Atomization of Liquids Relevant to Pharmaceutical Tablet Coating. Experiments and Droplet Size Modelling

Transcription

1 Atomization of Liquids Reevant to Pharmaceutica Tabet Coating. Experiments and Dropet Size Modeing A. Aiseda 1,, E. J. Hopfinger 1, J. C. Lasheras 1, D. M. Kremer,*, A. Berchiei, E. K. Connoy 1 Department of Mechanica and Aerospace Engineering University of Caifornia, San Diego 9500 Giman Drive La Joa, CA Current Address: Department of Mechanica Engineering University of Washington Stevens Way, Box Seatte, WA Ora Products Center of Exceence Pfizer, Inc. Goba Research and Deveopment Groton/New London Laboratories Eastern Point Road MS Groton, CT * Corresponding Author Ora Products Center of Exceence Pfizer, Inc. Goba Research and Deveopment Groton/New London Laboratories Eastern Point Road MS Groton, CT Phone: (860) Fax: (860) emai: [email protected]

2 Abstract This paper describes a coaborative theoretica and experimenta research effort to investigate the performance of co-axia atomizers utiized in pharmaceutica tabet coating when atomizing common tabet coating soutions under typica processing conditions. In pharmaceuticay reevant appications, the iquids being atomized are typicay compex, non-newtonian fuids which may contain poymers, surfactants and arge concentrations of insoube soids in suspension. The goa of this investigation is to produce a vaidated theoretica mode capabe making timey predictions of atomizer performance in pharmaceutica tabet coaters. The Rayeigh-Tayor mode deveoped by Varga et a. has been extended to viscous and non-newtonian fuids starting with the genera dispersion reation deveoped by Joseph et a. The theoretica mode is vaidated using dropet diameter data coected with Phase Dopper Partice Anaysis. The primary output from the mode is the Sauter Mean Diameter of the atomized dropets, which is shown to compare favoraby with experimenta data. Critica mode parameters and pans for additiona research are aso identified.

3 1. Introduction The atomization of a iquid jet by a co-fowing, high-speed gas is a process of considerabe practica interest in many industria settings. In particuar, the atomization of iquids is utiized extensivey in a variety of pharmaceutica manufacturing processes. The roe of atomization in pharmaceutica manufacturing can be organized into two broady defined categories. Some pharmaceutica manufacturing processes utiize atomization to ater the in vivo performance of the active pharmaceutica ingredient (API), often by modifying the bioavaiabiity of the API itsef. A common manufacturing process of this type is spray drying. During spray drying, API and other excipients are dissoved in sovents and the soution is atomized in a heated gas stream and dried to form powders (Masters 1976). Research has shown that the size distribution of the atomized dropets couped with the operating parameters of the spray dryer can infuence the morphoogy of the dried powder (Lin and Gentry 003). Additionay, scae-up of the spray drying process is notoriousy difficut due to the inabiity of modes to predict atomizer performance at different scaes, especiay for pharmaceuticay reevant soutions (Kremer and Hancock 006; Oakey 004). Thus, scae-up of this process can resut in unanticipated changes in the size and morphoogy of the dried powder which can deeteriousy impact the downstream manufacturing steps necessary to produce the fina dosage form. Another exampe of a pharmaceutica manufacturing process that utiizes atomization to ater the in vivo performance of the API is spray congeaing. In this process, the API is mixed with waxes and atomized, normay via a rotary atomizer, with the goa of producing very sma partices containing encapsuated API (Kawase and De 198; Mackapow et a. 006). Encapsuation can modify the reease profie of the API or target dissoution of the encapsuated partice to specific regions of the gastrointestina tract. In the other category of appications, atomization is utiized to modify the appearance or improve the in vivo performance of the fina dosage form. The most common exampe of this type of process is tabet coating, with a recent survey indicating that ~30% of pharmaceutica tabets manufactured in 004 were coated (IMS MIDAS Database 005). There are a number of reasons why such a arge percentage of pharmaceutica tabets are coated, which adds an additiona unit operation to the manufacture of the fina dosage form. Non-functiona tabet coatings improve the appearance and handing of tabets and may protect against counterfeiting by improving brand recognition. Functiona tabet coatings are appied to mask unpeasant taste or ater the tabet dissoution profie either by controing the rate of dissoution, normay via semi-permeabe membrane coatings, or by protecting the tabet from the acidic environment of the stomach via enteric coatings. As is the case for spray drying, scae-up of the tabet coating process is difficut as the operation aso invoves severa couped physica processes occurring simutaneousy. In addition to atomizing the coating soution, the tabet coating process invoves mixing a bed of tabets as we as drying the coating soution on the surface of the tabets resuting in the fina soid coating. Page 1 of 17

