The linalool essential oil and water mixtures, without stabilizing agent, immediately after its preparation (i.e., on the first day) are presented in Fig. 1S (Supplementary Material), with their respective nomenclature. These samples are slightly opaque with visible transparency, indicating low emulsification capacity, a characteristic usually reported for mixtures of immiscible liquids and distinct polarities (Shin et al. 2019). The oily and aqueous phases are not identifiable due to the staining of linalool essential oil, which is transparent and indistinguishable from water. The emulsions stabilized by CNC or CNF immediately after their preparation are presented in Fig. 2S. All emulsions showed the characteristic turbidity of emulsified samples, indicating that the emulsification methodology adopted was adequate. In the samples stabilized with CNC, no emulsion presented immediate phase separation after preparation, i.e., there was no immediate coalescence, and possibly the CNCs acted in the shielding of the droplet interface, avoiding its approximation (Tang et al. 2018). The emulsions stabilized by CNF presented a similar appearance, with high apparent turbidity, indicating successful emulsification.
3.1 Droplets' sizes
Two methods were used to measure the droplet sizes: DLS and optical microscopy. The techniques were used in a complementary way since DLS measures droplets that vary between nanometers and some micrometers, and this measurement is associated with Brownian movement of particles and drops, and the smaller the particle, the faster its movement. This technique does not measure the amount of light dispersed, but the dynamics of light disperses in short periods, at a fixed angle, being measured as a correlation function. Therefore, the size distribution calculation depends on the dispersion of light resulting from the emulsion droplets (Friberg et al. 2004). For optical microscopy, magnifying lenses and light incidence are used to visualize the emulsion microstructure. However, since it uses wavelength in visible light, there is a limitation of the measured particle sizes, restricted to micrometric particles.
The emulsions were submitted to DLS measurements on day zero after preparing the oil/water mixtures and dilution. The samples without stabilization did not present measurement readings, indicating that no drops were detected with sizes below the equipment's measurement limit, ~ 15 µm, regardless of the process conditions. Figure 2 shows the droplet sizes (DH hydrodynamic diameter, in µm) obtained for emulsions stabilized with CNC or CNF.
Among the emulsions stabilized with CNC (Fig. 2a), the majority presented similar diameters ranging between 0.02 and 1 µm, except for samples 7 (1% CNC, 10,000 rpm, 20% oil, 3 minutes) and 9 (1% CNC, 10,000 rpm, 30% oil, 7 minutes), which presented larger sizes, with 1.03 ± 0.005 µm and 2.6 ± 0.03 µm, respectively. Both samples have in common the lowest homogenization speed of 10,000 rpm and higher CNC concentrations (1%), indicating that, for Linalool/water/CNC emulsions, the rotation and concentration of cellulose nanocrystals are important parameters.
Figure 2b shows the DH of emulsions stabilized with CNF, and most samples showed similar droplet diameters. Two samples showed different size' pattern, with diameters greater than 35 µm: 3 (0.5% CNF, 10,000 rpm, 20% oil, 3 minutes) and 12 (0.5% CNF, 12,000 rpm, 30% oil, 3 minutes), and these emulsions have in common the preparation time and cellulose nanofibers content, indicating that both parameters have a significant influence on the emulsified system quality. Visually, both emulsions showed instability (Fig. 2S), e.g., sample 3 presented an aqueous layer, while sample 12 has a superior oily layer. These phenomena indicate a droplets coalescence tendency and, consequently, increased phase separation.
For the emulsified systems understanding, particle sizes' joint analysis is necessary, which is complimentary. Thus, the optical microscopy analysis was performed, and the average diameters results of the oil/water samples and stabilized with CNC or CNF are presented in Figs. 2c-e. Due to the difference in the light source and wavelength, significantly larger particles are measured by this technique. For the measurements, the samples were subtly agitated before the analysis.
Figure 2c shows the average diameters of samples without stabilizer, with values of ~ 100 µm, while after CNC stabilization, droplet sizes were ~ 50 µm (Fig. 2d). This variation in sizes is associated with stabilization efficiency with cellulose nanocrystals, which avoid the oil droplets approximation and coalescence due to their electrostatic charges generated by cellulose hydroxyls (Bai et al. 2019; Shin et al. 2019). For emulsions containing CNF (Fig. 2e), there was a significant variation between the samples, which can be explained by the gel-like formation, in which the drops are trapped inside the three-dimensional structure formed by cellulose nanofibers resulting in apparent larger sizes (Lu et al. 2019, 2021).
