Oil droplet breakup during pressure swirl atomization of food emulsions: Influence of atomization pressure and initial oil droplet size

Correspondence Martha L. Taboada, Institute of Process Engineering in Life Sciences, Chair of Food Process Engineering, Karlsruhe Institute of Technology, Kaiserstrasse 12, 76131 Karlsruhe, Germany. Email: martha.taboada@kit.edu Abstract Atomization of emulsions with pressure swirl atomizers is a common task in food process engineering. Especially in spray drying processes for food materials like dairy products, it is the technology of choice. During atomization, emulsions are subjected to high stresses, which can lead to deformation and breakup of the dispersed droplets. In this study, the influence of atomization pressure (5–20 MPa) and initial oil droplet size (0.26, 3.1, and 20.8 μm) on the oil droplet breakup during atomization of food based oil-in-water emulsions with pressure swirl atomizers was investigated. It was shown that a significant oil droplet breakup takes place upon atomization. The size of oil droplets with an initial value of 3.1 and 20 μm was reduced up to 0.36 μm. No breakup of oil droplets with an initial value of 0.26 μm was observed. The breakup was highly dependent on the atomization pressure. The results were analyzed based on existing knowledge on droplet breakup in laminar flow. A concept to estimate capillary numbers during atomization was developed based on common models from different applications. The results of this study can be used to control the resulting oil droplet size after atomization with pressure swirl atomizers.


| INTRODUCTION
Spray drying of emulsions is a common task in food engineering for the production of products with encapsulated oily components. Typical examples include products such as coffee creamers, infant formula, and the encapsulation of active ingredients, aroma, and coloring compounds (Reineccius, 2004). The process of spray drying starts with the atomization, by which the liquid emulsion is dispersed into small spray droplets. These droplets are subsequently dried to powder by contact with a hot air stream. Pressure swirl nozzles are widely used as atomization devices in the food industry (Barbosa-Cánovas et al., 2005).
After drying, the oil droplets should be encapsulated in a matrix material, which acts like a barrier, providing protection against oxidation or losses.
During atomization, emulsions are subjected to intense stresses, which do not only deform and atomize the feed, but can also lead to deformation and breakup of the dispersed droplets therein. A breakup of the dispersed oil droplets results in a modification of a previously adjusted oil droplet size distribution (ODSD). The ODSD affects the stability of the powder and of the reconstituted emulsion, as well as the functional properties of the product. For example, the release and bioavailability of active compounds are directly related to the oil droplet size (McClements & Li, 2010). In addition, the oil droplet size determines the color impression of food coloring powders (Haas et al., 2019). Furthermore, oil droplet breakup during atomization has been correlated to a reduced encapsulation efficiency in the powder (Jafari et al., 2008). Therefore, it is of upmost importance to control oil droplet breakup during atomization.
Breakup of oil droplets during atomization of oil-in-water (O/W) emulsions has been already studied for different types of atomizers: Schröder et al. (2012) and Kleinhans et al. (2016) studied oil droplet breakup for atomization with effervescent atomizers and an air core liquid ring atomizer. Munoz-Ibanez et al. (2015) studied oil droplet breakup during atomization with rotary and external mixing pneumatic atomizers. In these studies, the breakup was shown to depend on the energy input of atomization, as well as on the initial oil droplet size and on the viscosity ratio of the emulsions. Few studies using pressure swirl atomizers have also reported breakup of the disperse phase during atomization of emulsions (Bolszo et al., 2010). However, in spite of their wide industrial use, this aspect has not been systematically studied yet for pressure swirl atomizers. Most of the studies found in literature on these atomizers focus on the spray characteristics and not on the changes of the disperse phase (Davanlou et al., 2015;Tratnig et al., 2009).