4 Pharmaceutica researchers have deveoped thermodynamic modes to simuate the tabet coating process and guide scae-up; however these modes, whie usefu, make no attempt to predict atomizer performance at different scaes (am Ende and Berchiei 005). Atomization, and especiay of air-bast atomization, is a compex muti-parameter probem. It has, for this reason, euded a cear physica understanding and genera theoretica predictions of the dropet size as a function of the injector geometry and fuid properties. A physica mechanism that stands up to comparison with experimenta evidence is a two-stage instabiity mechanism, a primary shear instabiity foowed by a Rayeigh-Tayor (R-T) instabiity of the iquid tongues produced by the primary instabiity (Varga et a. 003). In this scenario, the iquid jet diameter is practicay irreevant; the thickness of the gas boundary ayer at the injector exit determines the waveength of the primary instabiity and the subsequent fuid mass that is suddeny exposed to the gas stream and acceerated. For ow viscosity fuids, the R-T waveength that determines the igament size and hence the drop size depends on surface tension, whie viscous effects are negigibe. In pharmaceuticay reevant appications, the iquids being atomized are typicay compex, non- Newtonian fuids which may contain poymers, surfactants and high concentrations of insoube soids in suspension. Tabet coating, regardess of the nature of the coating, and many pharmaceutica spray drying operations utiize co-axia air bast atomizers (Muer and Keinebudde 006). Athough the performance of co-axia airbast atomizers has been studied extensivey (Lasheras and Hopfinger 000; Varga et a. 003), very few of these investigations were focused on atomization of highy viscous or non-newtonian iquids (Mansour and Chigier 1995). In this paper, we describe a coaborative theoretica and experimenta research effort to investigate the performance of commercia co-axia atomizers utiized in pharmaceutica tabet coating when atomizing common tabet coating soutions under typica processing conditions. As such, the goa of this investigation is to produce a vaidated theoretica mode capabe making timey predictions of atomizer performance in pharmaceutica tabet coaters. Our theoretica study shows that for viscous or non-newtonian iquids, ike many common tabet coating soutions, the R-T waveength is strongy affected by the high viscosity or the non-newtonian behaviour of the soution. Joseph et a. (Joseph et a. 00) demonstrated this very ceary for viscoeastic iquid drops suddeny exposed to a high-speed gas stream. In this study, The R-T mode originay deveoped by Varga et a. (Varga et a. 003) is extended to viscous and non-newtonian fuids starting with the genera dispersion reation deveoped by Joseph et a.(joseph et a. 00). The theoretica mode is vaidated using data coected with Phase Dopper Partice Anaysis (PDPA). The primary output from the mode is the Sauter Mean Diameter of the atomized dropets, which is shown to compare favoraby with experimenta data. Critica mode parameters and pans for additiona research are aso identified. Page of 17

5 . Description of Experiment.1 Experimenta setup Atomization experiments were carried out using a Spraying Systems injector (1/8 JAC series with gas cap PA C, and iquid nozze PF8100NB) which has a we-characterized geometry shown in Fig. 1. The iquid was pressurized in a badder tank, fowed through a caibrated fow metre and injected through a sma orifice at the centreine of the nozze. Pressurized air was injected coaxiay with the iquid stream through an annuar gap ocated at the base of the iquid nozze. Between 10 and 0% of the pressurized air fowed through auxiiary ports ocated in the periphery of the gas cap and oriented at a 45 to the main iquid and gas streams (see Fig. 1 for detais). This pattern air induces an asymmetry in the veocity fied such that the cross section of the spray becomes eiptica. Additionay, the pattern air pays an important roe in the transport of sma iquid dropets inside the spray, which wi be discussed in greater detai beow. Further, because the pattern air merges with the main streams at a distance of more than ten iquid diameters downstream of injection, it pays a negigibe roe in the iquid atomization process which is dominated by a series of instabiities which form very cose to the iquid nozze discharge. A sketch of the atomizer, depicting the reevant characteristics and ength scaes, is as shown in Figure 1 beow. The air fow rate was measured by a fow metre and the outet pressure was measured by a pressure gage to correct for compressibiity effects at the fow-meter outet. The atomizer was secured to a two-dimensiona traverse system so that it coud be precisey positioned with respect to the measuring point aong the radia and axia coordinates of the spray. Figure presents a picture and sketch of the experimenta faciity.. Dropet size and veocity measurements The veocity and size of the dropets produced during atomization were measured by a Phase Dopper Partice Anayzer (TSI Inc., Minneapois, MN). A detaied description of this measurement technique can be found esewhere (Bachao 1994). Briefy, the nm beam from an Argon ion aser was spit and one of the beams passed through a Bragg ce which produced a 40 MHz frequency shift. These two beams were then transported through fiber optics to the experimenta setup where they cross, forming an interferometry fringe pattern at the probe voume. Light scattered from the dropets crossing through the beams intersection was acquired at three distinct points by the receiver and processed by three photodetectors. The frequency and phase shift in the signa were extracted to compute the dropet veocity and diameter, respectivey. In these experiments, the receiver was paced at a 30º ange with the transmitter to coect backscattered ight and a 150 µm sit was used in order to reduce the probe voume size. With the current optica setup the probe voume was 110 µm in diameter and 55 µm ong, and the resoution of the system aows the detection of dropets down to 1.5 µm in diameter. Page 3 of 17