Visually, due to the standard deviation, the samples have similar sizes, and it is not possible to identify significant variations in droplet size. For a more detailed analysis, a statistical treatment was applied, using the One-Way ANOVA method and the Tukey test to evaluate the systems' similarity. Table 2 presents the Tuckey test results, comparing pairs of oil/water samples without the nanocellulose use, where the colored cells represent statistically different samples.
Table 2. Tukey test for the mixtures of linalool essential oil and water, without stabilizer, for droplets measured by optical microscopy on day 0.
The results indicate that sample 21 (12,000 rpm, 30% oil, 7 minutes) presents statistically different results, with the largest drop sizes, which may be an indication that extreme preparation conditions, e.g., highest rotation, oil concentration, and time, can result in a non-beneficial system disturbance, inducing a spontaneous and immediate coalescence. Sample 19 (12,000 rpm, 30% oil, 3 minutes) presents the most significant similarity with the others since it has intermediate drop sizes. Compared to sample 21, its preparation characteristics indicate that time is a relevant factor for preparing emulsified systems and may be associated with its entropy and, consequently, its stability. Moreover, when doing a peer analysis, it was impossible to correlate the parameters of oil concentration and homogenization speed, which can be justified by the absence of stabilizers and miscibility between liquids. For example, sample 1 (12,000 rpm, 20% oil, 3 minutes) showed similarity with 2 (10,000 rpm, 20% oil, 7 minutes), 16 (10,000 rpm, 30% oil, 3 minutes) and 19 (12,000 rpm, 30% oil, 3 minutes). By comparing process parameters, there is no constant associated similarity or differences in drop size values.
Table 3 presents the statistical comparison for emulsions stabilized with cellulose nanocrystals, and the results presented samples with high statistical similarity, and only sample 6 (0.5% CNC, 10,000 rpm, 30% oil, 3 minutes) presented different diameters (and higher than the others). In this case, the samples' analysis on day 0 indicated a drop size pattern, and it was necessary a temporal follow-up, and complementary analyses identify possible processing effects. This result was expected since in Fig. 2S there was no visual distinction between the samples, indicating a possible similarity after preparation.
Table 3. Tukey test for linalool, water, and CNC emulsions for droplets measured by optical microscopy on day 0.
Table 4 illustrates the statistical comparison for emulsions stabilized with cellulose nanofibers. The samples with the highest statistical differences were those with larger droplet sizes, while those with similarities are attributed to smaller diameters. Sample 3 (0.5% CNF, 10,000 rpm, 20% oil, 3 minutes) showed statistical similarity with 4 (0.5% CNF, 10,000 rpm, 30% oil, 7 minutes), 12 (0.5% CNF, 12,000 rpm, 30% oil, 3 minutes) and 14 (0.5% CNF, 12,000 rpm, 20% oil, 7 minutes), and all have as a common factor the CNF concentration.
Table 4. Tukey test for linalool, water, and CNF emulsions for droplets measured by optical microscopy on day 0.
Unlike observed for CNCs, the CNFs present a limit concentration so that they are efficient against agglomeration, and above this value, the droplets have significantly larger diameters. Similarly, the samples stabilized with 1% CNF showed similarity to each other. However, due to the visual aspect, it is impossible to identify whether, in the samples stabilized with CNF, stability is directly related to the droplets' size or the stabilization mechanism (gel formation through a three-dimensional network). Analyzing the droplet sizes by DLS, optical microscopy, and statistics, it was found that the smallest droplets, measured by DLS, were found for 1% CNF. In contrast, for optical microscopy results, this behavior was reversed, which may indicate stability for smaller droplets and the larger droplets resulting from the oily phase retained between the polymeric nanofibers, possibly resulting in the absence of phase separation.
3.2 Emulsions' morphology
Optical microscopy was used for droplets shape visual analysis. The mixture between oil and water results in two distinct and completely identifiable phases, as shown in Fig. 3S, and shortly after mixing, a momentary formation of oily drops occurs. However, due to instability, during the analysis, a coalescence trend was observed.