The atomization principle of pressure swirl atomizers is based on the conversion of pressure to kinetic energy. In this type of atomizers the liquid flows through tangential holes or slots into a swirl chamber, and then to a discharge orifice (Walzel, 2003). Due to swirling motion of the liquid, an air core is created that extends from the rear of the swirl chamber to the discharge orifice. In the orifice, a thin liquid film is formed, which then leaves the atomizer in the form of an annular sheet that spreads to a conical hollow spray (Lefebvre & McDonell, 2017). A schematic view of a pressure swirl atomizer is depicted in Figure 1. Acceleration of the liquid due to the diameter contraction is expected to result in elongational stresses in both radial and axial directions. In the liquid film at the atomizer orifice, high shear stresses are expected due to the high velocities and the proximity to the wall. By means of numerical simulations of the internal flow in commercial pressure swirl atomizers, Renze et al. (2011) demonstrated that shear rates up to $100,000 s −1 and elongational rates up to 50,000 s −1 occur in the liquid film close to the nozzle exit for pressures of 0.2 MPa. These stresses can also lead to deformation and breakup of the disperse droplets in emulsions.
In order to estimate the stresses in the liquid film at the atomizer orifice, knowledge of the film thickness t is required. This information is not readily available and is not easy to determine experimentally at relevant industrial conditions. Several analytical and empirical correlations are available in the literature to estimate this value. A widely used theoretical model for the estimation of t is given by Suyari & Lefebvre (1986) in the form of Equation (1), where r o corresponds to the nozzle orifice radius, _ m to the mass flow rate, μ to the liquid viscosity, ρ to the liquid density and Δp L to the pressure differential during atomization. According to the original correlation by Rizk & Lefebvre (1985), the constant C is 3.66. The constant was corrected by Suyari & Lefebvre (1986) to 2.7 to better match experimental data. The correlation predicted the film thickness with high accuracy up to pressures of 3 MPa. Other recent studies with different atomizer geometries, pressures, and liquid properties have found that the expression estimates the liquid film thickness fairly well (Laurila et al., 2019;Wimmer & Brenn, 2013).
From the theory on breakup of dispersed droplets it is known that for droplet breakup the external forces imparted by the surrounding fluid must overcome the droplet capillary pressure (Karbstein & Schubert, 1995). When the external stresses are of simple shear nature, the breakup is characterized by the capillary number Ca: F I G U R E 1 Schematic representation of a pressure swirl atomizer where μ c is the viscosity of the continuous phase, _ γ the shear rate, x the droplet radius, and σ the interfacial tension between the continuous and the disperse phase. In the case of elongational flow, the shear rate is replaced by the elongational rate _ ε in Equation (2). For breakup to occur, a critical value of the capillary number Ca cri has to be exceeded (Grace, 1982). This value depends on the viscosity ratio between the disperse and the continuous phase μ d /μ c . The denominator is replaced with the emulsion viscosity for emulsions with high disperse phase fractions (Armbruster, 1990). The correlation of Ca cri with the viscosity ratio depends on the type of flow acting on the droplet interface. Grace (1982)  Ca cri from Grace (1982).
For droplet breakup, it is also necessary that the droplet deformation time τ def exceeds a critical value τ def,cri, that correlates with the droplet viscosity μ d divided by the deformation stress, see Equation (3) (Walstra & Smulders, 1998): In emulsions with high phase content, the resulting droplet size is not only a function of droplet breakup, but also of coalescence. In this study, the effect of coalescence was excluded by working at very low disperse phase fractions.
The aim of the present work was to investigate the impact of pressure swirl atomization on the oil droplet size of food emulsions.
Specifically, the influence of the atomization pressure and the initial oil droplet size were investigated. For this purpose, the ODSD of emulsions before and after atomization were compared. Additionally, stresses in the atomizer and capillary numbers were estimated in order to analyze the results based on the theory of droplet breakup in laminar flow.