6 The PDPA system was positioned in such a way that the measurement voume was ocated on the pane where the injector nozze evoved. By using the two degrees of freedom of the traverse system, the injector was moved reative to the probe voume. Thus, measurements were taken aong different radia and axia positions within a pane that cut diametricay across the spray. The origin of this pane was ocated at the center of the iquid nozze discharge, with the orientation of the coordinate system as indicated in Fig. 1. The axia veocity and size of individua dropets fowing through the probe voume were measured and statisticay anayzed. The arithmetic mean veocity of the dropets and the Sauter mean diameter (SMD) of the dropets were computed directy from the raw measurements using MATLAB (Mathworks, Natick, MA). High-speed visuaizations of the primary break-up process were captured by back-iuminating the region of interest at the outet of the iquid and gas jets. A Photron Fastcam 10k digita camera, at a resoution of 56x40 pixes, was focused through a Nikor 65mm Micro ens on a 5 mm x 5 mm region ocated at the outet of the iquid nozze. The camera operated at 1000 frames per second and the iumination came from a Kodak stroboscopic ight synchronized with the camera. Athough the exposure time of the camera was set at 1/000 sec., the ight puses from the stroboscopic ight were very short (approx. 10 microseconds) so that the dropet motion was frozen and the sharpness in the resuting images was enhanced. Images captured by this method for different experimenta conditions are shown in Fig Characterization of iquid rheoogy Five fuids with rheoogies of increasing compexity were utiized in this study, specificay water, two gycero-water mixtures, an Acetone/water/Poyethyene Gyco(PEG)/Ceuose Acetate (CA) mixture and two Opadry II water-based soutions, Y and 85F184 (Coorcon, West Point, PA). The CA:PEG coating was prepared by adding 9% (w/w) CA and 1% PEG to a soution consisting of 3% water and 87% Acetone. Both Opadry soutions were aqueous; however Opadry A (Y ) utiized 15% soids (w/w) whie Opadry B (85F184) utiized 0% soids in soution. The shear rate dependence of viscosity for the different fuids used in the atomization experiments was measured on a Brookfied DVII+Pro digita cone and pate viscometer. The viscosity of water and two different soutions of gycero in water were tested to vaidate the procedure and check the viscometer caibration. The measured vaues were constant across a vaues of shear rate tested, as expected for Newtonian fuids. The rheoogy of three soutions of interest was aso investigated within the range of shear rates avaiabe. Surface tension was measured with a Coe Parmer EW tensiometer. This system uses the du Nuoy ring method with a Patinum Iridium ring and a caibrated torque baance to measure the surface tension of iquids in air. The density, surface tension and viscosity at different shear rates of these fuids were measured prior to atomization and the resuts are given in Tabe 1 beow. Page 4 of 17

