL/ min; sample volume, 11 mL).
Figure 6. Arrhenius plots for the oxygen consumption of some oils and fats. Residence time: 9 min. Table 1. Linear Regression Parameters for the Oxidation Experiments
the reaction here takes place only on the surface of the sample. So if the oxidation is allowed to proceed to a substantial degree, depletion of both the oxygen in the reactor and of the oxidizable material on the surface of the sample (which must be replaced by diffusion) can have an effect on the observed rate. These limitations do not apply when the extent of oxidation is limited, i.e., at short reaction times (or at lower reaction temperatures). The data in Figure 5 show that for reaction times of e10 min (except for fish oil) at this temperature, the observed oxygen consumption is essentially linear with time. At constant oxygen concentration (as will occur with the oxidation of a sample exposed to air), this suggests a zero-order oxidation rate, as observed by others.4,24 Determination of Oxygen Consumed at Different Temperatures. The experimental results of oxidative-stability measurements for oil and fat samples at different temperatures are shown in Figure 6 in the form of an Arrhenius plot. It is clear that all of the samples exhibit Arrhenius behavior. The corresponding numerical data, including the calculated activation energies, are presented in Table 1. It is immediately apparent that several lines in Figure 6 intersect each other. This points out the problems inherent in determining the oxidation rates at any given temperature (e.g., 100 °C) and relying on such results to predict relative stabilities at a different temperature (e.g., a storage temperature of 30 °C). The present experimental arrangement not only permits ready measurement of oxygen consumption at different temperatures, the correspondence to Arrhenius behavior also makes it possible to extrapolate such data to another temperature. Direct measurement of oxidation rates at typical storage temperatures will be prohibitively slow for many samples. As pointed out previously, the ability of the experimental system to provide meaningful data only holds if the overall extent (24) Wewala, A. R. In Natural Antioxidants. Chemistry, Health Effects and Applications; Shahidi, F., Ed.; AOCS Press: Champaign, IL, 1997; pp 331345.
1696 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999
samples
coefficients of the linear equations log(O2 Consumed) ) -1000 a/T + b (a, b (linear r2 value, n ) 15))
activation energy (kJ/mol)
olive oil vegetable oil lard safflower oil fish oil cottonseed oil poultry fat peanut oil corn oil
2.38 ( 0.09, 7.29 ( 0.25 (0.9939) 2.33 ( 0.07, 6.73 ( 0.2 (0.9943) 2.32 ( 0.02, 6.87 ( 0.06 (0.9996) 2.26 ( 0.05, 7.01 ( 0.1 (0.9982) 1.97 ( 0.055, 6.79 ( 0.2 (0.9974) 1.85 ( 0.06, 5.64 ( 0.15 (0.9966) 1.88 ( 0.05, 5.78 ( 0.1 (0.9966 1.76 ( 0.06, 5.22 ( 0.15 (0.9967) 1.74 ( 0.05, 5.21 ( 0.1 (0.9955)
E ) 45.6 ( 1.8 E ) 44.5 ( 1.4 E ) 44.4 ( 0.45 E ) 43.3 ( 0.91 E ) 37.6 ( 1.1 E ) 36.4 ( 1.2 E ) 36.0 ( 0.90 E ) 33.6 ( 1.1 E ) 33.3 ( 8.8
of oxidation is low. Corn oil and fish oil were selected for experiments over a wide temperature range with a reactorresidence time of 9 min. The results are shown in Figure 7. For corn oil, the data generated by the instrument corresponds to Arrhenius behavior over the entire temperature range studied, up to 175 °C, while for the very rapidly oxidized fish oil the linear range extends only to ∼95 °C. In this and other cases, the nonlinearity observed at high temperatures is a measurement artifact (due primarily to the depletion of the oxygen), and linearity extends to higher temperatures if the reaction time is drastically reduced. Activation Energies Determined Are Independent of Experimental Variables. It is important to note that the activation energies that are determined from the experimental data are independent of adjustable experimental parameters such as reactor residence times, oxygen concentration, etc. With residence times of 5, 7, and 9 min in the reactor, the activation energies for oxygen consumption by cottonseed oil were determined, for example, to be 36.4 ( 1.0, 36.4 ( 1.3, and 36.4 ( 0.9 kJ/mol, respectively. These are indistinguishable within experimental error. Similar results were also obtained with poultry fat (mean activation energy from three different residence times 36.2 ( 0.4 kJ/mol). We have conducted a limited number of experiments in which the carrier
Figure 7. Extended-temperature-range Arrhenius plot of oxygen consumption for corn and fish oils. Residence time: 9 min.