In the inspected shear rate range the viscosity of the emulsions and of the oil were found to be independent of the shear rate. The viscosity of the MCT oil at 1000 s −1 was 28.8 ± 0.2 m PaÁs, while the viscosity of the emulsions was 32.3 ± 1.3 m PaÁs. The viscosity of the emulsion was used to calculate the viscosity ratio, as the viscosity of the emulsion and of the continuous phase are virtually the same. Thus, the viscosity ratio of the model system was 0.9, which is in the optimal region for droplet breakup in shear flow (Grace, 1982 Table 2. To ensure that the filter and the pump periphery have no effect on the initial ODSD, emulsion samples were taken right before the nozzle entry. No significant difference was observed between the ODSD of these samples and of the initial emulsions. During atomization, a sample of the spray was taken with a beaker $25 cm below the nozzle exit. The oil droplet size of the emulsion was measured offline with laser diffraction spectroscopy (HORIBA LA950, Retsch Technology
After atomization, the size distribution of the spray droplets was measured. Spray droplets are atomized emulsion droplets in which the oil droplets are dispersed (see Figure 1). Spray droplet size distributions were measured inline using a similar setup as in previous studies (Kleinhans et al., 2016). The spray rig was equipped with a laser dif- were processed according to the Fraunhofer theory and a time averaged mean value was calculated.

| Oil droplet size after atomization
To assess the influence of the atomization pressure on the oil droplet Therefore, the stresses that lead to oil droplet breakup are expected to increase with increasing atomization pressure.
It can also be noted from the results shown in Figure 3 that very similar ODSD resulted after atomization of emulsions with SMD i of 3.1 and 20.8 μm at each studied pressure. From these results it is clear that the oil droplets are broken up to the same value independently of their initial droplet size. This effect can be further seen in Figure 4, where the resulting SMD for these emulsions are depicted. Analysis of variance (ANOVA) was carried out to compare the resulting SMD for each pressure, and no significant difference (p < .05) between the SMD of emulsions with different initial droplet size was observed. To further investigate the influence of the initial oil droplet size on the breakup behavior during atomization, emulsions with SMD i of 0.26 μm were also atomized. The resulting SMD are also depicted in Figure 4. In the case of these submicron droplets, the SMD remained unchanged at all studied atomization pressures, indicating that no breakup of the oil droplets took place during atomization. The results imply that in the case of the small, submicron droplets, the capillary pressure is large enough to overcome the external stresses during atomization.
In the emulsification literature, the SMD of the disperse phase correlates with the energy input for emulsification according to the expression in Equation (4). In this equation, C is a constant that depends on the viscosity, and the exponent b gives insights on the breakup mechanisms of the disperse phase: for breakup due to laminar stresses, b takes a value close to one. For breakup due to inertial (turbulent) stresses, b takes values between 0.2 and 0.4 (Karbstein, 1994).
To evaluate the breakup mechanisms of the oil droplets during atomization, the resulting oil SMD were correlated to the expression in Equation (4). In this study, the energy input for emulsification corresponds to the atomization pressure. The resulting constants C and b for the emulsions with different SMD i , as well as the coefficient of determination R 2 are summarized in Table 3. The resulting fit is also depicted in Figure

| Spray droplet size
The resulting SMD of spray droplet size distributions at the different atomization pressures are also depicted in Figure 4 for emulsions with SMD i of 3.1 μm. No significant difference in the spray SMD was observed for emulsions with different SMD i at the same pressure (data not shown). A reduction of the spray SMD with increasing atomization pressure is observed. However, it is noticeable that the effect of increasing pressure on the spray droplet size is in relation much lower than on the oil droplet size. In fact, an increase in the atomization pressure from 5 to 20 MPa resulted in a reduction of the oil SMD by 74%, while the SMD of the spray droplets was reduced only by about 26%.
To evaluate the breakup mechanism of the spray droplets, the resulting spray SMDs were also correlated to the expression in Equation (4). This expression has been widely used to correlate the SMD of spray droplets with the atomization pressure in pressure nozzles (Lefebvre & McDonell, 2017;Stähle et al., 2017). In this case, b usually takes values between 0.27 and 0.4 (Lefebvre & McDonell, 2017). The resulting constants C and b for the spray droplets, as well as the coefficient of determination R 2 are listed in Table 3. The resulting fit is depicted in Figure 4. In the case of spray droplets in this study, b takes a value of 0.22, which is an indicative of breakup in turbulent flow (Karbstein, 1994). The results imply that different mechanisms underlie the breakup of oil and spray droplets.