Figure 3 presents representative photomicrographs for emulsions stabilized with cellulose nanocrystals, and these presented well-defined droplets with a spherical shape, with large size distribution, i.e., high polydispersity, which may be associated with an insufficient coating of particles by CNCs (Varanasi et al. 2018).
It is also observed droplets similar to small "dots" associated with DLS analysis' micrometric sizes. It is not possible to observe indications of nanoparticles' presence due to equipment limitations since the optical microscope has no resolution for the visualization of particles on the nanoscale, and this effect is more pronounced for samples 9, 11, 18, and 23. These samples have a common factor the longer preparation time (7 minutes), indicating that longer homogenization times contribute to the droplets breakage in smaller droplets due to the higher disruptive energy available, resulting in greater surface area to be covered by CNC (Bai et al. 2019). However, due to the CNC rigid characteristic, it was expected that the CNCs concentration was an influential factor for obtaining smaller drops due to its effect on the alteration of energy distribution, which was not verified. These results indicate that, in the present study, the emulsions' viscosity was not significantly altered by the nanoparticles' concentration.
There is no logical influence pattern of the oil concentration parameter since smaller and larger sizes are distributed randomly. These results indicate that, for linalool emulsions stabilized with CNC, an oily phase content value was not reached that can be adequately stabilized, dependent on the other parameters. Another characteristic of strong influence on stability is the oil polarity and viscosity, which can generate intrinsically larger drops to other oils to the aqueous phase. Due to the essential oil chains arrangement, larger drops can generate a very high kinetic barrier to be overcome and break the drops into smaller sizes (Saffarionpour 2020). As described above, there was no statistically significant variation in particle sizes, a factor that was corroborated by the images presented.
Figure 4 illustrates the morphology of emulsions stabilized with cellulose nanofibers. Unlike that observed for CNC emulsions, they have a cloudy background and indications of heterogeneity in the samples. Turbidity and irregularity are gel characteristics, and this behavior is more pronounced for samples 13, 17, and 10, which have as a common parameter the CNF concentration of 1%, indicating that smaller concentrations are not sufficient for the formation of a stable three-dimensional structure, resulting in the free movement of oil drops and their approximation (instability mechanism) (Kalashnikova et al. 2011; Angkuratipakorn et al. 2017). According to Li et al., the formation of a gel layer decreases the tendency to coalescence due to steric impairments caused by fibers (Li et al. 2019b). Lu et al. reported that, under appropriate process conditions, fibrillar nanoparticles' higher concentrations could increase the system's gelling, making it stronger and, consequently, more stable, reflecting in emulsions with long-term stability (Lu et al. 2021).
Of the three samples that stood out, it is possible to observe that sample 13 (1% CNF, 10,000 rpm, 30% oil, 3 minutes) presented droplet sizes larger than the other samples, which can be attributed to low rotation speed processing time, which did not allow an adequate breakage of droplets in smaller droplets. In addition to the gel structure, another important point is that all samples stabilized with CNF presented indistinguishable droplets (points), which can be attributed to the CNFs elongated and flexible morphology and their ability to bind or overlap on the water surface of the oil. The larger drops observed for all emulsions result from the three-dimensional network formed, in which the droplets tend to disperse in porous space, and this dispersion caused by the Brownian movement in the aqueous phase approximates the droplets, resulting in the coalescence phenomena (Lu et al. 2021).
Comparing emulsions stabilized with CNC and CNF, it was found that particle morphology is an essential factor for the stability of linalool/water essential oil systems due to electrostatic stability mechanisms, resulting from intrinsic loads from cellulose hydroxyls, particle flexibility, and migration of them to the oil/water interface. In this case, cellulose nanofibers may have presented more stable results due to their hydrodynamic interactions, such as the long-range capillary attraction forces induced by the deformation of fluid-fluid interfaces, dominating the CNF migration at the interface and result in the confinement of drops within the formed network (Kalashnikova et al. 2013).
3.2 Stability under storage
Oil/water samples were prepared for comparison with stabilized ones. It was verified that there was complete phase separation between oil and water from the preparation moment, i.e., the samples are entirely unstable and immiscible due to the polarity difference. Figure 5 shows the digital photographs obtained for days 0 and 30 to compare the mixtures' stability.
