assessment of ultrasonic sprays for spray drying -...

17th International Symposium on Applications of Laser Techniques to Fluid Mechanics Lisbon, Portugal, 07-10 July, 2014

Assessment of ultrasonic sprays for spray drying

Miguel Oliveira Panão1, António Luis N. Moreira2, João Vicente3 and Eunice Costa3

1 ADAI-LAETA, Mechanical Engineering Department, University of Coimbra, Rua Luis Reis Santos, 3030-788 Coimbra, PORTUGAL

2 IN+, Center for Innovation, Technology and Policy Research, Instituto Superior Técnico, Av. Rovisco Pais, 1049-001 Lisboa, PORTUGAL

3 Hovione FarmaCiencia SA, Sete Casas, Loures, 2674-506, PortugalCorresponding author: [email protected]

Abstract Ultrasonic atomization is a strategy used in industrial spray drying to produce droplets of nearly uniform size, regardless of volumetric flow rate and solvent. In this work, an ultrasonic spray is characterised to assess for spray drying considering different solvents (water, 10% ethanol/water, 20% ethanol/water, acetone). A Phase Doppler interferometer is used to characterise droplet size and velocity. The analysis of experimental data evidence that the volumetric flow rate induces some variability in the mean drop size, but liquid properties also play a non-negligible role. Liquid properties, particularly through the relation between density and surface tension (normalized by the value for water), ρ/σ*, generate different effects of the volumetric flow rate in droplet velocity. Namely, a higher ρ/σ* implied that higher flow rates produce lower inertia sprays, thus, more prone to interact with the drying airflow fluid structures. Finally, the spray uniformity index based on the normalised Shannon information entropy is suggested as a more reliable parameter to assess it than the relative span.

IntroductionSpray drying is an essential technique in particle engineering for obtaining high performance pharmaceutical products, namely, direct compressibles (for oral dosage forms) or respirable particles (for inhalation products). The process consists in the formation of a spray from a feed solution containing an Active Pharmaceutical Ingredient (API) and excipients inside a chamber in which a stream of hot gas causes evaporation of the solvent and precipitation of the solids forming a particle. For the preparation of inhalation drug products, these solutions are usually dilute and since the atomization process depends on the properties of the liquid, the solvent becomes the determining factor for droplet size and distribution. However in oral applications, both API and excipients (particularly when polymers are used) alter the feed solution properties, particularly viscosity [1]. Despite these alterations, the major effect exerted on atomization is related with the solvent used.

In spray drying, the relation between particle and droplet sizes can be approximated to equation (1). Namely, the aerodynamic diameter of particles, which is diameter of a unit-density sphere that has the same settling velocity as the measured particle, is primarily determined by feed solution concentration (cF) and initial droplet size (dD) by a mass balance [2]:

! of !1 9


(1)

where is the average particle apparent density and ρD is the liquid droplet density. Other parameters that affect the correlation between the droplet and particle sizes include the drying kinetics (temperature profile, residence time, and gas-droplet mixing pattern) and actual feed composition, being outside of the scope of this study. Nevertheless, changes in the size distribution of droplets cause a significant change in the final properties of the particles. Therefore, that knowledge and how are dynamic characteristics affected by operating conditions is important and one of the motivations for the present work.

Furthermore, in spray drying, a control parameter is the spray uniformity, because particles should ideally be of the same size. A narrow size distribution is a critical parameter for ensuring a good powder flowability and compressibility in preparing oral dosage forms and a well-defined aerodynamic performance for inhalation products. The standard index for assessing the spray uniformity, and used in the pharmaceutical industry, is the relative span (∆):

(2)

where dk is the quantile of the cumulative size distribution and k ∈ [0; 1] (see [3]). However, quantiles are discrete quantities that express some information about the distribution, but are limited in terms of expressing information about the entire distribution. This is why an new approach to assess spray uniformity is explored based on Information Theory, through the Shannon entropy (H), which is a measure of the uncertainty associated with the distribution itself and is obtained by

(3)

where Nbins corresponds to the number of classes in the discrete probability distribution, and pi is the probability associated with class i. This value can be normalized (Hn) by the maximum information entropy corresponding to a uniform distribution function, within the range of drop sizes measured, and given by Hmax = ln(Nbins). If the distribution is single class, i.e. all droplets have the same size, there is only one class and its probability is p = 1, thus, Hn = 0. On the other hand, if every class i has the same probability of occurrence (like in a uniform distribution), Hn = 1. The evolution of this value and ∆ in terms of interpretation is similar.

