Optimization of Steam Distillation Process and Chemical Constituents of Volatile Oil from Angelicae sinensis Radix

In this study, the steam distillation process of volatile oil from Angelicae sinensis Radix was optimized according to the concept of quality-by-design. A homemade glass volatile oil extractor was used to achieve better cooling of the volatile oil. First, the soaking time, distillation time, and liquid– material ratio were identified as potential critical process parameters by consulting the literature. Then, the three parameters were investigated by single factor experiments. The volatile oil yield increased with the extension in the distillation time, and first increased and then decreased with the increase in soaking time and liquid–material ratio. The results confirmed that soaking time, distillation time, and liquid–material ratio were all critical process parameters. The kinetics models of volatile oil distillation from Angelicae sinensis Radix were established. The diffusion model of spherical particle was found to be the best model and indicated that the major resistance of mass transfer was the diffusion of volatile oil from the inside to the surface of the medicinal herb. Furthermore, the Box–Behnken experimental design was used to study the relationship between the three parameters and volatile oil yield. A second-order polynomial model was established, with R2 exceeding 0.99. The design space of the volatile oil yield was calculated by a probability-based method. In the verification experiments, the average volatile oil yield reached 0.711%. The results showed that the model was accurate and the design space was reliable. In this study, 21 chemical constituents of volatile oil from Angelicae sinensis Radix were identified by gas chromatograph-mass spectrometer(GC-MS), accounting for 99.4% of the total volatile oil. It was found that the content of Z-ligustilide was the highest, accounting for 85.4%.


Introduction
Angelicae sinensis Radix is the dry root of Angelica sinensis (Oliv.) Diels [1]. It is widely used as a herbal medicine and has the effects of clearing heat and promoting dieresis, invigorating qi, and blood, etc. Volatile oil is an important component of Angelicae sinensis Radix [2,3]. The content of neutral oil in the volatile oil of Angelicae sinensis Radix is the highest. The volatile oil of Angelicae sinensis Radix mainly contains Z-ligustilide and other components [4,5]. Volatile oil is also considered as the main effective component of Angelicae sinensis Radix, which has the effects of treating hypertension [6], analgesic and antiinflammatory [7], and anti-tumor [8]. The preparation process of the Angelicae sinensis Radix volatile oil exists in the production of many Chinese medicines such as Danggui Tiaojing granules, Ruhe Sanjie tablets, Compound Herba Leonuri capsules, and Yangxueyin oral liquid, which are all included in the 2020 edition of the Chinese Pharmacopoeia (1st Part).

Distillation Method
Gansu Province is a genuine producing area of Angelicae sinensis Radix. Therefore, the Angelicae sinensis Radix from Gansu Province was used in this work. The steam distillation was used to prepare the volatile oil from Angelicae sinensis Radix. Figure 1 shows the essential oil extractor. With the homemade essential oil extractor, better cooling could be achieved. Compared to that of using a conventional extractor, more essential oil can be extracted. A proper amount of Angelicae sinensis Radix was taken and then crushed with a crusher (FD200T, Shanghai Traditional Chinese Medicine Machinery Factory, Shanghai, China). A total of 100 g of Angelicae sinensis Radix powder was taken and added into a 2000 mL flask. Water and several glass beads were added into the flask. After that, the flask was shaken to wet the herbal powders and was connected with a condenser tube. Before the experiment, water was added into the volatile oil extractor. An electric heating jacket (KDM type, Shandong Zhencheng Hualu Electric Equipment Co. Ltd., Shandong, China) was used to heat the flask. After boiling, the electric heating jacket was adjusted to keep boiling. The heating was stopped after a certain time of distillation. After more than 1 h, the height of the volatile oil was measured with a vernier caliper (0-150 mm, 3V type, Guilin Ganglu Digital Measurement and Control Co. Ltd., Guangxi, China). The volume of volatile oil was calculated, then the volatile oil was collected. The obtained volatile oil was dehydrated and dried with anhydrous sodium sulfate overnight to obtain an oil-like product until the amount of volatile oil no longer increased. Thee Angelicae sinensis Radix volatile oil was sealed in brown reagent bottles and stored in a refrigerator at 4 °C . The volatile oil yield was calculated using Equation (1).

