Influence of Oil Types and Prolonged Frying Time on the Volatile Compounds and Sensory Properties of French Fries

: In order to study the flavor of French fries (FFs) prepared in different frying oils, we identified and compared the volatiles of FFs fried in high-oleic sunflower oil (HSO), sunflower oil (SO), linseed oil (LO), and palm oil (PO) during prolonged 24 h frying time. 47 different kinds of volatiles were presented, and aldehydes were the most abundant compounds. The FFs prepared in SO were rich in alkadienals, especially the ( E , E )-2,4-decadienal, thus inducing the highest deep-fried odor. The content of alkenals was higher in FFs prepared in HSO, among which ( E )-2-nonenal and 2-undecenal provided the undesirable oily flavor. Whereas, FFs prepared in PO were rich in alkanals, and showed an undesirable green aroma because of hexanal. Besides, the aldehydes in FFs fried in LO were the least with more undesirable flavor substances (e.g. ( E , E )-2,4-heptadienal). In addition, except for the FFs fried in LO, the aldehydes in other FFs showed an increasing trend. While, the volatiles from the Maillard reaction (e.g. pyrazines) showed no clear pattern. Meanwhile, frying process had optimum frying window (approximately 12 h with total polar compounds content of 14.5%-22.2% in different oils), and the French fries prepared in this period obtained higher flavor score. Therefore, the comparison related to volatiles of FFs provided a basis for the flavor control to a certain extent.

cellent source of carbohydrates and protein of high relative biological value while with a little fat 6 . Numerous researchers paid more attention to explore the volatile compounds of frying oils after frying FFs under various conditions 7 11 . However, considering the consumers more important concern such as the flavor and nutritional quality of food, more attention should be paid to the flavor compounds present in FFs. The flavor formation of fried food is related to many factors, including the type of oils, the fatty acid composition of oils, and the temperature and time of frying 4 . Multari et al. studied the volatiles of six oils after frying to evaluate the suitability of these oils. But the study did not take the sensory properties into consideration 12 . Dobarganes et al. conducted frying experiment of FFs at 175 in three oils to study the changes of oils and FFs and proved that the frying oil and fried food existed with the interaction reactions, therefore the flavor of frying oil would have significant effect on the fried food 13 . Besides, the increased frying time resulted in the degradation of oils and act on the volatile aldehydes of FFs, thus making a difference in the flavor desirability scores of FFs 7 . Nonetheless, the relationship between the changes involved in total volatile compounds and sensory properties of FFs prepared in oils with different fatty acid composition were uncertain. Therefore, the influence of oil types and prolonged frying time of oils on the sensory properties of FFs need further investigation.
Many kinds of vegetable oil are available for deep-frying considering their availability and thermal resistance. Sunflower oil is generally accepted worldwide. Recently, higholeic sunflower oil become very popular because monounsaturated fatty acids can provide a broader spectrum of functionality and health benefits with an inherent higher thermal oxidation stability than polyunsaturated fatty acids 14 . Besides, linseed oil attracts people s attention due to the high content of Omega 3 fatty acids, which is beneficial to health. Relatively, palm oil is widely used in the manufacturing of different type of food on an industrial scale, as well as in fast food restaurants. In this study, HSO, SO, LO and PO were selected to prepare the FFs during prolonged frying period. The headspace solid-phase microextraction HS-SPME and gas chromatography-mass spectrometry GC-MS were used to analyze the changes in volatile substances of FFs in different oils and frying time. The hierarchical cluster analysis was applied to obtain the qualitative discrimination of FFs prepared in different frying oils and frying periods. Overall, this study on the volatiles and sensory evaluation of FFs will provide a significant reference for industrial production of more delicious FFs.

Materials and chemicals
HSO, SO, LO and PO were obtained from Riqing Food Co., Ltd., Yihai Grain & Oil Industry Co. Ltd., Beijing hongjingyuan Trading Co., Ltd. and Cargill investment Co., Ltd China , respectively, and all without antioxidant. Fresh potatoes were purchased from a farm in Shandong Province, China. The reagents and flavor standards were purchased from Sinopharm Chemical Regent Co., Ltd Shanghai, China and Sigma St. Louis, MO, USA , respectively.