gas was bubbled through the hot sample in a continuous-flowthrough mode. Even in this case, the activation energies obtained were the same as those obtained with the static surface reactor. Experiments were also conducted for olive oil, safflower oil, and poultry fat at two different oxygen concentrations of the carrier gas, 0.1% and 5%. Within experimental error, the activation energies were identical despite the 50-fold change in oxygen concentration: olive oil, 45.6 ( 1.9 and 45.9 ( 1.1; safflower oil, 43.3 ( 0.9 and 43.3 ( 0.9; poultry fat, 36.0 ( 1.1 and 35.4 ( 0.8 kJ/mol at 0.1% and 5% oxygen, respectively. However, once again, with very easily oxidizable fish oil, so much of the sample gets oxidized with 5% oxygen that different results are obtained at higher oxygen concentration if other experimental parameters (e.g., reaction temperature range and tR) are not altered to reduce the extent of oxidation. Preference for Low Oxygen Concentration in the Carrier Gas. One of the attractive aspects of the present instrument is that multiple temperature runs with replicate measurements at each temperature can be conducted with a sample loaded into the reactor only once. This requires that the sample be altered as little as possible during the entire process. Lack of sample deterioration can be verified by measuring the PV of the sample before and after the measurement sequence. This is indeed observed in a typical experimental run. The PV of a cottonseedoil sample changed only by 2% before and after the experiments when the sample was put through our oxygen-consumption study over 75 °C to 115 °C using 0.1% O2 as a carrier gas. This included triplicate measurements, at 10 °C intervals, for a total of 15 measurements. Unsaturated linkages are oxidized to peroxides and destroyed during the oxidation process. These can be measured by the Iodine Absorption Number, IAN, vide infra. The original PV value for this sample was 3.88 mequiv/kg and was measured to be 3.96 mequiv/kg after the experiment. The IAN value was measured to be 81 both before and after the experiment (an IAN value of 81 corresponds to over 6 mol of unsaturation/ kg). A 2% change in PV is tantamount to an insignificant
consumption of the unsaturated linkages ( vegetable oil ∼ cottonseed oil > olive oil ∼safflower oil . fish oil. In contrast, the increase in PV shows no such correlation with the measurement of oxygen uptake. Indeed, the PV for the fishAnalytical Chemistry, Vol. 71, No. 9, May 1, 1999
1697
Figure 8. A plot of IAN decrease and PV increase versus relative oxygen consumption. Closed symbols: decrease in IAN, left ordinate. Open symbols: increase in PV, right ordinate.
oil sample actually decreases after the oxidation. Similar experiments were conducted with these oil samples for shorter oxidation periods of 4, 13, and 25 min, and similar results were obtained in all cases. The PV increase for peanut oil was the maximum and that for fish oil was the minimum. These results thus suggest that the measured PV increase under these conditions is a reflection on the stability of the specific type of hydroperoxide formed as well as on the stability of the oil. This conclusion is supported when the profile of PV as a function of time for an experiment involving aeration at an elevated temperature is examined in detail for different oils and fats. For example, while the general pattern of ascending and descending PV resembles a Gaussian curve, the maximum of the PV curve obtained with fish oil reaches only a few milliequivalents per kilogram whereas the maximum for poultry fat under the same conditions can reach 250-300 mequiv/kg. The presently proposed method eliminates these confounding issues in the interpretation of PV analyses. CONCLUSIONS A new, rapid method has been developed which can accurately measure the oxidative stability of an oil or fat. An important
1698 Analytical Chemistry, Vol. 71, No. 9, May 1, 1999
advantage of the method developed is the ability to predict the stability of samples at ambient storage temperatures that may be very difficult to be done experimentally. This will be very useful in practice. The proposed approach is much faster than conventional methods used for the assessment of lipid stability. Only 2-3 h are needed for the stability determination of a typical oil/fat sample. The instrument described is simple, easy to automate, and yields reproducible results. ACKNOWLEDGMENT This research was supported by Novus International, St. Louis, MO. We gratefully acknowledge the help and guidance provided by Dr. William D. Shermer during many phases of this investigation.
Received for review December 8, 1998. Accepted February 17, 1999. AC981365T