7 The data presented in Tabe 1 ceary shows that the Opadry soutions exhibit a strong non- Newtonian behaviour. The other fuids had amost constant viscosity, with the variation in the different measurements being attributabe to sight interna heating at higher shear rates. The shear-thinning (pseudopastic) behaviour of the Opadry soutions was characterized for ow and intermediate vaues of the shear rate. The use of the highest measured shear rate viscosity in the atomization mode yieds a great improvement over use of the ow viscosity vaues, which woud grossy overestimate the effect of viscosity on atomization. The shear rate at the outet from the nozze is estimated to be higher than the range tested here, thus it woud be beneficia to measure the viscosity of the soutions at higher shear rates. It is aso important to note the arge differences in surface tension, ranging from mn/m for the acetone based soution to 7 mn/m for water. This physica property has a very strong impact on atomization dynamics. If the poymer soutions, which have the highest viscosity, did not have such ow surface tensions the resuting dropet size for these fuids woud be orders of magnitude arger than water. 3. Atomization Mode Varga et a. (Varga et a. 003) have demonstrated that the atomization of a iquid jet by a cofowing, high-speed gas stream occurs via a series of instabiities. Initiay, the primary Kevin Hemhotz instabiity deveops in the annuar shear ayer present at the iquid nozze discharge foowed by a secondary Rayeigh-Tayor instabiity at the interface of the acceerating iquid tongues. The initia stages of this process are represented graphicay in Fig. 3. The wave ength of the primary instabiity, λ 1, depends on the gas boundary ayer thickness,δ g, at the gas discharge pane and is given by the foowing expression (Marmottant 001): ρ λ δ, (1) 1 g ρg where ρ and ρ g are the iquid and gas densities, respectivey. For a convergent nozze, such as the PA C air cap used here, the gas fow at the nozze exit is being acceerated and remains aminar such that the boundary ayer thickness is Cb g δ g =, () Re bg where Re U b / ν and the coefficient of proportionaity C depends on nozze design. For the bg Gas g Gas vaues of gas fow rate investigated here, the Reynods number was approximatey The convective veocity of the iquid tongues resuting from this instabiity is Page 5 of 17

8 U c = ρ U Liquid g Gas ρ + + ρ U ρ g. (3) For the primary instabiity to deveop rapidy it is necessary that the Reynods number of the iquid shear ayer is sufficienty arge ( Uc ULiquid) λ1 Re λ 1 = > 10. (4) ν This condition is necessary even though the instabiity is driven by the gas. For non-newtonian fuids the iquid viscosity ν is the effective shear viscosity, which in this investigation is assumed to be the viscosity measured the highest shear rate. The tongues of the primary instabiity, of thickness b, grow rapidy and are exposed to and acceerated by the high-speed gas stream. These tongues are thus subject to a R-T instabiity simiar to a fattened drop in a high-speed gas stream. For non-newtonian fuids the dispersion reation is given by Joseph et a. (Joseph et a. 00) in the form (when ρ g << ρ ) 1 σ k k α k α = ( ak ) 4 4 ( ) ( ) 0 q k n ρ n ρ n ρ, (5) where k is the magnitude of the wave vector, n the ampification rate, a the acceeration of the iquid tongue, σ the surface tension, α the effective shear viscosity of the iquid inτ ij = α e ij, whereτ ij and e ij are, respectivey, the stress and strain tensors of the iquid, and q is given by: q = k + nρ / α. (6) When viscous effects are negigibe, as in atomization of water, the wave number corresponding to maximum ampification is k σ = aρ 3σ. (7) Page 6 of 17

9 When viscous terms are important, as is the case for the water-gycero mixtures and the tabet coating nρ soutions under investigation here, α is arge and it can be assumed that <<1 such that k α nρ ( q1 k) in Eq. (6). The simpified dispersion reation from Eq. (5) then reads: kα k α k α k σ n= + / ( ka). (8) ρ ρ ρ 4 3 Disturbances wi grow when the second term in Eq. (8) is positive and arger than the first term. It is usefu to rewrite equation Eq. (8) in the form: n = k α ρ 1+ aρ k 3 α σρ kα 1/ 1. (9) From Eq. (9) it is seen that the ampification rate is zero when k = aρ σ, which is the capiary cut-off wave number, and when k = 0. The wave number of maximum ampification is given by the third order equation 4 α ρ k 3 3σ k + a = 0. (10) ρ The exact soution of this equation is too compex to be of practica interest. However, for the high viscosity fuids studied, the Ohnesorge number based on the wave ength is arge and the second term in Eq. (10) is sma compared to the first one, so that the wave number of maximum ampification is: k aρ. (11) 3 max α Page 7 of 17

