A further study of the kinetics and optimization of the essential oil hydrodistillation from lavender flowers

Aleksandra Perovi?,Mihajlo Z.Stankovi?,Vlada B.Veljkovi?,2,Milan D.Kosti?,Olivera S.Stamenkovi?,

1 University of Ni?,Faculty of Technology,Bulevar oslobodjenja 124,Leskovac 16000,Serbia

2 The Serbian Academy of Sciences and Arts,Knez Mihailova 35,Belgrade 11000,Serbia

Keywords:Essential oil Lavender Lavandula officinalis L.Kinetics Hydrodistillation Modeling

ABSTRACT Hydrodistillation has commonly been used to recover essential oil from various plant materials,including lavender(Lavandula officinalis)flowers.The main objectives of the present study were to model the kinetics of the lavender essential oil (LEO) hydrodistillation using a phenomenological model,to evaluate the statistical significance of the hydromodule and hydrodistillation time on LEO yield combining a 42 full factorial design with the response surface methodology,to model statistically LEO yield by multiple non-linear regression and to determine the optimal process conditions that provided the maximum LEO yield.The fast-essential oil distillation(washing stage)in the initial period and the slow diffusion stage until the saturation occurring simultaneously were observed,justifying the use of the phenomenological model.With increasing the hydromodule,the saturation LEO yield and the washable fraction of the LEO decreased while the washing and diffusion rate constants increased.Knowledge of the LEO oil yield and the hydrodistillation kinetics is important from the techno-economical point of view.

1.Introduction

Lavender(Lavandula officinalis,L.angustifolia,or L.vera)is a perennial evergreen subshrub,native to the Mediterranean region growing on rocky and calcareous soils but is cultivated elsewhere [1].Lavander-based products,especially its essential oil,are commonly used in pharmaceuticals,phytomedicine preparations,as well as cosmetic and food products[1,2].Essential oil is contained in oil glands of flowers (2%–3%) and is commonly obtained from fresh and dried flowering tops by steam distillation or hydrodistillation[2].The content and composition of the lavender essential oil(LEO)depend on genotype,climatic conditions,growing location[3],stage of development[4],drying method and conditions[5–9],storage conditions[10],and distillation conditions like pressure,temperature,rate,and duration[11,12].The LEO contains more than 300 components,including linalool(9%–69%)and linalyl acetate(1%–59%)as the major compounds[2].

Several recovery methods including hydrodistillation,steam distillation,solvent extraction,supercritical CO2extraction(SCE),and novel techniques like ultrasound-,microwave-,ultrasound-microwaveassisted and pressurized fluid extraction have been applied for the LEO recovery.LEO is mainly produced by the conventional hydro-and steam distillation despite their drawbacks as they are time-consuming processes operating at a high temperature,which can degrade the thermal labile constituents[13].However,its advantages are low investment,simple extraction procedure,and no negative impact on the environment.Therefore,it is the most frequently used method for the commercial production of essential oils.Solvent extraction of LEO is utilized to a lesser degree[14]due to the extraction of unwanted compounds like waxes and pigments and the residual solvent in the product[15].SCE[15–22],ultrasound-[18],microwave-,ultrasoundmicrowave-assisted[23],and pressurized fluid[24,25]extractions,as well as steam[12]and microwave-accelerated steam distillation[26],can be promising alternatives to the conventional techniques for the LEO recovery.Several studies have compared different LEO recovery techniques with hydrodistillation[15,17,21,23].

The hydrodistillation of essential oil from lavender flowers has been rarely statistically optimized and kinetically modeled.Yu et al.[27]optimized the essential oil yield from air-dried lavender flowers achieved by the ultrasonic enhanced salt-containing hydrodistillation,regarding ultrasonic power,enhancing time,water-to-flower ratio(hydromodule),and MgSO4concentration using the Box–Behnken design coupled with the response surface methodology(RSM).The ultrasonic treatment of flowers ensured twice higher essential oil yield compared to the conventional hydrodistillation.Among the tested process factors,the hydromodule was the most important.The process variables of pressurized-hot water extraction(PHWE)and pressurized fluid extraction of essential oil from lavender flowers were optimized using a central composite [24]and Box–Behnken [25]design,respectively.Besides that,Rashed et al.[23]combined the RSM with the Box–Behnken design to optimize the LEO extraction conditions by the ultrasonic-microwave-assisted hydrodistillation preceded by an enzymatic pretreatment.