Figure 6 shows the digital photographs obtained for days 0 and 30 to compare the CNC-emulsions stability. For these samples, the interest phase was the emulsified part, considered as stable. The instability phenomenon was more significant for samples 7 (1% CNC, 10,000 rpm, 20% oil, 3 minutes), 8 (0.5% CNC, 12,000 rpm, 20% oil, 3 minutes), and 23 (1% CNC, 12,000 rpm, 20% oil, 7 minutes), with the stability value below 30%, as shown in Table 5, and the three samples have as a common factor the lowest essential oil concentration.
The samples showed a non-adsorbed CNC sedimented fraction, an aqueous phase, and the top's emulsified phase. The sedimentation results from the nanoparticles' attraction to each other, which are strong enough to avoid their deposition at the oil/water interface, forming particle clusters later flocculated. This behavior is called entropy-driven (Lu et al. 2021).
The two most stable samples stabilized with CNC were 9 (1% CNC, 10,000 rpm, 30% oil, 7 minutes) and 11 (0.5% CNC, 12,000 rpm, 30% oil, 7 minutes), with values above 39%, indicating that for cellulose nanocrystals, time can be a decisive factor for obtaining stable emulsions. These results indicate that the emulsion stability is dependent on a factor combination that needs to occur to be a good coating of oil drops by nanoparticles (Bai et al. 2019; Shin et al. 2019).
Table 5. Stability values over time for linalool and water mixtures and their emulsions. Red lines indicate samples stabilized with CNC, and blue lines, samples stabilized with CNF.
* The symbol "-" represents no emulsified phase, and therefore it was not possible to calculate the samples' stability.
For CNF-emulsions (Figure 7), the stability values were significantly higher than that of CNC after 30 days, and the lowest stability was verified for 3 (0.5% CNF, 10,000 rpm, 20% oil, 3 minutes), 4 (0.5% CNF, 10,000 rpm, 30% oil, 7 minutes), and 12 (0.5% CNF, 12,000 rpm, 30% oil, 3 minutes), with values of 71.9 , 72.7 and 72.4%. These samples have the lowest CNF content in common, indicating that, for nanofibrillar morphology, higher contents induce greater emulsification. The most stable samples were 10 (1% CNF, 12,000 rpm, 30% oil, 7 minutes) and 17 (1% CNF, 12,000 rpm, 20% oil, 3 minutes), with values of 97 and 100%. These high stability values are associated with van der Waal forces, hydrogen bonds, colloidal interactions associated with DLVO theory (Derjaduin-Landau-Verwey-Overbeek), and capillary forces (Li et al. 2019a).
For linalool essential oil, essential oil concentration and time are factors that do not influence stability, while the others are an essential combination for stabilization (Jiménez Saelices and Capron 2018). Similar results regarding the concentration of CNFs were reported by Paximada et al. that investigated the use of bacterial cellulose nanofibrils as a stabilizer of oil/water emulsions (Paximada et al. 2016). Besides, while the linalool-CNC emulsions presented an emulsified behavior of a more liquid aspect, the linalool-CNF emulsions showed a gel-like behavior, and no excess oil layers were observed.
The variation of values between morphologies is associated with the aspect ratio and stabilization methodology, since CNCs stabilize via electrostatic interactions (stabilization associated with nanoparticle and essential oil intrinsic characteristics, such as surface charges and chemical structures), and CNFs form a three-dimensional network that traps the oil droplets inside and avoids their approximation (Lu et al. 2021).
3.3 Design of experiments
The DoE was carried out in a simplified way, using each parameter's minimum and maximum values to prepare emulsions. The results used for this analysis were the stability values on day 30 and the particle sizes in the micrometric range, obtained by optical microscopy, also for day 30. The results of the oil/water mixtures were not considered because, as previously reported, the mixtures are unstable. The design of experiments was used to analyze the sample set and verify, in a simplified way, the parameters that significantly influence the emulsion formation and stability.
Considering the stability results, for statistical analysis, ANOVA was used for the reduced 2FI model, and the F model value was 5.49, i.e., the model is significant. The R2 was 0.9083 for stability and 0.9795 for particle size, indicating a good fit. Also, P values less than 0.050 indicate that the terms of the model are significant. In this case, the DoE is a meaningful and valid model term. Thus, the design results are correct and can be used with statistical confidence.