10 π The R-T waveength is λ RT = and utimatey the dropet diameter is a fraction of k max λ RT (Varga et a. 003). Therefore, assuming viscous and surface tension effects are additive to the eading order according to the dispersion reation, we ook for a correation in the form: λ RT = π 3σ α + C 3 aρ aρ. (1) F F The acceeration a in (1) is simpy a = =, where the force F is the drag force exerted by the gas m ρ V stream on a iquid eement, here the iquid tongue of the primary instabiity, 1 F = C ρ ( U U ) A (13) D g Gas c e where C D is the drag coefficient and A e the projected area. The mass of the iquid to be acceerated is m = ρ b A e with b λ 1. The expression for a is therefore given by: ρg( UGas Uc) a. (14) ρ b Substitution of Eq. (14) in Eq. (1) gives: λ RT 1/ 1/6 1/3 σλ ρ ( ) 1 ' g Ug Uc α 1 C +. (15) ρg( Ug Uc) λσ 1 ρσ Further substituting for λ 1 from Eq. (1) and using Eq. () and taking the drop diameter, say the Sauter Mean Diameter (SMD), proportiona to λ RT gives: 1/ 1/4 1/6 1/1 SMD bg ρ / ρ g 1 D Re bg 1/6 /3 = C 1(1 + m r ) 1 + C We D Oh. (16) D D Re bg We b g ρ / ρ D g In Eq. (16), the mass oading effect in the form (1+ m r ) is obtained from energy arguments previousy outined by Mansour and Chigier (Mansour and Chigier 1995), where Page 8 of 17

11 m r m ρ U A m ρ U A Liquid = = and A and A g are the areas of the iquid and gas nozze exit sections, respectivey. g g Gas g Furthermore, this equation indicates a dependency of the SMD on 5/4 U Gas andσ 1/. The drop diameter increases with b g 1/4 if the coefficient of proportionaity C in Eq. (3) remains constant when b g is changed. As wi be shown beow, this woud ony be the case if the ength of the gas jet potentia cone is much arger than the iquid jet s intact ength which is not typica of atomizers designs. The SMD in Eq. (16) has been made dimensioness by the iquid diameter D and the Weber and Ohnesorge numbers are based on D foowing the usua convention. However, it shoud be emphasized that the drop diameter does not depend on iquid diameter but rather on the gas boundary ayer thickness at the nozze exit. This has been ceary demonstrated by Varga et a. (Varga et a. 003) where the iquid diameter was changed by a factor of 3 and the drop diameter remained practicay identica for the same gas fow conditions. For competeness, the various non-dimensiona parameters in Eq (16) are defined as foows: Weber number: We D ρg ( UGas Uc) D =, σ Ohnesorge number:oh = α ρ σ D, (17) Reynods number: Re bg UGasbg =, ν g Mass fux ratio: m r ρuliquid A =. ρ U A g Gas g The coefficients C 1 and C in Eq. (16) are order 1 and vaues for both coefficients are determined from experiments. The vaue of C 1 depends on the gas nozze geometry in genera, and on the contraction ratio in particuar, because for a given nozze size the gas boundary ayer thickness at the iquid nozze discharge depends strongy on the contraction ratio. C characterizes the viscosity dependence of the critica wavenumber in the R-T instabiity, compared to the surface tension dependence. This vaue is associated to the additivity and inearity of both cohesive effects, surface tension and viscosity, which Page 9 of 17

12 determine the growth rate of the instabiity. The vaidity of the inear theory for R-T instabiity has been confirmed for a wide parameter range via quaitative observation of the jet break-up process. Another important parameter, which does not appear expicity in Eq. (16), is the dynamic pressure ratio M that determines the rate of atomization and hence the intact ength of the iquid stream (Lasheras & Hopfinger, 000). This ratio is defined as M ρ U =. (18) ρ g Gas U Liquid The dimensioness intact ength of the iquid stream can be defined as L D 6 M and in the present investigation M is typicay arge (of the order 100). The gas potentia cone ength is approximatey 6b g. For efficient atomization it is desirabe that the gas potentia cone ength is equa or arger than the iquid intact ength so that the primary atomization is competed before the gas veocity starts to decrease. This requirement is expressed by b g M > 1. (19) D It is worth noting that for the fow rates and atomizer utiized in this investigation, Eq. (19) is satisfied easiy, with vaues typicay exceeding 10, strongy suggesting that atomization in pharmaceutica tabet coating is typicay quite rapid and efficient. Finay, the fuid jets under the conditions of interest here are aminar, but woud potentiay become turbuent if the fow rates are significanty increased. Turbuent conditions of the iquid stream at the nozze discharge pane woud have itte effect on the atomization process, whie turbuent conditions in the high-speed gas stream woud require atering the exponent of Re bg in Eq (16). 4. Rheoogica properties The non-newtonian behaviour in Eq. (5) is expressed by the effective viscosity, α reating the stress tensor with the strain tensor τ = α e. (0) ij ij Page 10 of 17