Only Stanojevi? et al.[28]modeled the kinetics of the LEO hydrodistillation from ground dried lavender flowers using Ponomaryov's equation [29].However,this simple linear model describes only the kinetics of the slow hydrodistillation period where the diffusion of the essential oil from the internal oil reservoirs toward the external medium occurs.These researchers did not analyze the effect of hydromodule on the kinetics of the essential oil hydrodistillation.Besides that,Zheljazkov et al.[12]used the asymptotic model to describe the variation in LEO yield during the steam distillation of dried lavender flowers.On the other hand,the kinetics of the SCE extraction of essential oil from lavender flowers has been frequently modeled[17,20,22].The knowledge of the essential oil extraction kinetics is important for proper operation,design,and scale-up of a hydrodistillation process.

In this work,the extraction of LEO from ground dried lavender flowers by hydrodistillation was investigated.Hydrodistillation was performed using a Clevenger apparatus at water-to-flower ratios(hydromodules)ranging from 10 to 25 ml·g?1.The main objectives were to model the kinetics of the LEO hydrodistillation process using a phenomenological model,to evaluate the statistical significance of the process variables (hydromodule and hydrodistillation time) on LEO yield using a 42full factorial design coupled with the RSM,to model LEO yield statistically by multiple non-linear regression and to determine the optimal process conditions that provided the maximum LEO yield.

2.Materials and Methods

2.1.Materials

Air-dried lavender(L.officinalis)flowers were obtained from the Institute for Medicinal Plant Research“Dr.Josip Pan?i?”Belgrade,Serbia.The plant material was ground by an electrical mill(15,000 r·min?1,1 min)before the hydrodistillation(average particle size 0.5 mm).The total LEO content was determined to be 6.47 ml.(100 g)?1of dry plant material by a prolonged hydrodistillation(6 h)at the hydromodule of 10 ml·g?1.Diethyl ether and anhydrous sodium sulfate,both p.a.quality,were purchased from Reahem(Novi Sad,Serbia)and Merk-Alkaloid(Skopje,North Macedonia),respectively.

2.2.Hydrodistillation

All experiments were carried out using the same Clevenger apparatus and electrical heater.The flowers(15 g)were placed in the vessel of the Clevenger apparatus,submerged with an appropriate amount of water (corresponding to the hydromodule of 10,15,20 and 25 ml·g?1),and the oil was distilled off.After hydrodistillation for 5,10,15,30,45,60,90,120,150,180,210,and 240 min,the volume of the essential oil collected in the measuring tube of the Clevenger extension was measured.

2.2.1.Kinetic modeling

The phenomenological kinetic model,derived for a batch hydrodistillation vessel,where the lavender flowers were immersed in water,was used[30]:

where y and ySare the LEO yields at the time of t and the saturation,respectively,kwand kdare the washing and diffusion rate constant,respectively and f and(1 ?f)are the washable fraction of the essential oil located on external surfaces of the plant particles and the diffusible fraction of the essential oil uniformly distributed within the plant particles.The parameters of Eq.(1)were calculated by fitting this equation to the experimental oil yield data by the nonlinear least-squares method and minimizing the sum of the squared deviations between the experimental and calculated oil yield values.The computer program employed the Levenberg–Marquardt algorithm,which combined the Gauss-Newton method and the steepest descent method,to adjust the parameter values in the iterative procedure.The mean relative percentage deviation(MRPD)between the predicted and actual oil yield values was the criterion used to evaluate the goodness of fit of the developed kinetic model.

2.2.2.Statistical analysis

Within the framework of this paper,the influence of two factors,viz.process conditions (hydrodistillation time and hydromodule)on the LEO yield was investigated.To optimize the LEO hydrodistillation,a 42full factorial design was used in combination with the RSM.The experimental matrix with actual LEO yields is given in Table 1.In order to avoid errors due to the influence of process variables on oil yield,experiments were performed randomly.The Kolmogorov–Smirnov normality test verified that the LEO yield data were normally distributed at the 0.05 level of significance.The significance of the factors and their combinations was evaluated using the variance analysis method(ANOVA),using the computer software Design Expert 7.0.0.Multiple non-linear regression equations were derived showing the dependence of LEO yield on process factors and their interactions.

The experimentally obtained LEO yield values were analyzed by a multiple regression model to fit the quadratic equation.Testing the impact of two factors at 4 levels allows modeling the response surface curve and makes it easier to find a relationship between the response(LEO yield) and each factor individually (hydromodule and hydrodistillation time).The general regression equation for this model is as follows:

where y is response(dependent variable,i.e.LEO yield),b0is the constant regression coefficient,and b1and b2are linear regression coefficients,b11and b22are quadratic regression coefficients,b12is twoparameter interaction regression coefficient,and X1and X2are factors(independent variables,i.e.hydromodule and hydrodistillation time,respectively).