From the literature on pressure swirl atomization it is known that spray droplets are generated due to the high relative velocity between the liquid and the gas outside of the atomizer. The liquid leaves the atomizer as a conical sheet and disintegrates into spray droplets by Kelvin-Helmholtz instabilities or by turbulence (Walzel, 2003). In the case of oil droplets, laminar shear stresses inside the atomizer, and specifically in the thin liquid film before the atomizer outlet, are expected to dominate the droplet breakup.

| Estimation of stresses and capillary numbers
To explain the observed dependences of the oil droplet breakup on the atomization pressure and on the initial oil droplet size, the laminar stresses in the atomizer and the capillary numbers are estimated. The dominant stresses leading to oil droplet breakup are expected to occur in the thin liquid film at the atomizer outlet. Shear and elongational stresses can also occur in the slots of the slotted core, as well as in the swirl chamber. However, the named stresses are expected to be much lower in comparison to the stresses in the liquid film close to the atomizer exit (Nonnenmacher & Piesche, 2000;Rezaeimoghaddam et al., 2010). At this point, the small thickness of the liquid film leads to very high liquid velocities and velocity gradients.
In the liquid film, shear stresses are expected to dominate. A study of the flow inside similar pressure swirl nozzles confirms this: Renze et al. (2011) showed that the elongational rates in the liquid film at the nozzle outlet had a magnitude of half of the shear rates. It should be noted, however, that the critical capillary numbers for elongational flow are much lower than in shear flow (Grace, 1982).
Therefore, oil droplet breakup in elongational flow requires reduced stresses compared to shear flow.
A schematic drawing for the liquid film in the orifice of the pressure swirl nozzle (as in Figure 1) is depicted in Figure 5. A model of the flow profile for the estimation of the shear rate in the liquid film is also depicted. A simplified linear flow profile is assumed, in which the liquid velocity at the wall u w is zero and the maximum velocity u a occurs at the air-liquid interface. In reality, the maximum liquid velocity occurs probably somewhere before the interface, as the liquid is slowed down by the air. This discrepancy is however not expected to change the rough magnitude of the shear rate estimation. Another important assumption is, that due to the high magnitude of the axial T A B L E 3 Constants C and b as well as coefficients of determination R 2 for oil and spray droplets for the fit using Equation (4) F I G U R E 5 Model for the flow profile in the liquid film in the atomizer orifice. r o : nozzle inner radius; t: liquid film thickness; u a : velocity at the interface with the air core; u w : velocity at the wall velocity, the radial velocity in the liquid film can be neglected (Rezaeimoghaddam et al., 2010).
The definition of the shear rate _ γ is presented in Equation (5), while the solution for our model system is presented in Equations (6) and (7). In these equations u is the axial velocity and y is the coordinate perpendicular to the flow direction. t corresponds to the thickness of the liquid film. To solve Equation (6), the following boundary conditions are applied: the velocity at the wall u w is equal to zero, while the velocity at the interface with the air core u a is twice the average velocity u of the liquid.
The average velocity u is calculated according to Equation (8), in which Q L corresponds to the experimentally measured volume flow and A L is the flow area of the liquid. The flow area is calculated from the area of the nozzle orifice minus the area of the air core (Equation (9)), in which r o corresponds to the radius of the nozzle Solving Equation (9) requires knowledge of the liquid film thickness inside the atomizer. Equation (1) has been used to estimate the liquid film thickness in this study with a constant C value of 2.7.
Unfortunately, no study has been found in literature, which validates the use of this correlation in the pressure range of this study. However, the correlation is based on a theoretical analysis of the flow conditions in the atomizer, which are expected to be valid at high pressures too.
The estimated film thickness, mean velocity, and shear rate in dependence of the atomization pressure are summarized in Table 4.
As expected, the film thickness decreases and the velocity of the liquid increases with increasing pressure. By this, an increase in the calculated shear rate with increasing atomization pressure is observed.