Optimization of Distillation Process Parameters
After the literature review, the process parameters of distilling the volatile oil from herbal medicines by steam distillation were found and are shown in an Ishikawa diagram ( Figure 2). A proper amount of Angelicae sinensis Radix was taken and then crushed with a crusher (FD200T, Shanghai Traditional Chinese Medicine Machinery Factory, Shanghai, China). A total of 100 g of Angelicae sinensis Radix powder was taken and added into a 2000 mL flask. Water and several glass beads were added into the flask. After that, the flask was shaken to wet the herbal powders and was connected with a condenser tube. Before the experiment, water was added into the volatile oil extractor. An electric heating jacket (KDM type, Shandong Zhencheng Hualu Electric Equipment Co., Ltd., Qingdao, China) was used to heat the flask. After boiling, the electric heating jacket was adjusted to keep boiling. The heating was stopped after a certain time of distillation. After more than 1 h, the height of the volatile oil was measured with a vernier caliper (0-150 mm, 3V type, Guilin Ganglu Digital Measurement and Control Co., Ltd., Guilin, China). The volume of volatile oil was calculated, then the volatile oil was collected. The obtained volatile oil was dehydrated and dried with anhydrous sodium sulfate overnight to obtain an oil-like product until the amount of volatile oil no longer increased. Thee Angelicae sinensis Radix volatile oil was sealed in brown reagent bottles and stored in a refrigerator at 4 • C. The volatile oil yield was calculated using Equation (1).

Optimization of Distillation Process Parameters
After the literature review, the process parameters of distilling the volatile oil from herbal medicines by steam distillation were found and are shown in an Ishikawa diagram ( Figure 2).
According to the pre-experiments, it was considered that the potential CPPs affecting the volatile oil yield by steam distillation were distillation time (A), soaking time (B), and the liquid-material ratio (C). In this study, the above three factors were investigated by single factor experiments. During the investigation, a total of 100 g of Angelicae sinensis Radix powder was taken, soaked for a certain time, and then heated to distill the volatile oil. After reaching the distillation time, the heating was stopped. After the volatile oil was cooled, the volume of the volatile oil was accurately read, and the volatile oil yield was calculated. When the distillation time was studied, the liquid-material ratio was 10:1 mL·g −1 and the soaking time was 2 h. When the soaking time was studied, the ratio of the liquid to solid According to the pre-experiments, it was considered that the potential CPPs affecting the volatile oil yield by steam distillation were distillation time (A), soaking time (B), and the liquid-material ratio (C). In this study, the above three factors were investigated by single factor experiments. During the investigation, a total of 100 g of Angelicae sinensis Radix powder was taken, soaked for a certain time, and then heated to distill the volatile oil. After reaching the distillation time, the heating was stopped. After the volatile oil was cooled, the volume of the volatile oil was accurately read, and the volatile oil yield was calculated. When the distillation time was studied, the liquid-material ratio was 10:1 mL·g −1 and the soaking time was 2 h. When the soaking time was studied, the ratio of the liquid to solid was 10:1 mL·g −1 and the distillation time was 4 h. When the liquid-material ratio was studied, the distillation time was 4 h and the soaking time was 2 h.
Next, the Box-Behnken design was used to optimize the distillation process of the volatile oil from Angelicae sinensis Radix with the distillation time, soaking time, and liquid-material ratio as the factors. Compared with some other commonly used response surface designs, the number of experiments in the Box-Behnken design is relatively smaller when there are three 3-level factors. For example, there are 29, 17, and 13 runs for the full factorial design, central composite design, and the Box-Behnken design when there are three repetitions of the center point, respectively. The volatile oil yield was considered as the evaluation index. The factor levels are shown in Tables 1 and 2.

Analysis of Chemical Constituents of Volatile Oil from Angelicae sinensis Radix
A total of 100 µL of volatile oil was measured accurately, diluted to 10 mL with ethanol, filtered with a 0.22 µm microporous filter membrane, and loaded into a sample bottle. The chemical constituents of the volatile oil samples were determined by gas chromatographymass spectrometry (Agilent 7890A/5975C, Agilent, Santa Clara, CA, USA).
The identification of volatile compounds was performed by computer matching their mass spectra with those stored in a digital library of mass spectral data (NIST 14). The identification results were tentatively identified by the EI-MS spectrum and further experiments are planned to confirm their identification by authentic standards.