Table 1 Experimental matrix

3.Results and Discussion

3.1.Kinetic modeling

Fig.1 shows the variations of the LEO yield during hydrodistillation at various hydromodules.As it was expected,the curves show the existence of the fast LEO distillation in the initial period,known as the washing stage,where the LEO yield increased rapidly because of the quick LEO removal from the outer surface of the plant particles.This mechanism of the LEO hydrodistillation is closely related to the anatomy of lavender flowers and their grinding.Most of the essential oil is produced in the hairs on the tubular calyces(buds)[31].By grinding of lavender buds,some of the reservoirs of essential oil were destroyed,thus releasing partly the essential oil on the external surface of the bud particles,which then rapidly distilled off during the washing stage.During the fast hydrodistillation,greater than 60%of the LEO was extracted from lavender flowers at all applied hydromodules;the highest extraction degree of about 81% was achieved with the hydromodule of 10 ml·g?1within 60 min.At the same time,the essential oil from the undestroyed reservoirs diffuses toward the external surfaces of the bud particles,slowing down the distillation of essential oil with time until the saturation(equilibrium)LEO yield was achieved.During this period,known as the diffusion stage,the diffusion of LEO through the plant particles controlled the overall process rate.In this way,the mechanism of the hydrodistillation processes at various hydromodules was the same and included fast(washing)and slow(diffusion)distillation of essential oil from lavender flowers,which occurred simultaneously.

Based on the observed mechanism of the LEO hydrodistillation,the phenomenological model,Eq.(1),was tested using the experimental data on LEO yield determined during the hydrodistillation processes at various hydromodules.The calculated values of the kinetic parameters,namely yS,f,kwand kd,are presented in Table 2.With increasing the hydromodule,the equilibrium LEO yield and the washable fraction of the LEO decreased while the washing and diffusion rate constants increased(Fig.S1,Supplementary material).Furthermore,the kinetic parameters were connected to hydromodule through the following linear relationships:

Fig.1.Variations of LEO yield with time at different hydromodules (hydromodules/ml·g?1:10 ○,15 Δ,20 □and*25).

Table 2 Parameters of the phenomenological model,Eq.(1)

The observed relationships of the model parameters,Eqs.(3)–(6),were explained by the effects of the increased volume of water at higher hydromodules on the LEO solubility and the viscosity of the suspension.The increased volume of water dissolved a higher amount of the hydrophilic components of the LEO,thus reducing the collected amount of LEO and consequently the equilibrium LEO yield and the washable fraction of the LEO.The reduction of the equilibrium LEO yield was observed with increasing the hydromodule in the range of 10–50 ml·g?1for the hydrodistillation of essential oil from aerial parts of creeping thyme(Thymus serpyllum)[32].?On the other hand,the increased volume of water,compared to the amount of plant material,reduced the viscosity of the suspension and increased the available solid–liquid contact area,thus both improving the external diffusion from the flower surface to the liquid phase and the internal diffusion through the flower particles,i.e.from the reservoirs of essential oil to the flower surface,thus enhancing the kwand kdvalues.

The simulation curves for different hydromodules agreed fairly well with the experimental points,as can be seen in Fig.1,and as indicated by the relatively small MRPD value of ±7.5% (48 data).This kinetic model has been already verified for modeling of the hydrodistillation of essential oils from fennel seeds,thyme and savory aerial parts,as well as juniper berries[30],thus indicating its general importance.

3.2.Statistical modeling and optimization

The essential oil hydrodistillation from lavender flowers was statistically analyzed,modeled,and optimized combining the RSM with the 42full factorial design.The analysis included two factors at four levels:hydromodule(10,15,20,and 25 ml·g?1)and hydrodistillation time(30,60,90,and 120 min).The quadratic equation,Eq.(2),was used to connect LEO yield with the process factors.The acceptability of the polynomial models of different orders was firstly examined by the sequential sum of squares and model summary statistic tests that aimed at the selection of the model with the highest order and maximized R2and R2-values,respectively.As can be seen in Tables S1 and S2(Supplementary material),the quadratic model had the highest F-value and thelowest p-value,as well as the high R2-,R2-and R2-values,and hence,it was adopted for further evaluating by the ANOVA.

Table 3 ANOVA results for the quadratic model

The results of the ANOVA evaluation of the statistical significance of the process factors(hydromodule and hydrodistillation time)and their interaction on the LEO yield,which is based on their F-and p-values,are presented in Table 3.The greater the F-value of a process factor or interaction,the greater its statistical significance whereas the values of p ＜0.05 indicate the significant factors and their interaction at the 95%confidence level.Both the process factors and their individual squares had a statistically significant effect on the LEO yield while their interaction did not influence it significantly.Having a higher F-value,the hydromodule had a statistically more significant effect on LEO yield than the hydrodistillation time.The ANOVA results were validated by the normally distributed population of the experimental data and Cook's distance which was significantly below the limit value(0.935),as can be seen in Fig.S2a and b(Supplementary material),respectively.