13 Mansour & Chigier (Mansour and Chigier 1995) considered air-bast atomization of power aw iquids with the shear viscosity of the form: m 1 α s = µγ. In Eq. (1), the subscript s is added to distinguish shear dependent viscosity from eongation strain dependent viscosity. Athough eongationa strain is dominant within the iquid nozze (Mansour and Chigier 1995), during atomization shear is anticipated to be much arger than eongationa strain. When m = 1 in Eq. (1) the shear viscosity is just the iquid viscosity, for m <1 the iquid is shear thinning for m>1 it is shear thickening. An estimate of the shear rate in the atomization process is given by U / λ 1 such that α in (16) may be repaced by Bµ ( U / λ ) m c 1 1 where B is a constant to be determined from experiments. In the simpest rheoogica mode of the fuids used, we can assume the Non-Newtonian behaviour wi manifest in three different stages according to the vaue of the oca shear rate experienced by the fuid. For very ow shear rates, viscosity can be modeed by an inverted paraboa α = µ 0 1 γ γ 0. For an intermediate range of shear rates, the fuid behaviour is captured by the cassica power aw γ α = µ 1 γ 1 m 1. Finay, for the argest vaues of the shear rate viscosity assumes an asymptotic vaue that can be estabished from the end of the vaidity of the power aw. Typicay this asymptotic behaviour determines the effective viscosity for the break-up process µ, as the fuid been atomized is subjected to very arge shear deformations (Mansour and Chigier 1995). Thus, the vaue of viscosity obtained at the argest shear rate is utiized for the mode. This vaue represents a conservative estimate of the shear during atomization, but avoids extrapoation based on the power aw constitutive aw outside the range of shear rates tested. Despite this simpifying assumption the mode data wi be shown to compare favouraby with experimenta data. c (1) 5. Resuts and Discussion 5.1 Quaitative observations Images of iquid jet break-up are shown in Fig. 4, for severa Weber and Reynods numbers for both the water and the 85% gycero-water soutions. The gycero-water mixture is of interest because it aows determination the coefficient C in Eq. (16) and eucidates the effect of viscosity on the atomization process. The vaue of the coefficient C is found by estimating the waveength of the R-T instabiity using Page 11 of 17

14 image processing. For the range of parameters and atomizer geometry studied, C was found to be equa to Effect of Pattern Air The mode for the break-up of the iquid jet into dropets does not take into account the existence of the two side jets that inject the pattern air into the spray. The geometry of the atomizer used in this investigation is such that the pattern air impinges on the main stream at a distance downstream that is arge compared to the break-up ength. As was discussed previousy, the growth of the R-T instabiity whose waveength utimatey determines the dropet size is assumed to take pace in the potentia cone of the main gas jet, a condition which is necessary for efficient atomization. For the conditions under investigation here, this distance is typicay of the order of 3 mm. The impinging distance of the pattern air, based on the atomizer geometry, is ~7 mm. For visua reference, the axia extent of the images shown in Fig 4. is 5mm, so the pattern air jets impinge downstream of the fied of view shown. Therefore, the effect of the pattern air can be negected in the break-up mode which determines the dropet SMD, but the pattern air pays an important roe in the transport of the dropets of different sizes. The pattern air induces an asymmetry in the veocity fied of the spray such that the cross section becomes eiptica rather than the circuar cross section typica of a co-axia atomizer. The patterned spray is narrower aong the axis of the side jets, where the pattern air momentum forces the spray inwards (minor axis of the eipse), and broader in the perpendicuar axis (major axis of the eipse). This is shown ceary in Fig. 5, where the dropet SMD distribution is potted for both the minor and major axis, at two distances downstream of the atomizer. To generate the data presented in Fig. 5 water was atomized using a gas fow rate of 59.5 SLPM using the atomizer detaied in Fig. 1. The effect of the pattern air on the dropet size distribution is rather compicated. The size distribution is concave aong the major axis with arger dropets found at increasing radia distances from the center of the atomizer, due primariy to the tendency of arger dropets to conserve radia momentum onger than smaer dropets. A contrary situation occurs aong the minor axis, where the effect of the side jets is more noticeabe for the arger dropets that are subject to higher aerodynamic focusing towards the center of the spray axis due to their arger inertia and ower diffusivity. The smaest dropets quicky adjust to the oca gas veocity, minimizing any oca perturbation inficted by the pattern air. This expains the convex shape of the diameter distribution found aong the cross section minor axis. Furthermore, some arge dropets acquire a significant radia veocity component as they interact with the pattern air and, because of their arge inertia, do not ose that momentum as they cross the spray s axis. These dropets are found in reativey high proportion at the outskirts of the jet, where the dropet number density is ow and a sma number of very arge dropets have a significant impact on the average diameter of the distribution. This expains the Page 1 of 17