Kinetic Models
Volatile oil yield was fitted by Equations (2)-(4), as shown below. First-order kinetic model: where C eq is the concentration of solution at equilibrium; k is the total distillation rate constant; and t is time. Peleg's model: where k 1 is the mass transfer rate constant and k 2 is the concentration of the solution at equilibrium. The diffusion model of spherical particle: where D is the solute diffusion coefficient in the solvent and R is the particle radius. In the calculation, only the first three terms (n = 3) were taken for the sum of the infinite order. In order to evaluate the fitting effect of each model, R 2 can be calculated according to Equation (5).
where C act is the experimental value; C act is the average value of the measured value, and C fit is the model fitting value. The software MATLAB R2019b (American Math Works Company) was used to analyze and process the data. Except for R 2 , other indices of RMSE, MSE, SSE, and MAE were also calculated according to Equations (6)- (9).
The software Microsoft Office Excel (American Microsoft Company) was used to analyze and process the data.

Data Processing of Volatile Oil Distillation Rate Obtained from Experimental Design
The software "Design-Expert 8.0.6" was used to analyze the experimental data collected from the Box-Behnken designed experiments. The volatile oil yield was taken as the evaluation index (Y), and a second-order polynomial equation fitting was carried out. Equation (10) was adopted as the mathematical model between the evaluation index and the three process parameters.
where b 1-9 is the partial regression coefficient and b 0 is the intercept.

Optimization of Distillation Parameters of Volatile Oil by the Monte Carlo Method
The design space was calculated by the Monte Carlo method with a parameter optimization software compiled by MATLAB R2018b (Math Works Company of America) [20]. According to previous work, the design space calculated with this method is more reliable [20]. Three factors affecting the volatile oil yield were simulated randomly. The combination of parameters with the probability of no less than 0.80 of attaining the preset volatile oil yield value was taken as the design space. In the calculation, the step lengths of distillation time, soaking time, and liquid-material ratio were set to 0.04, 0.02, and 0.08, respectively. The simulations were conducted 2000 times.

Critical Process Parameters of Volatile Oil Distillation
The distillation time was changed to 2, 4, 6, 8, and 10 h, respectively, and the volatile oil yield is shown in Figure 3. The volatile oil yield increased continuously within the first 8 h of distillation time. When the distillation time reached 8 h, the volatile oil yield reached 0.68%, and the yield tended to be stable when the distillation time was prolonged.
The soaking time was changed to 0, 1, 2, 3, and 4 h, the yield of volatile oil from Angelicae sinensis Radix is shown in Figure 4. When the soaking time was 3 h, a maximum oil yield of 0.67% was obtained. After soaking time reached 3 h, the oil yield decreased with the extension in the soaking time. The reason may be that the aqueous extract became viscous when soaking for too long, which was not conducive to the distillation of the volatile oil. of distillation time, soaking time, and liquid--material ratio were set to 0.04, 0.02, and 0.08, respectively. The simulations were conducted 2000 times.