Sample preparation
Upon washing and peeling of potato, it was cut into strips of 1 cm 1 cm 6 cm, soaked in water before frying, and then blotted dry with filter paper. A commercial deep fryer 2.5 L, Aigoli, China was used for frying. Then, 2 kg frying oil was poured into the deep fryer, and heated to 170 2 within 15 min. 100 2 g of potatoes was fried in HSO, SO, LO, and PO for 4 min, then wait for the next frying cycle for 25

Determination of fatty acid composition
The fatty acid composition was analyzed by gas chromatography using Agilent 7820, equipped with a FID and a TRACE TM TR-FAME Column 60 m 0.25 μm, Thermo Fisher . And the conditions were referred to our previous research 15 . The injection block temperature was set at 250 . The oven temperature was firstly kept at 60 for 3 min, then set as follows, 60 175 at 5 /min, 175 for 15 min, 175 220 at 2 /min, and finally 220 for 10 min. The carrier gas was nitrogen with a flow rate of 25 mL/min and the split rate was 1/100.

Determination of AV, PV, and TPC
AV and PV were determined according to AOCS official method 3a-63 16 , Cd 8b-90 17 , respectively. The determination of TPC was based on the AOCS official method Cd 20-91 with minor modifications 18 .

Determination moisture and oil content
The moisture content of FFs, expressed as g/100 g wet basis, and the oil content expressed as g/100 g dry basis, were determined according to Santos et al. 6 .

Determination of volatile compounds
The SPME method was referred to Santos et al. 6 with minor modification. Volatile extraction from the samples was carried out using a 2 cm-50/30 μm divinylbenzene/carboxen/polydimethylsiloxane DVB/CAR/PDMS Stable Flex solid-phase microextraction SPME fiber Supelco Inc., Bellefonte, PA, USA . Before extraction, about 1.5 0.02 g of sample, 5 mL of saline solution, and 50 μL of 1,2-dichlorobenzene 59 μg/mL , as an internal standard solution, was put into a 20 mL vial. And in the incubator, the sample vials were equilibrated at 60 , 500 rpm for 30 min. Then the fiber was inserted into the GC/MS injection port at 250 to desorb for 5 min.
The GC/MS analysis was performed with the GC/MS apparatus using a TSQ Quantum XLS Thermo, USA operating in electron ionization mode EI, 70 eV . The DB-WAX 30 m 0.25 mm i.d.; 0.25 μm film thickness column was set from 45 for 2 min to 180 at a rate of 3 /min then to 240 at 10 /min for 5 min . The quantification of each compound was calculated by the internal standard method.

Sensory evaluation
For aroma profile analysis of different FFs, the aroma intensities were rated from 0 not perceivable to 3 strongly perceivable by the sensory panel, which was consisted of 12 panelists from Jiangnan University. A reference solution in refined high-oleic sunflower oil was provided for each aroma: E, E -2,4-decadienal for deep-fried, 3-methylbutanal for malty, E -2-nonenal for fatty, methional for cooked potato-like and 2-ethyl-3,5-dimethylpyrazine for roasty.

Statistical analyses
All analyses were conducted in triplicate. The relevant results were denoted as mean standard deviation SD . The heatmap was conducted by the Metabo Analyst 3.0 for unsupervised clustering.

3.1
The change of fatty acids of four kinds of vegetable oils As shown in Table 1 Table 1 The change of fatty acids of four frying oils in different frying time.
to 15.97 . The differences were also apparent in linolenic acid, as it decreased in HSO, SO, and LO from 51.24 to 48.58 , and increased in PO.

