3.1. Optimization of GC-MS condition
In this study, we have investigated the suitability of HP-5 and HP-INNOWAX for the analysis of 19 FAs by the comparison of peak shape, resolution and sensitivity. The results showed that the peaks of several alcohol analytes were tailed when analyzed with HP-5 column. What’s more, citronellol, cis-citral and geraniol with similar retention times and same quantitative ion could not be separated. By contrast, when HP-INNOWAX column was utilized, all the FAs showed sharp and symmetrical peaks, and could be identified by retention times or characteristic ions. Therefore, the HP-INNOWAX column was chosen in subsequent analysis.
The injection port temperature would affect the desorption efficiency of the analytes. Considering the limits of operating temperature of SPME fiber coating and chromatographic column, the injection port was held at 270 oC to shorten the desorption time and improve the speed of analysis. Under the optimal GC-MS conditions, the total ion chromatogram (TIC) of 19 FA standards and 2 internal standards was shown in Fig. 1, in which most analytes obtained good separation except limonene with a tailed peak.
3.2. Optimization of sample pretreatment conditions
The factors affecting the extraction efficiency of SPME include the coating and thickness of the SPME fiber, equilibration temperature, equilibration time, extraction method, extraction temperature, extraction time, desorption temperature, desorption time and so on. Based on the published literatures [5, 6, 16], the HS-SPME method was selected as the extraction method. The method is simple and environment friendly, which can not only ensure the detection sensitivity of 19 FAs with boiling point below 300 oC, but also avoid the interference of impurities with high boiling points. In the subsequent optimization experiments, 270 oC and 4 min were selected as the desorption temperature and desorption time, separately. And 10 µL of the mixed standard solution with each FA of 10 µg/mL was spiked into the blank sample for optimization of sample pretreatment conditions. The optimal conditions were evaluated in terms of the total peak areas.
3.2.1. Optimization of SPME extraction fibers
The coating polarity and the thickness of the SPME fiber have a great influence on the extraction efficiency and equilibration time. Since all the analytes are volatile or semi-volatile organic compounds, and most of them are medium or strong- polar compounds, the extraction efficiencies of five SPME fibers (85 µm PA, 100 µm PDMS, 65 µm PDMS/DVB, 75 µm CAR/PDMS and 50/30 µm DVB/CAR/PDMS) for 19 analytes were investigated. The extraction conditions were fixed as follows: the samples were extracted by SPME at 80 oC for 20 min and desorption at 270 oC for 4 min. The results of extraction with five fibers were shown in Figure S1. The total peak areas of CAR/PDMS, PDMS/DVB and DVB/CAR/PDMS were similar, while the peak areas of other two fibers were significantly lower than those of these three. However, for furfuryl alcohol and cinnamyl alcohol, the peak area extracted with DVB/CAR/PDMS was less than 60,000, which indicated that the high LODs of these two FAs would be obtained with this fiber. Compared with CAR/PDMS, the PDMS/DVB fiber showed better extraction efficiency for citral, geraniol, hydroxy citronellal, eugenol, isoeugenol and amylcinnamyl alcohol, but an unsatisfied extraction for furfuryl alcohol. The CAR/PDMS fiber had the best extraction efficiency for limonene, furfuryl alcohol, benzyl acetate, benzyl alcohol, phenethyl alcohol, cinnamaldehyde and 7-methylcoumarin, and also good for other analytes. Therefore, the fiber of 75 µm CAR/PDMS was selected for subsequent experiments after comprehensive consideration.
3.2.2. Optimization of extraction temperature, equilibration time and extraction time by CCD
In the procedure of HS-SPME, lower extraction temperature would lead to longer extraction time, and also affect the adsorption efficiency of the fiber. However, if the extraction temperature is too high, the extraction efficiency of the analytes with low boiling points would be reduced. The extraction efficiency of HS-SPME was determined by the distribution of the analytes in the sample matrix, HS air and the coating of the fiber. Equilibration time, extraction temperature and extraction time together affect the dynamic equilibrium of HS-SPME, and the three factors might interact with each other [8]. Therefore, an experimental design of CCD was adopted to explore the effect of the three factors on the extraction efficiency and their interactions, so as to obtain a suitable model and further determine the optimal conditions. The optimized experiment design of extraction temperature, equilibration time and extraction time generated by Minitab and the total peak areas of 19 FAs in each experiment was presented in Table S2.