Critical Process Parameters of Volatile Oil Distillation
The distillation time was changed to 2, 4, 6, 8, and 10 h, respectively, and the volatile oil yield is shown in Figure 3. The volatile oil yield increased continuously within the first 8 h of distillation time. When the distillation time reached 8 h, the volatile oil yield reached 0.68%, and the yield tended to be stable when the distillation time was prolonged. The soaking time was changed to 0, 1, 2, 3, and 4 h, the yield of volatile oil from Angelicae sinensis Radix is shown in Figure 4. When the soaking time was 3 h, a maximum oil yield of 0.67% was obtained. After soaking time reached 3 h, the oil yield decreased with the extension in the soaking time. The reason may be that the aqueous extract became viscous when soaking for too long, which was not conducive to the distillation of the volatile oil.  The liquid-material ratio was changed to 6:1, 8:1, 10:1, 12:1, and 14:1 (mL·g −1 ), respectively, and the volatile oil yield is shown in Figure 5. When the liquid-material ratio was not higher than 10:1, the volatile oil yield increased as the liquid-material ratio increased. The reason may lie in the more uniform heating when the liquid-material ratio increased, which was beneficial to the diffusion and dissolution of the volatile oil. In Figure 4, when the liquid-material ratio was greater than 10:1, the volatile oil yield showed an obvious downward trend, which may be due to the loss caused by the dissolution of the volatile oil in water. The liquid-material ratio was changed to 6:1, 8:1, 10:1, 12:1, and 14:1 (mL·g −1 ), respectively, and the volatile oil yield is shown in Figure 5. When the liquid-material ratio was not higher than 10:1, the volatile oil yield increased as the liquid-material ratio increased. The reason may lie in the more uniform heating when the liquid-material ratio increased, which was beneficial to the diffusion and dissolution of the volatile oil. In Figure 4, when the liquid-material ratio was greater than 10:1, the volatile oil yield showed an obvious downward trend, which may be due to the loss caused by the dissolution of the volatile oil in water.
The above results indicate that the distillation time, soaking time, and liquid-material ratio all had a great influence on the yield of the volatile oil from Angelicae sinensis Radix and were all indeed the CPPs. tively, and the volatile oil yield is shown in Figure 5. When the liquid-material ratio was not higher than 10:1, the volatile oil yield increased as the liquid-material ratio increased. The reason may lie in the more uniform heating when the liquid-material ratio increased, which was beneficial to the diffusion and dissolution of the volatile oil. In Figure 4, when the liquid-material ratio was greater than 10:1, the volatile oil yield showed an obvious downward trend, which may be due to the loss caused by the dissolution of the volatile oil in water. The above results indicate that the distillation time, soaking time, and liquid-material ratio all had a great influence on the yield of the volatile oil from Angelicae sinensis Radix and were all indeed the CPPs.

Study on Kinetics of Volatile Oil Distillation
Three models were used to fit the kinetic data of the volatile oil distillation, and the fitting results are shown in Figure 2. The first-order kinetic model fitting results C eq and k were 0.6644% and 0.5861 h −1 , respectively, and R 2 was 0.8198. The Peleg's model fitted k 1 and k 2 were 1.6156 h·% −1 and 1.3105 % −1 , respectively, and the R 2 was 0.9338. The diffusion model of spherical particle fitted C eq and D R 2 were 2.5317% and 0.0007 h −1 , respectively, and the R 2 was 0.9710. Among the three models, the diffusion model of spherical particle fitted R 2 was the highest, which could explain more than 97% of the variance in experimental data. Table 3 shows the errors between the predicted values and the actual values of different models. The prediction values of the diffusion model were closest to the actual values. The diffusion model of spherical particle assumed that the major resistance of mass transfer in the distillation process of volatile oil was the diffusion of volatile oil from the inside to the surface of the medicinal materials. The mass transfer rate of volatile oil from the surface of the medicinal materials to the extraction solution and the mass transfer rate in the air inside the extractor were all much faster. These results indicate that a rapid distillation of volatile oil can probably be realized by lowering the average particle size of Angelicae sinensis Radix.

Data Processing and Model Fitting
By modeling the oil yield data, the multivariate binomial regression model obtained was as follows: Y = 0.710 + 0.026A + 0.010B − 0.026C + 2.50 × 10 −3 AB − 5.00 × 10 −3 AC − 7.50 × 10 −3 BC − 0.068A 2 − 0.030B 2 − 0.082C 2 . The model determination coefficient  Table 4. The F value of the model was 93.13, and p < 0.0001, which showed that the model was extremely significant. The model R 2 exceeded 0.99, which showed that the model could well explain the data variation in the experiment. A (p < 0.0001) and C (p < 0.0001) were extremely significant, and B (p < 0.05) was significant. Among the quadratic terms, A 2 (p < 0.0001), B 2 (p = 0. 0003), and C 2 (p < 0.0001) were all extremely significant terms, which showed that the influence of the three parameters on the oil yield was nonlinear. The results were consistent with the previous single factor experimental results. The p values of AB, BC, and AC were all greater than 0.10, which showed that the interaction between the three factors was small. Figure 6 shows the Pareto chart. It can be concluded that all the linear terms and quadratic terms were significant.