3.2
The change of TPC, PV and AV of four kinds of oils TPC, PV, and AV were detected for the determination of the physical-chemical properties of oils. As shown in Fig.  1a, the initial TPC of LO was 10 higher than that of the other oils. Further, it increased rapidly with time, reaching up to 25.50 after 12 h frying. The TPC of fresh SO was about 7 , slightly lower than that of fresh PO 8 . However, it increased rapidly the next time then exceeded that of the PO at 4 h. In addition, the TPC of SO and PO exceeded the upper limit 27 , reaching the maximum value of 32 and 28 at 24 h, respectively. The level and growth rate of TPC of HSO 4 were both the lowest among the four kinds of oils, and the maximum 15 was far below the upper limit of the Chinese national standard 27 which indicated greater frying oil stability. These results were consistent with the previous study reported by Warner et al. 20 .
The initial PV of LO 15.48 meq/kg was the highest among the four kinds of oils, which was even higher than that of the other three oils at 24 h, and reached 34.36 meq/ kg at the end of frying time. While the initial and final PV of SO were 7.66 and 15.40 meq/kg, respectively. The initial PV of PO was 3.86 meq/kg and then increased up to 14.60 meq/kg at 24 h. The PV of HSO was the lowest, increasing from 2.14 meq/kg to 9.22 meq/kg Fig. 1b .
The AV of LO was low even at 12 h 1.61 mg/g , which was far from the limit 5 mg/g, from the Chinese national standard . The initial AV of HSO 1.02 mg/g was higher than that of SO 0.74 mg/g , but after 24 h frying, it reached up to 3.79 mg/g, compared with 4.01 mg/g in SO. For PO, its AV was the highest reaching up to 7.10 mg/g at 12 h, exceeding the Chinese national standard Fig. 1c .

The moisture and oil content of four kinds of FFs
As shown in Table 2, the moisture content of potatoes decreased significantly approximately 25 after frying. While the fat absorption was accompanied by moisture reduction, with the final oil contents ranging from 22 Table 1 Continued.  33.14 . The moisture and oil content of FFs fried in the four kinds of oils fluctuated within 24 h frying time, which was similarly observed in the previous study 6 . From the average level, the oil content of FFs fried in HSO was slightly lower, while that in PO, SO and LO was all slightly higher. The moisture content, that in FFs prepared in PO was also slightly higher, compared with that in HSO, SO, and in LO.

Generation of Volatile Compounds of French Fries During Prolonged Deep-Frying
3.4 Analysis of volatile compounds of FFs prepared in four kinds of oils 3.4.1 Heat map analysis Figure 2 A clustering heat map study was used to distinguish the FFs fried in different oils over frying time and 7 similar groups were observed. However, more specifically, the FFs fried in LO and SO were distinguished well from the other FFs. In addition, raw potatoes and FFs fried in HSO 1-4 h were also separated from others. However, FFs fried in PO and HSO 8-24 h particularly had a certain consistency. Besides, the higher E, E -2,4-decadienal and E, Z -2,4-decadienal typical deep-fried odor were observed in FFs fried in SO at 18 h and 24 h. Further, the hexanoic acid, octanoic acid, and nonanoic acid increased in FFs fried in HSO at 24 h, causing stronger undesirable flavor. Furthermore, Z, Z -2,4-heptadienal and E, E -2,4-heptadienal with fatty aroma was excellent in FFs prepared in LO.

Analysis of volatile compounds
47 kinds of volatile compounds related to FFs prepared in four kinds of vegetable oils are shown in Table 3, including aldehydes 20 , acids 7 , nitrogen-containing heterocycles NCHs, 6 , alcohols 5 , hydrocarbons 4 , ketones 2 , oxygen-containing heterocycles OCHs, 2 , sulfur-containing volatiles SCVs, 1 . Besides, data was analyzed more comprehensively in order to explain the reason for sample clustering, and to reveal the differences that existed between the volatile compounds in different FFs and their related changes. As shown in Fig. 3, aldehydes were the main volatiles, a protagonist to decide the FFs flavor, followed by SCVs and NCHs. Additionally, although the relative concentration of OCHs, acids and ketones were lower, they contributed significantly. In comparison, alcohols and hydrocarbons with a moderate level of content contributed little to flavor.