Figure 8 presents the response surface obtained for stability, and the CNC-emulsions indicated that the essential oil concentration is a factor that does not influence the stability values, not being a key parameter for the preparation of the emulsions.
The results indicated that for shorter process times and speeds, stability increased (red color) with higher EO concentrations and lower CNC concentrations. By maintaining the speed and increasing the time, there was a linear trend, with better oil concentration remaining at 30% and the CNC concentration being variable, according to the selected configuration. By DoE, considering the preparation time and its impact on stability, it was impossible to identify the best conditions, and therefore, the experimental results were essential, as described earlier. Moreover, CNC concentration was not shown as a significant influence parameter, i.e., the appropriate concentration required an analysis of the combination with the other factors.
The same analysis was performed considering the particle sizes, as shown in Fig. 9, verifying no linearity of the process parameters' actions. Each parameter possibly directly affects the others, e.g., the staging time and the homogenization speed are directly linked.
Considering the theoretical and experimental stability results, samples 5 (1% CNC, 12,000 rpm, 30% oil, 3 minutes), 6 (0.5% CNC, 10,000 rpm, 30% oil, 3 minutes), 9 (1% CNC, 10,000 rpm, 30% oil, 7 minutes) and 11 (0.5% CNC, 12,000 rpm, 30% oil, 7 minutes) stand out. The other parameters were analyzed together with the results previously presented, and the samples that presented better emulsification capacity, stability, and adequate droplet sizes, for CNC-emulsions were 9 (1% CNC, 10,000 rpm, 30% oil, 7 minutes) and 11 (0.5% CNC, 12,000 rpm, 30% oil, 7 minutes), having in common 30% essential oil and 7 minutes of preparation time. This result may be correlated with the linalool chemical structure, resulting in hydrogen bonds with CNCs, altering emulsion stabilization due to its free hydroxyls. For this reason, the highest times are necessary to guarantee an adequate CNC-adsorption onto oil/water interfaces since the system receives more energy during longer emulsification processes.
The same analysis was performed for CNF-emulsions, as shown in Fig. 10. Different from that observed for CNC emulsions, there is a well-defined pattern for each parameter influence. The highest stability values (red color) were observed for samples containing 1% cellulose nanofibers, i.e., it was necessary to use larger CNF amounts to achieve a stable and adequate three-dimensional network for this emulsions type, and, therefore, of the 8 compositions investigated, the DoE analysis decreases the initial number to 4 samples: 10 (1% CNF, 12,000 rpm, 30% oil, 7 minutes), 13 (1% CNF, 10,000 rpm, 30% oil, 3 minutes), 17 (1% CNF, 12,000 rpm, 20% oil, 3 minutes) and 24 (1% CNF, 10,000 rpm, 20% oil, 7 minutes). The concentration of 1% was associated with the necessary concentration to reach a critical level to guarantee long-range hydrodynamic interactions between the fibers, generating a percolation path and ensuring that the suspension passes to the gel state and that most drops were well coated, avoiding their coalescence.
Another parameter that influences emulsion stability is the homogenization speed, and the highest stability values and the trend of more stable emulsions were observed for 12,000 rpm. The high homogenization speed reflects the need for a high amount of energy. The oily phase dissociation and droplet reorganization into smaller droplets are necessary for cellulose nanofibers' dispersion (long and flexible fibers) and forming a gel-like structure. By the response surface graphs, it is verified that lower rotations do not provide enough energy for efficient homogenization and a possible adequate droplets coverage, indicating that the system energy is an essential parameter to ensure a stable emulsion, being associated with the free energy necessary to overcome the phase separation energy barrier.
Figure 11 presents the response surface graphs obtained for the droplet size for the evaluation of process variables. Similar to that observed for stability, the main parameters are speed, and at 10,000 rpm, the largest drop sizes were observed (i.e., they have the greater surface area and lower coverage of the interface containing CNFs). The essential oil again presented itself as a parameter without significant influence. The most stable samples were 10 (1% CNF, 12,000 rpm, 30% oil, 7 minutes) and 17 (1% CNF, 12,000 rpm, 20% oil, 3 minutes).
Figure 12 presents the main results obtained for the emulsions studied in this work, considering the processing parameters.