Response Surface Diagram and Contour Diagram
The response surface diagram and contour diagram of the distillation process of the volatile oil from Angelicae sinensis Radix are shown in Figure 7. The figure reflects the effects of the distillation time, soaking time, and liquid-material ratio on the oil yield. When the liquid-material ratio is fixed, the oil yield increases first and then decreases gradually with the increase in distillation time. When the distillation time is too long, some volatile components in the volatile oil may volatilize and be lost. When the soaking time is fixed, the volatile oil yield increases first and then decreases gently with the increase in the distillation time, and increases first and then decreases with the increase in the liquid-material ratio. When the distillation time is fixed, the oil yield increases first and then stabilizes with the increase in soaking time, and increases first and then decreases with the increase in the liquid-material ratio. The acceptable lower limit of volatile oil yield was set at 0.65%, and the lowest pro ability of target attainment was set at 0.80. The design space calculated by the probabili based method is shown in Figure 7. Figure 7a is a three-dimensional design space d

Design Space Calculation and Verification
The acceptable lower limit of volatile oil yield was set at 0.65%, and the lowest probability of target attainment was set at 0.80. The design space calculated by the probability-based method is shown in Figure 7. Figure 7a is a three-dimensional design space diagram, and Figure 8b is a two-dimensional space diagram after fixing the liquid-material ratio at 9.3:1. It can be seen from the figure that the design space was close to the shape of a football. The design space was verified with a selected condition of 8.4 h of distillation time, 3.2 h of soaking time, and 9.3:1 (mL·g −1 ) of liquid-material ratio. Under this condition, the probability for target attainment was 1.0 and the predicted volatile oil yield was 0.715%. Three parallel validation tests were carried out, and the oil yield of Angelicae sinensis Radix volatile oil was 0.711%, 0.709%, and 0.712%, respectively. The average oil yield was 0.711% and the relative standard deviation was 0.21%. The density of the collected volatile oil was 1.0075 g/cm 3 at room temperature. The mathematical model was developed according to the results of the Box-Behnken designed experiments. The measured values were close to the predicted values, indicating that the prediction of the model was accurate. The volatile oil yield was higher than the preset standard, which shows that the design space was reliable.

Qualitative Analysis of Chemical Constituents of Volatile Oil
The GC-MS analysis of the volatile oil from Angelicae sinensis Radix was carried out, and the total ion flow diagram is shown in Figure 9. The results of the mass spectrometry are shown in Table 5. Twenty chemical constituents were identified, accounting for 99.426% of the total volatile oil. Among them, Z-ligustilide had the highest relative content, and its relative content reached 85.385%. Its structural formula can be found in SciFinder [21]. Other components with relatively high contents were α-pinene, β-ocimene, 2-methyl−1,3-benzoxazole, butylidenephthalide, and E-ligustilide. The design space was verified with a selected condition of 8.4 h of distillation time, 3.2 h of soaking time, and 9.3:1 (mL·g −1 ) of liquid-material ratio. Under this condition, the probability for target attainment was 1.0 and the predicted volatile oil yield was 0.715%. Three parallel validation tests were carried out, and the oil yield of Angelicae sinensis Radix volatile oil was 0.711%, 0.709%, and 0.712%, respectively. The average oil yield was 0.711% and the relative standard deviation was 0.21%. The density of the collected volatile oil was 1.0075 g/cm 3 at room temperature. The mathematical model was developed according to the results of the Box-Behnken designed experiments. The measured values were close to the predicted values, indicating that the prediction of the model was accurate. The volatile oil yield was higher than the preset standard, which shows that the design space was reliable.

Qualitative Analysis of Chemical Constituents of Volatile Oil
The GC-MS analysis of the volatile oil from Angelicae sinensis Radix was carried out, and the total ion flow diagram is shown in Figure 9. The results of the mass spectrometry are shown in Table 5. Twenty chemical constituents were identified, accounting for 99.426% of the total volatile oil. Among them, Z-ligustilide had the highest relative content, and its relative content reached 85.385%. Its structural formula can be found in SciFinder [21]. Other components with relatively high contents were α-pinene, β-ocimene, 2-methyl−1,3benzoxazole, butylidenephthalide, and E-ligustilide.