15 change from convex to concave in the diameter radia distribution. Note that the infexion point associated with this change in character moves outwards from the spray s axis with distance downstream, as one woud expect the arge dropets to continue to move radiay, occupying the outer region of the spray. This inhomogeneity in the diameter distribution aong the radia coordinate in the spray was found consistenty for a fuids tested. 5.3 Dropet size measurements and comparison with mode prediction Predictions from the quantitative mode were compared with experimenta resuts for the five fuids discussed previousy. Conditions for the atomization experiments were invariant over the range of fuids tested, and were conducted using a iquid fow rate of 10 g/min and a gas fow rate of 59.5 SLPM. For the atomizer and fow rates utiized in this investigation, this resuts in a gas veocity in the annuar gap at the base of the iquid nozze of ~0 m/s and the veocity of the iquid stream was approximatey 0.4 m/s. The dynamic pressure ratio, M, for water was ~317 which was a typica vaue over the range of fuids under the conditions of interest here. The data from water and the two gycero-water mixtures, shown in Fig. 6, was used to vaidate the mode and to determine if appropriate vaues were assigned to the two adjustabe constants in the mode. As Fig. 6b demonstrates, setting both constants equa to 1 produces acceptabe mode predictions for conditions of efficient atomization and a aminar gas-stream boundary ayer. The vaues of dropet SMD predicted by the mode are 19 µm for water, 4 µm for 59% gycerowater and 43 µm for 85% gycero-water. It is important to emphasise that the mode is predicting the dropet SMD after the atomization process is compete, which generay occurs at an axia distance of ~50mm downstream from the atomizer. Upstream of the competion of atomization arge dropets and igaments are present that affect the PDPA measurement of SMD. In pharmaceutica tabet coating the distance from the atomizer to the tabet bed is generay greater than 100 mm and does vary somewhat depending on the scae and manufacturer of the equipment. The kinematic viscosity of the 59 % and 85 % gycero-water mixtures are approximatey 10 and 65 times that of water. It can be seen in Fig. 6 that the dropet SMD is not affected when the kinematic viscosity is ten times that of water; however the dropet SMD becomes noticeaby arger when the viscosity is of the order of sixty-five times that of water. This behaviour is we represented by the additive terms in Eq. (15). When viscosity is ow, the first term, which depends soey on surface tension, is dominant and the dependence on viscosity is negigibe. When viscosity is high, however, the term mutipying C is much arger and the dependency on surface tension becomes weaker, resuting in a dominant contribution from the viscous term. Additionay, increasing viscosity aso increases the distance at which atomization is compete, due to fact that both the primary and secondary instabiities growth sower when the viscosity is arge. Variation of the dropet SMD with increasing effective Page 13 of 17

16 viscosity was found to be we captured by the additive dependency of the Ohnesorge and the Weber numbers incorporated within the mode. Data for the three tabet coating soutions is presented in Fig 7. The vaues of SMD predicted by the mode were 59 µm for CA-PEG, 57 µm for Opadry A and 41 µm for Opadry B. These vaues compared favoraby with the experimenta vaues measured at downstream from the injector after atomization is compete, see Fig 7b for detais. 6. Concusions This work presents a mode deveoped to predict the performance of co-axia atomizers utiized in pharmaceutica tabet coating when atomizing common tabet coating soutions under typica processing conditions. This mode has been vaidated using fuids of increasing rheoogica compexity. Output from the mode is the SMD of the atomized dropets after competion of atomization. The mode resuts were found to compare favoraby with experimenta data over the range of fuids tested. In addition, deveopment of the mode has yieded usefu insights into the characteristics and performance of the atomizers frequenty encountered in pharmaceutica tabet coating. For exampe, this investigation has ceary demonstrated that, for typica processing conditions, pattern air pays a negigibe roe in the atomization process, which occurs via a series of instabiities which form very cose to the atomizer discharge pane. However, the patter air was shown to pay an important roe in the transport of dropets within the spray. This work has aso identified areas which require further investigation. Severa of the fuids under investigation here are non-newtonian, and the range of shear rates at which viscosity was measured is beow what woud be encountered during atomization. Ceary, characterization of the rheoogica properties of the fuids at conditions which more cosey approximate the atomization process woud be expected to improve the accuracy of this mode. However, the resuts presented here suggest that the current mode is capabe of making timey predictions of atomizer performance in pharmaceutica tabet coaters, and offers a practica too to guide scae-up and optimization in these systems. Acknowedgements Hepfu discussions concerning tabet coating were hed with severa Pfizer coeagues, specificay A. G. Thombre, B. A. Johnson and P. E. Luner. The support of this research by the management of Pfizer Goba Research and Deveopment (S. M. Herbig, C. A. Oksanen and C. M. Sinko) is gratefuy acknowedged. Finay, J. Rodriguez at UCSD offered vauabe assistance with severa aspects of this project. Page 14 of 17