⇢p

� =d0.9 � d0.1

d0.5

H = �NbinsX

i=1

pi ln(pi)

! of !2 9


The work presented here considers an ultrasonic atomization process to produce droplets from several solvents that are used in spray drying. The ultrasonic nozzle is an atomization system with the ability to form large droplets for relatively small flow rates (<50 ml/min), being an ideal laboratorial/pilot scale atomizer for generating product with a particle size similar to the one obtained in commercial-scale spray dryers. The analysis is limited to solvents atomization without the API or any excipients, in order to investigate the isolated effect of solvents properties on ultrasonic atomization, as well as the effects exerted by different flow rates.

Operating Conditions and Diagnostic Techniques

The atomizer is an ultrasonic one used in actual laboratorial/pilot scale spray dryers. The flow rate of solvent is varied between 5 and 50 ml/min. The fluids considered are water, acetone, and water-ethanol mixtures of 10% and 20%. The power supplied to the atomizer has been set to 1.6W for a fixed ultrasonic frequency of ≈35Hz. Table 1 summarizes the measurement locations, flow rate conditions and estimated thermophysical properties of the liquids used. The analysis performed in the following section is organized as follow:

• effect of volumetric flow rate at center axis on:‣ mean drop size;‣ mean drop velocity;‣ spray uniformity.

The size and axial velocity of droplets has been measured using a Dantec Phase Doppler Interferometer using a scattering angle of 30º. The laser power is set to 350mW, and only one velocity component is measured with the wavelength of 514.5nm. The beam spacing is 60mm and both transmitting and receiving lens have a 500mm focal length. The

transmitting optics unit is a 55X, the unit receiving the light emitted by particles is a

! of !3 9

Fig. 1 (Left) Droplet mean size for several measurement planes below the atomizer tip (z = 5, 10, 20 mm); (right) longitudinal profiles of the average velocity of droplets.


57×10 PDA and the BSA P80 processor is used to process signals and obtain the information on droplets characteristics.

Table 1 Summary of solvents used, measurement grid, operating conditions and liquid properties

Results and DiscussionThe size and velocity of droplets have been measured according to the measurement grid indicated in Table 1 for several planes below the ultrasonic atomizer tip. With a larger distance between the atomizer tip and the measurement plane, the number of points in the grid increases. Fig. 1 shows the results obtained for the several planes in terms of Sauter Mean Diameter (SMD or d32). The SMD is used because it is a moment of a drop size distribution that is correlated with the heat transfer that eventually leads to the desired

Fluid (x, y) [mm] z [mm]Volumetric flow

rate [ml/min]Liquid properties

@ 35ºC

water

typical pattern

x ∈ [-2; 2]; y ∈ [-2; 2] δx = 2; δy = 2; center @ (0,0)

5 5density (= 994 kg/m

kinematic viscosity (ν)

= 7.24×10

surface tension (σ) = 70.4 mN/m

x ∈ [-4; 4]; y ∈ [-4; 4] δx = 2; δy = 2; center @ (0,0)

10 5

x ∈ [-6; 6]; y ∈ [-6; 6] δx = 2; δy = 2; center @ (0,0)

20 5

water

(0,0) 20 5, 10, 20, 30, 40, 50

acetone

ρD = 773.4 kg/m

ν = 4.72×10σ = 21.8 mN/m

water - ethanol (10%)

ρD = 973.8 kg/m

ν = 7.67×10σ = 65.5 mN/m

Estimation by weighted average

water - ethanol (20%)

ρD = 953.6 kg/m

ν = 8.1×10σ = 60.61 mN/m

Estimation by weighted average

! of !4 9


micro-encapsulation forming the particle. Statistical errors in the SMD are less than 4.3% considering an information-theory approach [4].