Aldehydes
The volatile aldehydes identified in FFs were composed of alkanals, alkenals, alkadienals, and substituted aldehydes. As shown in Fig. 3, the accumulation of aldehydes occurred during frying time. Moreover, aldehydes from thermal degradation and hydrolysis of triacylglycerols under frying conditions were the most abundant volatiles 21 . In general, previous studies have shown that alkanals were positively correlated with oleic acid content. Whereas, alkenals were positively correlated with the content of linoleic acid and linolenic acid. As shown in Fig.  3, the total relative content of aldehydes with flavor contribution detected in FFs prepared in LO was less than that in the other three FFs. These findings may be attributed to      the relatively lower content of oleic acid and linoleic acid. Alkanals: Hexanal, with grass and green aroma, increased apparently especially in FFs prepared in PO, reaching the maximum 1484.70 ng/g at 24 h. This was about 16 times higher than that of raw potato 88.25 ng/g and about 2 times of the other FFs. Moreover, we detected the remaining alkanals, such as heptanal, octanal, nonanal and decanal. These were also slightly higher in FFs prepared in PO. The relative content of alkanals was generally the highest in FFs prepared in PO. Nevertheless, these results were consistent with the previous study carried out by Zhang et al. 7 .
Alkenals: Alkenals were mainly existed with trans-forms with ten carbon or less, including E -2-octenal, E -2nonenal and E -2-decenal all with fatty aroma. Among them, E -2-octenal was higher in FFs prepared in PO with a maximum value of 22.32 ng/g at 24 h. Whereas, E -2nonenal and E -2-decenal produced by homolytic β-scission of 9-hydroperoxy oleic acid exhibited the relatively higher levels in FFs prepared in HSO, with maximum values of 111.60 ng/g and 1112.39 ng/g at 24 h, respectively. However, the E -2-octenal showed an initial increasing trend and then decreased subsequently, which cannot be fully explained by the oxidation of unsaturated acid. In fact, previous studies have also regarded E -2-octenal as the oxidative decomposition product of 3-nonenal from the β-homolysis of L-9-hydroperoxide 6 . From the perspective of cis forms, it was composed of Z -2-heptenal, Z -2decenal and 2-undecenal. The content of Z -2-heptenal pungent and penetrating odor increased with time in FFs fried in HSO, SO, and PO, and then became stable at certain level. The study carried by Warner et al. 22 . has reported the generation of Z -2-heptenal from linoleic acid and linolenic acid. Moreover, the content of 2-undecenal fatty, green, and soup aroma was highest in FFs prepared in HSO and showed an obvious increasing trend to 856.95 ng/g at 24 h , which was consistent with the conclusion that 2-undecenal was produced by homolytic β-scission of 8-hydroperoxy oleic acid 23 .
Alkadienals: E, E -2,4-decadienal was the main component observed during frying as a deep-fried aroma. The relative content of E, E -2,4-decadienal was significantly higher in FFs prepared in SO, high linoleic acid content as it was the oxidation product of linoleic acid 20 . Moreover, even the minimum 631.25 ng/g at 1 h was 6 7 times larger than the other three FFs. In addition, E, Z -2,4-decadienal deep-fried flavor also showed a higher content in FFs prepared in SO, reaching 56.00 ng/g after 18 h frying. Warner et al. also reported that 2,4-decadienal levels had the positive correlation with intensity of friedfood flavor of FFs and the flavor score of FFs was higher prepared in cottonseed oil with higher content of linoleic acid compared with high-oleic sunflower oil 20 . In addition, Z, Z -2,4-heptadienal and E, E -2,4-heptadienal existed simultaneously in the frying system. The E, E -2,4-heptadienal fatty, oily, and rancid was the degradative product of hydroperoxides from oxidation of linoleic acid 24,25 . It increased with time in all four kinds of FFs and was observed higher in FFs 220.89 ng/g at 24 h prepared in LO rich in linolenic acid than the other FFs. The same trend was observed in the Z, Z -2,4-heptadienal. While E, E -2,4-nonadienal fatty, green, and waxy aroma was higher in FFs 66.35 ng/g at 18 h prepared in SO, which was produced by the cracking of the linoleic acid. Substituted aldehydes: The Z -4,5-epoxy-E -2-decenal and E -4,5-epoxy-E -2-decenal typically metallic odor accumulating with time, were produced from the oxidative degradation of linoleic acid, causing the undesirable flavor to FFs 26 . Moreover, the content of benzaldehyde fruity and woody aroma increased in FFs fried in HSO but decreased moderately in the other three FFs. It was considered as the decomposition product of linoleic acid 27 . It was worth noting that 4-oxononanal, produced by long-term frying of food, was considered to be potentially toxic and pathogenic 28 . It was perceived that 4-oxononanal was not detected in FFs prepared in LO, which may be due to the shorter heating time. Meanwhile, its relative content in the other three FFs increased significantly with the maximum of 3.17 ng/g at 24 h in FFs prepared in SO. From the perspective of the possibility of inducing toxicological effects, these compounds should be paid special attention.
Pyrazines were kinds of volatile substances in many processed foods, such as the fermentation and drying process of cocoa beans and barbecue meat products 4 . They are produced almost from the Maillard reaction one of the important chemical reactions in the frying process 29,30 , and are the important contributor of roasty flavor in FFs. During our study, six types of pyrazines were detected, among which ethylpyrazine roasted nut flavor and popcorn odor was the most common, whereas, 3-ethyl-2,5-dimethylpyrazine was one of the main substances in FFs that presented a roasty aroma.
Our results showed that the content of pyrazines in each FF did not change significantly with frying time. Moreover, no significant difference was observed in four kinds of FFs. The main reason may be that pyrazines were mainly produced by the Maillard reaction, and had almost no relationship with the fatty acid composition of oils.