17 REFERENCES am Ende, M. T., and Berchiei, A. (005). "A Thermodynamic Mode for Organic and Aqueous Tabet Fim Coating." Pharmaceutica Deveopment and Technoogy, 10, Bachao, W. D. (1994). "Experimenta Methods in Mutiphase Fows." Internationa Journa of Mutiphase Fow, 0, IMS Midas Database, 005. IMS Heath, CT, USA. Joseph, D. D., Beaver, G. S., and Funada, T. (00). "Rayeigh-Tayor Instabiity of Viscoeastic Drops at High Weber Numbers." Journa of Fuid Mechanics, 453, Kawase, Y., and De, A. (198). "Ligament-Type Disintegration of Non-Newtonian Fuid in Spinning Disk Atomization." Journa of Non-Newtonian Fuid Mechanics, 10, Kremer, D. M., and Hancock, B. C. (006). "Process Simuation in the Pharmaceutica Industry: A Review of Some Basic Physica Modes." Journa of Pharmaceutica Sciences, 95(3), Lasheras, J. C., and Hopfinger, E. J. (000). "Liquid Jet Instabiity and Atomization in a Coaxia Gas Stream." Annua Review of Fuid Mechanics, 3, Lin, J.-C., and Gentry, J. W. (003). "Spray Drying Drop Morphoogy: Experimenta Study." Aeroso Science and Technoogy, 37, Mackapow, M. B., Zarraga, I. E., and Morris, J. F. (006). "Rotary spray congeaing of a suspension: Effect of disk speed and dispersed partice properties." Journa of Microencapsuation, 3, Mansour, A., and Chigier, N. (1995). "Air-bast Atomization of non-newtonian Liquids." Journa of Non-Newtonian Fuid Mechanics, 58, Marmottant, P. (001). "Atomisation d un courant iquide dans un courant gazeux," Ph.D. Thesis, Institut Nationa Poytechnique de Grenobe, Grenobe. Masters, K. (1976). Spray Drying: an introduction to principes, operationa practice and appications, Wiey, New York. Muer, R., and Keinebudde, P. (006). "Comparison Study of aboratory and production spray guns in fim coating: Effect of pattern air and nozze diameter." Pharmaceutica Deveopment and Technoogy, 11, Oakey, D. E. (004). "Spray Dryer Modeing in Theory and Practice." Drying Technoogy, (6), Varga, C. M., Lasheras, J. C., and Hopfinger, E. J. (003). "Initia Breakup of a Sma-Diameter Liquid Jet by a High Speed Gas Stream." Journa of Fuid Mechanics, 497, Page 15 of 17

18 Tabes Tabe 1 Physica and rheoogica properties of the fuids utiized in the atomization experiments. Page 16 of 17

19 ρ (kg/m 3 ) σ (N/m) µ x s -1 µ x s -1 µ x s -1 µ x10-3 5s -1 T (ºC) Water % Gycero - Water % Gycero - Water CA:PEG 10% Soids Opadry A (Y ) 15% Soids Opadry B (85F184) 0% Soids

20 Figures Figure 1 Atomizer schematic. Figure (a) Sketch and (b) photograph of the atomization experiment. Figure 3 Sketch of the primary instabiity in the iquid stream caused by the high-speed, co-axia air stream. Figure 4 High-speed visuaizations of the atomization process for water (eft) and 85% gycero-water (right) at various atomization conditions. (a) We D = 60; (b) We D = 153; (c) We D = 640. Figure 5 Sauter Mean Diameter of water dropets as a function of radia position at two distances downstream of the injector. Figure 6 Sauter Mean Diameter of water and the two gycero-water mixtures downstream of the atomizer. (a) entire fied of view; (b) detai view showing competion of atomization and mode prediction. Figure 7 - Sauter Mean Diameter of three tabet coating soutions downstream of the atomizer. (a) entire fied of view; (b) detai view showing competion of atomization and mode prediction. Page 17 of 17

21 High Momentum Coaxia Air Jet D go =.85 mm D gin = 1.78 mm D = 0.71 mm L nozze = 1.0 mm D PA = 0.71 mm Side Air Jets (Pattern Air) Low Momentum Liquid Jet R Z

22 (a) (b)

23 λ 1 b U C U Gas b g U Liquid R Gas R Liquid R Z

24 (a) (b) (c)

25

26 (a) (b) Fig. 6

27 (a) (b) Fig. 7