Although in the center, the average drop size of droplets in relatively larger, the variation between the minimum average size and the several measurement points is 22.6% for z = 5mm, 32.3% for z = 10mm and 26.9% for z = 20mm. In the following section, the analysis is focused on the plane z = 20mm and (x, y) = (0, 0). A preliminary comparison between the distribution obtained at that location, and the distribution containing data from all measurement points, at all measurement locations, evidence a deviation of ±3.35% for the mean drop size, deviation of ±8.4% for the Sauter mean diameter and, considering that the normalized Shannon entropy depends on the distribution itself, it has been observed a deviation of ±2.58% of that value. Therefore, it can be assumed that drop size distributions measured at (x, y, z) = (0, 0, 20) are adequately representative of the entire spray.

Effect of volumetric flow rate on the mean drop size

The results evidence the effect of the volumetric flow rate on the Sauter Mean Diameter (SMD or d32) and plots in Fig. 2 show some induced variability between 14.2% for a water-10%ethanol mixture, and 35% for acetone. With different fluids, ultrasonic atomization produces droplets within a similar range, and follow similar patterns in the evolution with flow rate. Initially, the size mildly grows and afterwards, either decreases (water and water-10% ethanol mixture), or tends to stabilize (acetone and water-20% ethanol mixture).

While between water and acetone, the range of SMD is similar, introducing a percentage of ethanol in water changes the result in terms of drop size, implying that liquid properties play some role in the final outcome of atomization. This implies a certain dependency between drop size range and the solvent used that should be taken into account when defining particle size in the design of a micro-encapsulation drying process.

! of !5 9

acetonewater

Fig. 2 Effect of the volumetric flow rate on the Sauter Mean Diameter.


Effect of volumetric flow rate on the mean velocity and fluctuation

Usually, an increase of the volumetric flow rate is expected to induce an increase in droplet velocity (Fig. 3). However, two behaviors have been measured: 1) a monotonic increase (water; water-ethanol:10%); 2) or a curve with a maximum (acetone; water-ethanol:20%). Also, the dispersion of velocity values given by the standard deviation increases with the flow rate. The reason for using different liquids is to investigate the effect of liquid properties on the atomization outcome.

Since atomization forces are predominantly inertia and surface tension, the dimensionless number that best expresses this balance is the Weber number, which for a droplet can be formulated as

�

In this case, the Sauter mean diameter (d32) is considered, as well as the mean droplet velocity. Between fluids, considering the droplet Weber number, what changes is the ratio between density and surface tension (ρ/σ). Thus, this parameter is used for comparison, relatively to a certain ratio taken as reference, (ρ/σ)reference. In the present work, water is considered as the reference fluid. Therefore, the parameter expressing the fluid properties is defined by ρ/σ obtained for each fluid normalized by the reference:

Fig. 4 shows the evolution of droplet Weber number, as a function of the volumetric flow rate, considering different fluids. It is noteworthy that, ρ/σ* increases from water, to a water-ethanol mixture to acetone. Considering the thermophysical properties indicated in Table 1, from water, to a mixture with 10% and 20% of ethanol, to acetone, individually, both density and surface tension decrease. Therefore, the increase observed for ρ/σ* evidences that a variation of surface tension between fluids is predominant over the variation of density.

WeD =⇢U2

dd32�

! of !6 9

acetonewater

Fig. 3 Effect of the volumetric flow rate on droplet mean velocity at (x, y, z) = (0, 0, 20) mm. Bars indicate the velocity fluctuation.


In Fig. 4, it is also clearer by the similarity between the evolution of Wed and Ud, that the effect of the volumetric flow rate is more important on droplet inertia than its size for higher ρ/σ*. Usually the relation between the particle response time and the airflow timescale is several orders of magnitude smaller than unity. Therefore, particles usually follow the main airflow. This result for the WeD is particularly important if droplets are trapped in the recirculating flow in the outer region where flow velocities are lower and smaller turbulent structures are present during spray drying, haven smaller timescales, increasing the relation between droplet and airflow timescales. Thus, in this context, when using an ultrasonic atomizer, a higher ρ/σ*, and volumetric flow rate, leads droplets to interact more with the airflow, and if some droplets are transported near the wall, the probability of deposition is higher, which would jeopardize the spray drying process.