Discussion
Li Tao et al. [22] found that the content of volatile oil from wild Angelicae sinensis Radix (1.14%) was more than twice as high than that from the artificially cultivated Angelicae sinensis Radix (0.4%). Yan Hui et al. [23] found that longer sunshine was not conducive to the increase in volatile components that mainly consisted of ligustilide. Lin Haiming [24] found that the content of the volatile oil and alcohol-soluble extract decreased gradually with the increase in the drying temperature. Ji Peng et al. [25] determined and analyzed the chemical components of volatile oil in raw Angelicae sinensis Radix and different processed products of Angelicae sinensis Radix by GC-MS. It was found that different processing methods would affect the total amount of volatile oil in Angelicae sinensis Radix and the contents of ligustilide and butenyl phthalolactone in the volatile oil [25]. Li Runhong et al. [26] studied the difference in the volatile oil composition and content in Angelicae sinensis Radix from different producing areas. The results showed that the highest volatile oil content of wild Angelicae sinensis Radix in Linzhi (Tibet Province) was 2-hydrobutenyl phthalide (70.184%), followed by n-butenyl phthalide (9.288%) [26]. In the volatile oil of wild Angelicae sinensis Radix in Pingwu (Sichuan Province), the content of 2-hydrobutenyl phthalide (92.551%) was the highest, followed by butenyl phthalide (3.037%) [26]. These results were different compared with those reported in this study and most of the literature, in which ligustilide was found to be the most abundant constituent. Generally speaking, the composition and content of volatile oil in Angelicae sinensis Radix are affected by the growth environment, harvest time, drying method, processing method, extraction technology, etc. [14,27]. Therefore, the composition and content of volatile oil in Angelicae sinensis Radix may have significant differences.
The results of some published works [28] have shown that the yield of volatile oil obtained from Angelica sinensis by steam distillation was low (about 0.3-0.5%). In industry, the yield is even lower. Sometimes, only the volatile oil aqueous solution can be obtained. In this work, an improved steam distillation with enhanced cooling was used. Compared with that of using a traditional volatile oil extractor, the collected volatile oil could be increased by 30-58%. The collected volatile oil in Angelica sinensis was different to that in the literature [22,29,30], which may be attributed to germplasm resources, harvesting time, planting altitude, processing technology, and so on. However, the efficiency of steam distillation was not high. More volatile oil may be obtained when using other techniques such as supercritical fluid extraction [31].

Conclusions
In this work, an improved volatile oil extractor with better cooling was applied to collect the volatile oil from Angelicae sinensis Radix by steam distillation. The distillation process was optimized according to the concept of quality-by-design. The soaking time, distillation time, and liquid-material ratio were determined as CPPs. It was observed that the volatile oil yield increased with the increase in the distillation time. Furthermore, three models were used to fit the distillation kinetics data of the volatile oil. The fitting effect of the diffusion model of spherical particle was best, and the R 2 exceeded 0.97. This indicates that a more rapid distillation process can probably be realized when smaller Angelicae sinensis Radix can be used. The volatile oil yield increased first and then decreased with the increase in the soaking time and liquid-material ratio. Then, the distillation time, soaking time, and liquid-material ratio were investigated with the Box-Behnken designed experiments. The R 2 of the established second-order polynomial model exceeded 0.99. Then, the design space for distilling the volatile oil from Angelicae sinensis Radix was obtained with a probability-based method. The design space was verified at a condition shown as follows: Distillation time: 8.41 h, soaking time: 3.20 h, liquid-material ratio: 9.3:1 (mL·g −1 ). The probability of the target attainment was 1 at this condition. The actual volatile oil yield was 0.711%, which was close to the predicted value of 0.715% and higher than the preset standard, indicating that the model could predict accurately and the design space was reliable. The volatile oil of Angelicae sinensis Radix was also analyzed by GC-MS, and 21 chemical constituents were found, accounting for 99.426% of the total volatile oil. The content of Z-ligustilide in the volatile oil was the highest, and the relative content reached 85.38%.

Informed Consent Statement: Not applicable.
Data Availability Statement: All data generated or analyzed during this study are included in this published article.