The data analysis of CCD would be elaborated by taking limonene as an example. The suitability of the model was verified by model analysis, lack-of-fit tests and R2 test. First, the relative magnitude and statistical significance of the main effect, the square effect and the interaction effect could be compared by the Pareto chart, and the factors that contribute significantly to the peak area (Y) could be screened. The reference line was set with the confidence interval of 95%. If the main effects or interactions exceed the reference line, it is considered to have a significant effect, and the longer the length, the greater the effect. As shown in the Pareto chart of limonene (Fig. 2), the significant parameters included A, AA, AC, AB, BB and C. What’s more, ANOVA F-test was performed to evaluate whether the correlation between the peak areas and the parameters was statistically significant. The p of this factor was compared to the significance level ɑ=0.05 to test the null hypothesis. The results showed the p of the six factors mentioned above were all less than 0.05, which indicated that the correlation between the extraction efficiency and these six factors was statistically significant.
Then, the model was simplified to improve the prediction accuracy by eliminating the factors without statistical significance. Since both AB and BB were statistically significant, the linear term equilibration time (B) was also taken into consideration when simplifying the model. The model was established by the following equation:
Y = 367760 + 177171A − 49595B + 117698C − 1273A2 + 7531B2- 3607AB − 2408AC + 4423BC
Besides the statistical significance of various factors in ANOVA, the significance of the model and the insignificance of lack-of-fit are also important parameters to evaluate the reasonableness of RSM[13]. On the premise that all the data conformed to normal distribution, the p value of the ANOVA F-test of the simplified model was less than 0.01, which indicated that the model was significant. And the F of lack-of-fit tests was more than 0.05 (0.392), indicating that the model could explain all the data. The R2 and the predict coefficient of determination (Rpre2) could evaluate fitness of the model and the data. Rpre2 could assess the ability of model to predict the response for new values of the factors. A larger R2 value indicated a better fit of the model, and a larger Rpre2 value indicated a better predictive ability of the model. For the model of limonene, R2 value was 97.18% and Rpre2 value was 86.78%. The results showed that the model was well simulated and all the response variables could be explained accurately.
The 3D response surface can intuitively show the influence of two factors on the peak area. The greater the slope of the surface, the more significant the influence of the factors on the response. Figure S2 showed that when the extraction time was 20 min, the peak area of limonene first increased and then decreased with the increase of extraction temperature and equilibration time. Figure S3 showed that when the equilibration time was 10 min, the response value of limonene first increased and then decreased with the increase of extraction temperature and extraction time. Figure S4 showed that when the extraction temperature was 80 oC, the peak area of limonene first decreased and then increased with the increase of equilibration time and extraction time. The response value could reach the maximum when both the equilibration time and the extraction time were the lowest.
Finally, the optimum conditions for limonene were obtained by Minitab were as follows: extraction temperature 40 oC, equilibration time 15 min and extraction time 30 min.
The results of the RSM for all the analytes exhibited that: (1) The extraction temperature had a significant impact on extraction efficiency of each analyte. Five analytes with lower boiling points include limonene, linalool, furfuryl alcohol, diethyl maleate and benzyl acetate showed better extraction efficiencies with low extraction temperature. However, other 14 analytes preferred high extraction temperature to obtain satisfied extraction efficiencies. Besides, cinnamyl alcohol and amylcinnamyl alcohol had relatively low response when compared with other target FAs. Thus, a higher extraction temperature was beneficial for the extraction of most target FAs and 80 oC was finally selected as the extraction temperature. (2) The Pareto chart showed that equilibration time had a significant effect on the response of 10 analytes. Among them, the optimal equilibration time for citronellol was 5 min, the optimal equilibration time for eugenol, cinnamyl alcohol and iso-eugenol was 12 min, and the optimal equilibration time for limonene, citral, hydroxycitronellol, jasmonal, amyl cinnamol was 15 min. Considering the total analytical time and manual injection of SPME, 12 min was selected as the final equilibration time. (3) The extraction time had a significant effect on all the 19 analytes, and the peak area showed an upward trend with the increase of extraction time. This might be due to the insufficient range of extraction time in this design, so a single-factor optimization experiment of extraction time was further performed to achieve the optimal extraction time.