Effect of volumetric flow rate on spray uniformity

An important issue in spray drying is whether the spray produced is able to form droplets of a relatively uniform size, in order to dry into particles, as equally sized as possible. Therefore, the assessment of the spray uniformity is essential. As aforementioned in the introduction, two methods have been considered. The most used is the relative span (∆), based on the size range between 10% and 90% of the volume spray, and the median mean diameter, using the cumulative size distribution. The second method relies on the on the entire distribution itself. It is based on an information theory approach and uses the normalized Shannon entropy in the analysis.

Fig. 5 depicts the results obtained for these two methods considering different fluids and the effect of the volumetric flow rate is assessed. While the relative span has a local maximum, the normalized Shannon entropy suggests that the polydispersion degree of drop sizes stabilizes for volumetric flow rates above 20 ml/min.

It is noteworthy that the relative span for water as solvent does not vary significantly in the cases of 5ml/min and 40ml/min, and the same occurs for acetone in the cases of 5ml/min and 50ml/min. The histograms for both cases are depicted in Fig. 6.

! of !7 9

Fig. 4 Effect of the volumetric flow rate and surface tension, expressed in ρ/σ*, on droplet Weber number at (x, y, z) = (0, 0, 20) mm.


It is clear that drop size distributions for the lowest flow rate are more uniform. However, the relative span values are similar. On the other hand, the result provided by Hn(d) based on information theory appears to be more consistent with the spray uniformity degree. If Hn(d) is decreasing toward zero, the distribution has a higher uniformity of drop sizes, and if it approaches unity, the distribution is closer to the case of a uniform distribution function, where all classes have the same probability of occurrence, for the range of drop sizes considered. The examples in Fig. 6 evidence how the relative span (∆) decreases with an increase of the flow, when the distribution characterizing drop sizes has actually less uniformity. This does not occur for the normalized Shannon Entropy, thus suggesting the later as a more reliable criterion for assessing spray uniformity than the relative span.

An additional detail corresponds to the size distributions obtained for 5ml/min between water and acetone where spray uniformity has a higher degree. The class with the highest probability in the distribution has a larger value for water than for acetone, indicating than the former has a slightly larger spray uniformity than the later. The values obtained for Hn(d) indicated in Fig. 6 support that observation, while the values of the relative span indicate otherwise.

Finally, as depicted in Fig. 7, the ultrasonic spray uniformity does not seem to be significantly affected by liquid properties, but it depends to some extent on the volumetric flow rate until 20ml/min, above which it appears to attain stabilisation.

! of !8 9

Fig. 5 Effect of the volumetric flow rate on spray uniformity at (x, y, z) = (0, 0, 20) mm.

acetonewater

Fig. 6 Histograms of drop size

acetone


Concluding remarks

Ultrasonic atomization is used in industrial spray drying with the purpose of producing droplets of almost the same size regardless of volumetric flow rate. The purpose of this work is to use the Phase Doppler diagnostic technique to assess the actual effect of the volumetric flow rate in the characteristics of the spray, using different solvents and the uniformity degree of size distributions.

The experimental results evidence that:

• the volumetric flow rate induces some variability in the mean drop size, but liquid properties also play a non-negligible role;

• liquid properties, particularly through the relation between density and surface tension (normalized by the value for water), ρ/σ*, generate different effects of the volumetric flow rate in droplet velocity. Namely, a higher ρ/σ* resulted in a local maximum, implying that higher flow rate produce sprays with lower inertia, thus, more prone to interact with the drying airflow fluid structures.

• the spray uniformity index based on the normalised Shannon information entropy is suggested as a more reliable parameter to assess it than the relative span, because it relies on the entire distribution itself instead of the size associated with a percentage of the spray droplets volume, which are less sensitive to eventual multimodality effects of the distribution.

References[1] Vehring R, “Pharmaceutical particle engineering via spray drying”, Pharmaceutical Research, 25 (2008), pp. 999-1022.[2] Gil M, Vicente J, Gaspar F, "Scale-up methodology for pharmaceutical spray drying", chimica oggi / Chemistry Today, 28 (2010), pp. 18-22.[2] Lefebvre A, Atomization and Sprays, Hemisphere Publishing Corporation, 1989.[3] Panão MRO, Moreira ALN (2008) A real-time assessment of measurement uncertainty in the characterization of sprays, Measurement Science and Technology, 19, 095402.

! of !9 9

Fig. 7 Overall results for spray uniformity assessment based on an information theory approach.

assessment of ultrasonic sprays for spray drying -...

Documents