The parameters of the quadratic model were determined by multiple non-linear regression using the data from Table 1,and the resulting model was given in the form of the following quadratic equation:

where X1and X2are actual (uncoded) values of hydromodule and hydrodistillation time,respectively.This model equation is valid only within the applied experimental ranges of the process factors.

The reliability of the developed quadratic model was assessed based on several statistical criteria.First,a high F-value(164.2)and a low pvalue(＜0.0001)pointed out the statistical significance of this model.Moreover,the high value of the coefficient of determination (R2=0.988)indicated a good prediction of LEO yield as 98.8%of the variation in the LEO yield could be explained by this model.In addition,the values of(0.982)and(0.974)agreed to each other.The model accuracy was also proved by a low value of the coefficient of variation(2.42%)and a large adequate precision(41.2),which was much higher than the limit value of 4.Finally,the reliability of the quadratic model was verified by a small MRPD-value(±1.7%,16 data).

Fig.2 shows the response surface and contour plots of LEO yield as a function of hydromodule and hydrodistillation time.These illustrations were suitable for visualizing the effects of the process variables on the LEO yield and the optimal conditions providing the maximum LEO yield.

The LEO yield decreased with the increase of the hydromodule independently of the hydrodistillation time,which was also confirmed by the negative value of the linear coefficient of the hydromodule in Eq.(7).Such a change in the LEO yield was explained by the influence of the specific heating power(heating power per unit mass of the suspension)and the variation in the amount of hydrophilic components of the LEO in the floral water with increasing the hydromodule.Using the input heating power,the specific heating power decreased with increasing the hydromodule,which slowed down the distillation of essential oil and reduced the equilibrium LEO yield.In addition,as the amount of water increased,the amount of the hydrophilic components of the LEO decreased in the vapor phase due to their dissolution in the floral water,thus reducing the LEO yield.On the other hand,with increasing the hydrodistillation time,the LEO yield increased.Obviously,the longer hydrodistillation allowed the extraction of more LEO.From Fig.2,it can be also concluded that the effect of the hydrodistillation time on the LEO yield is more pronounced at smaller hydromodules.

The hydrodistillation process optimization was based on the developed quadratic model and the desirability function.This model predicted the maximum LEO yield of 5.60 ml.(100 g)?1at the hydromodule of 10 ml·g?1in 108 min.The actual LEO yield achieved under the optimum hydrodistillation conditions was 5.57 ml·g?1.The agreement between the predicted and actual LEO yield was another confirmation of the adequacy of the developed quadratic model.

4.Conclusions

Fig.2.Response surface(a)and contour(b)plots of LEO yield as a function of hydromodule and hydrodistillation time.

Conventional hydrodistillation of the essential oil from lavender flowers was studied at different hydromodules in the range of 10–25 ml·g?1.Independently of hydromodule,this process occurred by the same mechanism including simultaneous fast(washing)and slow(diffusion)distillation of essential oil.This mechanism was proven by kinetic modeling using the phenomenological model.Statistical modeling showed that the effect of the hydromodule on the LEO yield was statistically more significant than hydrodistillation time;the former had a negative influence on LEO yield while the latter affected it positively.The optimum operating conditions for The LEO hydrodistillation were the hydromodule of 10 ml·g?1and the hydrodistillation time of 108 min,which provided a LEO yield of 5.6 ml.(100 g)?1,corresponding to the recovery degree of 87%.

Nomenclature

b0Constant regression coefficient

biLinear regression coefficient(i=1,2)

biiQuadratic regression coefficient(i=1,2)

bijRegression coefficient of two–factor interaction(i=1,j=2)

C.V. Coefficient of variation

F Fisher's test value

h Hydromodule defined as water-to-lavender flower ratio,ml·g?1

f Fraction of lavender essential oil extracted by washing

(1-f) Fraction of lavender essential oil extracted by diffusion

kwRate constant of washing,min?1

kdRate constant of diffusion,min?1

MRPD Mean relative percentage deviation,%

p Probability value

R2Coefficient of determination

t Time,min

XiProcess variable (i=1:hydromodule;and i=2:hydrodistillation time)

y Lavender essential oil yield,ml·(100 gd.b.)?1

ysLavender essential oil yield at equilibrium(dry biomass),ml·(100 g)?1

Acknowledgements

This work has been funded by the Ministry of Education,Science and Technological Development of the Republic of Serbia(Project assigned to the Faculty of Technology,Leskovac,University of Ni?,Research Group III 45001,No.451-03-68/2020-14/200133).

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.cjche.2020.06.028.