Ketones and Sulfur-containing volatiles
Ketones and SCVs detected in FFs fried in four kinds of oils included 1-octen-3-one, E -6,10-dimethyl-5,9-undecane-2-one, and methional. The 1-octen-3-one was generated from arachidonic acid, which recovered with a noticeable mushroom aroma to FFs. However, the relative content of it did not show any clear pattern. Methional generated from the Strecker degradation of methionine recovered with a clear flavor of cooked potato-liked, which was one of the most important flavors related to FFs. The relative content of methional in FFs prepared in LO was higher than that in the other three FFs, but the change was not obvious with time.
Oxygen-containing heterocycles Two kinds of OCHs were detected in the four kinds of FFs including, 5-butyldihydro-2 3H -furanone and 2-pentylfuran Fig. 3 . Among which 2-pentylfuran was an important flavor compound. Moreover, 2-pentylfuran butter, fruity, green and bean aroma was mainly the oxidative decomposition product of linoleic acid, which often appear during the process of oil heating. At the same time, 2-pentylfuran may also be the result of cyclization of 1,3,6-nonotrienyl radical and hydroxyl radical from the decomposition of linolenic acid 31 . In this study, the relative content of 2-pentylfuran in FFs prepared in SO rich in linoleic acid was slightly higher, and reached the highest value at 18 h 161.86 ng/g . In addition, 5-butyldihydro-2 3H -furanone was one of the types of furan derivative, which probably generated from thermal decomposition of carbohydrates, or Maillard reaction between amino acids and reducing sugars 11 . Acids During our study, we detected volatile acids, and 7 of them were main saturated alkyl acids recovered, i.e. hexanoic acid, octanoic acid, nonanoic acid, tetradecanoic acid, pentadecanoic acid, n-hexadecanoic acid, and octadecanoic acid. Among them, octanoic acid and nonanoic acid were important due to their rancid flavor, which showed a negative impact on the flavor of FFs 4 . Besides, octanoic acid and nonanoic acid showed a moderate increasing trend in FFs with frying time, especially, in FFs fried in HSO, corresponding to the stronger rancid aroma of HSO. Moreover, the hexanoic acid presented the higher contents compared with other acids in the FFs samples which was a key aroma compound contributed to the off-flavor of FFs. The hexanoic acid may result from the secondary decomposition of hexanal and 2,4-decadienal 32 , therefore, the decrease of these precursors may lead to the increase of hexanoic acid in the end of frying time. Some studies also indicated that these short-chain fatty acids are generally produced from the oxidative decomposition of unsaturated fatty acids such as oleic acid and linoleic acid in oil 33 . At the same time, saturated fatty acids were oxidized by hydrogen peroxides to produce saturated aldehydes, which were oxidized at high temperature to form saturated monoacids with reduced carbon number 8 .