The relatively high values of velocities are expected for atomizers with small orifices, as in this study (Wimmer & Brenn, 2013). The corresponding capillary numbers for shear flow for the different initial oil droplet sizes and atomization pressures were calculated by means of Equation (2) and are summarized in Table 5. These values are compared to the critical capillary number from Grace (1982) for droplet breakup in shear flow, also shown in Table 5. The estimated capillary numbers for the emulsions with a SMD of 3.1 and 20.8 μm are, for all atomization pressures, well above the critical capillary number. Therefore, oil droplet breakup in shear flow is possible for these emulsions at the studied atomization conditions. In the case of emulsions with a SMD of 0.26 μm, the capillary numbers are below the limits of droplet breakup in shear flow calculated by Grace (1982). In the case of the smaller oil droplets, the shear stresses during atomization are not high enough to overcome the capillary pressure and no oil droplet breakup is possible.
According to the work of Renze et al. (2011), the elongational stresses in the liquid film at the outlet of the atomizer are estimated to have a magnitude of half of the shear stresses. Based on this information and on the estimated shear rates listed in Table 4, the capillary numbers for elongational flow were also estimated and are listed in Table 5. Similar to the previous analysis on shear flow, the capillary numbers are compared to the critical capillary number from Grace (1982) for elongational flow. From Table 5 it can be seen, that for emulsions with a SMD i of 3.1 and 20.8 μm the estimated capillary numbers are well above the critical capillary number for all atomization pressures. Therefore, in spite of the lower magnitude of the elongational stresses, oil droplet breakup due to elongational flow is also possible for these emulsions. The lower values of critical capillary numbers for elongational flow, compared to shear flow (Grace, 1982) explain this. In the case of emulsions with a SMD of 0.26 μm, the cap- conditions. These results may be explained by several reasons. First of all, the critical capillary numbers from Grace (1982)  To explain the fact that the same oil droplet size after atomization was achieved with emulsions with SMD i of 3.1 and 20.8 μm, the residence time in the high stress area was estimated. From the emulsion theory it is known that larger droplets require longer stressing times for reaching the deformation state that corresponds to the stress applied (Walstra, 1993). Therefore, to achieve the equilibrium value of oil droplet size, the residence time in high stress areas must be long enough to allow deformation and breakup of the large droplets. By means of Equation (3)

| CONCLUSION
The experimental study of atomization of O/W emulsions with pressure swirl atomizers showed that a significant oil droplet breakup takes place during atomization. The oil droplet breakup is highly dependent on the atomization pressure, as the stresses in the liquid film of the atomizer orifice correlate with the atomization pressure. The impact of the pressure on the spray droplet size is relatively low compared with the impact on the oil droplet size. These results have the practical implication that an increase in the atomization pressure to achieve an adequate spray droplet size for the spray drying process, will necessarily lead to a reduction of a previously adjusted oil droplet size in emulsions. The results also suggest that oil droplet breakup occurs under laminar flow conditions, whereas spray droplet breakup is dominated by turbulent flow. In addition, the results suggest that the stresses in the atomizer and the residence time of the droplets are large enough to reduce the SMD to submicron values, even when emulsions with large initial oil droplet sizes are atomized. Therefore, the oil droplet size after atomization can only be controlled to a limited extent with the initial oil droplet size.
A theoretical approach for the estimation of stresses and capillary numbers during atomization with pressure swirl atomizers was developed. The good agreement of the experimental results with the theory on droplet breakup in laminar flow supports the hypothesis, that the stresses in the liquid film at the atomizer outlet dominate oil droplet breakup during atomization. The results also indicate that both shear and elongational stresses can lead to oil droplet breakup under the studied conditions. To confirm this, detailed stress-time profiles in the atomizer should be analyzed.
The concept developed in this study for the estimation of the capillary numbers can be used to control oil droplet breakup during atomization under given process conditions. The findings suggest that to avoid oil droplet breakup, the emulsion properties and operating conditions must be adjusted to obtain capillary numbers below the critical values. This concept could be used as a tool to control oil T A B L E 5 Estimated capillary numbers and critical capillary number after (Grace, 1982) for shear and elongational flows writing-review and editing.