3.2.3 Single-factor optimization experiment of the extraction time
The volatile compounds could be rapidly absorbed by the fiber at the beginning of extraction procedure, and then the speed of absorption would gradually slow down. If the extraction time increases continuously after reaching equilibrium, the adsorption efficiency might remain unchanged or even decrease. The response surface showed that within the previous range of time (10–30 min), the peak areas were increased with the increase of extraction time. Therefore, the range of extraction time was expanded for further optimization, which were set at 10 min, 30 min, 40 min and 60 min, respectively. With other conditions unchanged, the extraction efficiencies with different extraction time were compared. The experimental results (Figure S5) showed that the peak areas of limonene, linalool, furfuryl alcohol, benzyl acetate, diethyl maleate, citronellol and hydroxy citronellal were the highest with the extraction time of 30 min, and the peak area decreased with further increase of extraction time. When the time increased from 30 to 60 min, the peak areas of citral, citronellol, benzyl alcohol, phenethyl alcohol, eugenol and carvacrol did not change significantly. Six compounds with higher boiling points including cinnamaldehyde, amyl cinnamaldehyde, cinnamyl alcohol, isoeugenol, amylcinnamyl alcohol and 7-methylcoumarin had the highest response when the extraction time was 60 min. For all the 19 analytes, the total peak area reached the highest with the extraction time of 30 min (57841515). The result indicated that the extraction time over 30 min was not beneficial for better extraction efficiency, which might be due to the competition among the analytes for active sites of the fiber as the absorption of compounds on SPME fiber is a dynamic process. In the process of water bath, the FAs with low boiling points volatilized quickly and could occupied the active sites on the fiber more rapidly so that their extraction effect was better in the first ten minutes of extraction. As the time of extraction increased, the concentration of FAs with higher boiling points on the fiber gradually increased, and their competitiveness for the active sites was improved. They could occupy the active sites, and even replace FAs with low boiling points that has been adsorbed on the fiber. Therefore, with the increase of extraction time, the extraction efficiency of the FAs with high boiling points became higher, while the extraction efficiency of FAs with low boiling points was unchanged or lower. Considering the extraction efficiencies of most target analytes and a relatively short analytic time, 30 min was selected as the extraction time.
According to the results of RSM and single-factor optimization of extraction time, the extraction temperature of 80 ℃, equilibration time of 12 min and extraction time of 30 min were selected as the final extraction condition. The verified experiment of the optimal condition was carried out, and the results of the verified experiments were all within the 95% prediction interval of the model, indicating the fitness of the predicted and the experimental results.
3.4 Application of the proposed method
The established HS-SPME-GC-MS method was applied for the determination of FAs in various paper personal care products, including 2 kinds of baby wipes, 9 kinds of wet tissues, 3 kinds of wet toilet papers, 3 kinds of makeup remover cleansing wipes, 1 kind of baby diaper and 2 kinds of sanitary napkins. The detection rates and content ranges of 19 FAs were shown in Table 3, and 12 kinds of FAs were detected in these samples. The detected contents higher than LOD but lower than LOQ was defined as “<LOQ”. The detection rate of benzyl alcohol was the highest (80%) with the content from 0.009 mg/kg to 5.73 mg/kg, followed by linalool (over 75%) with the content up to 12.30 mg/kg in makeup remover cleansing wipes. For FAs that commonly added in flower-scented wet wipes and sanitary napkins, phenethyl alcohol was detected in 13 samples, benzyl acetate in 15 samples, citronellol in 12 samples, geraniol in 8 samples and hydroxy citronellol in 6 samples. For fruit-scented FAs, limonene had a high detection rate of 70%, while only two samples had limonene concentrations above 1 mg/kg. Citral was detected in 9 samples, in which the content of 8 samples ranged from 0.022 mg/kg to 2.21 mg/kg. In addition, hydroxy citronellal, cinnamaldehyde, carvacrol and isoeugenol with contents below 1 mg/kg were detected. The other 7 kinds of FAs were not detected in these 20 kinds of samples. The TIC of samples B1 and D3 were shown in Fig. 3 and Fig. 4. The sample with most FAs detected was B1, and 10 FAs were detected in it.
Table 3
The results of the 19 FAs in the samples detected by HS-SPME-GC-MS
FAs
|
Detection rate(%)
|
Content range(mg/kg)
|
FAs
|
Detection rate(%)
|
Content range (mg/kg)
|
FAs
|
Detection rate(%)
|
Content range (mg/kg)
|
Limonene
|
70
|
<LOD
-1.07
|
Hydroxy citronellal
|
30
|
<LOD
-0.926
|
Geraniol
|
45
|
<LOD
-4.27
|
Diethyl maleate
|
|
<LOD
|
Cinnamyl alcohol
|
0
|
<LOD
|
Eugenol
|
0
|
<LOD
|
Citronellol
|
60
|
<LOD
-3.83
|
Phenethyl alcohol
|
65
|
<LOD
-3.17
|
Linalool
|
75
|
<LOD
-12.30
|
Cinnamaldehyde
|
15
|
<LOD
-detection
|
Amylcinnamyl alcohol
|
0
|
<LOD
|
Citral
|
45
|
<LOD
-2.21
|
Benzyl acetate
|
75
|
<LOD
-4.87
|
Furfuryl alcohol
|
0
|
<LOD
|
Carvone
|
10
|
<LOD
-<LOQ
|
Benzyl alcohol
|
80
|
<LOD
-5.73
|
Amyl cinnamaldehyde
|
0
|
<LOD
|
|
|
|
Isoeugenol
|
10
|
<LOD
-<LOQ
|
7-Methylcoumarin
|
0
|
<LOD
|
|
|
|
All the FAs tested in the samples did not exceed the limits in Chinese national standard for daily fragrances. As a reference, the determination results of the samples meet the requirement of cosmetics in EU standards.