Other volatiles
Apart from the above mentioned several types of substances, the other common volatiles detected were hydrocarbons and alcohols.
Hydrocarbons identified in four types of FFs were as follows: Z -3-3-methyl-1-butenyl cyclohexane, E, E -4-ethyl-3-nonen-5-yne and 3-ethyl-2-methyl-1,3-hexadiene. Generally, hydrocarbons are produced by the free radical reaction of fatty acid decomposition products. At the same time, they are affected by other factors, which ultimately result in cis-trans isomerization, carbon-carbon double bond C C position isomerization of intermediate products. This further forms a wide variety of hydrocarbons consisting of various alkanes, alkenes, alkynes, etc.
However, in general, the relative contents of hydrocarbons and alcohols which were not the main contributor to the flavor in the four kinds of FFs, did not change significantly with time and most of them fluctuated at a certain level Fig. 3 .

Sensory evaluation
The data shown in Figs. 4a-4d illustrates the sensory changes of the same oil over the 12 h of deep-frying, and Figs. 4e-4h revealed the comparison between four kinds of FFs in the same frying time.
The extended frying time of oils caused noticeable difference in sensory profiles of FFs. Among all FFs prepared in four kinds oils, the sensory score of deep-fried aroma intensity presented maximum at 12 h, which was due to the increased content of E, E -2,4-decadienal. Then, as the oil deteriorated, the fatty score also increased obviously owing to the increasing fatty aroma from E -2-nonenal and 2-undecenal. Moreover, the malty and roasty aroma decreased moderately, while the cooked potato-like aroma increased with frying time.
A significant difference in the intensity of deep-fried aroma was observed with the highest score for FFs prepared in SO than that in HSO, LO and PO. This was consistent with the high content of E, E -2,4-decadienal and E, Z -2,4-decadienal deep-fried aroma , generating from linoleic acid. The fatty aroma, mainly from E -2-nonenal the product of homolytic β-scission of oleic acid was stronger in FFs prepared in HSO rich in oleic acid . While, the cooked potato-like aroma, generating from the Strecker aldehydes main methionine in FFs and the roasty aroma produced by pyrazines from Maillard reactions, suggested similar profiles in all FFs, which were consistent with the study conducted by Thurer et al. 9 Moreover, FFs prepared in PO observed with a higher level of malty aroma, which also contributed to the desirable flavor of FFs. In summary, the FFs existed with the highest sensory score when prepared in SO at the 12 h than that in HSO, LO and PO, due to more E, E -and E, Z -2,4-decadienal from linoleic acid. This result was actually different from the conclusion from Petersen et al. that high-oleic edible oils had higher sensory evaluation and several researches got some different conclusion 35 37 . Therefore, whether high-oleic oil can provide better flavor for fried food in the frying process is worth further discussion. In addition, FFs prepared in HSO presented with higher fatty flavor due to high levels of E -2-nonenal and 2-undecenal. Further, the green aroma was more prominent in FFs prepared in PO. In general, the type of oil caused significant differences in the sensory score of FFs samples, which was consistent with the results conducted by Blanca et al. 38 .

Conclusion
The type of frying oils and frying time played an important role in the formation of volatile compounds during the preparation of FFs. The result revealed that FFs prepared in HSO, SO, LO and PO were mostly comparable. Aldehydes were the main volatiles in FFs prepared in the four kinds of oils. And E, E -2,4-decadienal was the dominant aroma-active compounds in the FFs fried in SO, thus leading to higher levels of deep-fried odor compared with FFs fried in other oils. While FFs prepared in HSO and PO were enriched with few undesirable volatile aldehydes i.e. 2-undecenal, 2-undecenal and hexanal compared with that in the SO. Further, the desirable odor of FFs prepared in LO dropped the fastest due to the rapid degradation of oil. Although the present study proposes phenomena and reasons for the formation and changes of volatile compounds in FFs, further research is needed to provide a basis for the improvement and control of the flavor of FFs.