Computational Fluid Dynamic Predictions and Experimental Results for Particle Deposition in an Airway Model

. An area identi® ed as having a high priority by the National Research ( ) Council NRC 1998 relating to health effects of exposure to urban particulate matter is the investigation of particle deposition patterns in potentially-susceptible subpopulations. A key task for risk assessment is development and re® nement of mathematical models that predict local deposition patterns of inhaled particles in ( ) airways. Recently, computational ¯ uid dynamic modeling CFD has provided the ability to predict local air¯ ows and particle deposition patterns in various struc - tures of the human respiratory tract. Although CFD results generally agree with available data from human studies, there is a need for experimental particle deposition investigations that provide more detailed comparisons with computed local patterns of particle deposition. Idealized 3-generation hollow tracheo - bronchial models based on the Weibel symmetric morphometry for airway lengths ( ) and diameters generations 3± 5 were constructed with physiologically-realistic ( bifurcations. Monodisperse ¯ uorescent polystyrene latex particles 1 and 10 m m ) aerodynamic diameter were deposited in these models at a steady inspiratory ¯ ow ( ) of 7.5 L rrrrr min equivalent to heavy exertion with a tracheal ¯ ow of 60 L rrrrr min . The models were opened and the locations of deposited particles were mapped using ¯ uorescence microscopy. The particle deposition predictions using CFD for 10 m m particles correlated well with those found experimentally. CFD predictions were not available for the 1 m m diameter case, but the experimental results for such particles are presented.

ABSTRACT.An area identi® ed as having a high priority by the National Research ( ) Council NRC 1998 relating to health effects of exposure to urban particulate matter is the investigation of particle deposition patterns in potentially-susceptible subpopulations.A key task for risk assessment is development and re® nement of mathematical models that predict local deposition patterns of inhaled particles in ( ) airways .Recently, computation al ¯uid dynamic modeling CFD h as provided the ability to predict local air¯ows an d particle deposition patterns in various structures of the human respiratory tract.Although CFD results generally agree with available data from human studies, there is a need for experimental particle deposition investigations that provide more detailed comparisons with computed local patterns of particle deposition.Idealized 3-generation hollow tracheobronchial models based on the Weibel symmetric morphometry for airway lengths ( ) an d diameters generations 3± 5 were constructed with ph ysiologically-realistic ( bifurcations.Monodisperse ¯uorescent polystyrene latex particles 1 and 10 m m ) aerodyn amic diameter were deposited in these models at a steady inspiratory ¯ow ( ) of 7.5 L r r r r r min equivalent to heavy exertion with a tracheal ¯ow of 60 L r r r r r min .The models were opened and the locations of deposited particles were mapped using ¯uorescence microscopy.The particle deposition predictions using CFD for 10 m m particles correlated well with those found experimentally.CFD predictions were not available for the 1 m m diameter case, but the experimental results for such particles are presented.

INTRODUCTION
In recent years there has been interest in identifying research needs regarding the issue of human health effects of urban par-( ticulate matter EPA 1998;Phalen and Lee ) 1998;NRC 1998 .In its review, the Committee on Rese arch Priorities for Airborne Particulate Matte r of the National Research Council listed 10 research topics that were classi® ed as having the highest priority.One of these 10 areas was the investigation of deposition patterns in respiratory tracts of potentially susceptible individuals.
A key task identi® ed to accomplish this priority was the development and re® nement of mathematical models for predicting regional and local deposition in subpopulation s such as elderly people and individuals with lung diseases.The advent of computational ¯uid dynamic modeling ( ) CFD of particle deposition in airway structure s has the potential to provide predictions for the amount and rate of particle deposition in airways of virtually any speci-® ed structure.CFD has been used to study air¯ow and inhaled particle deposition in ( laboratory animals Morgan et al. 1991;Kimbell et al. 1993;Cohen Hubal et al. ) 1996 and humans.Although CFD has been used to predict particle deposition in various regions of the human respiratory tract, no experimental bench-top particle deposition data obtained using the actual airway geometry used in CFD predictions are available for comparison.Such a comparison would enable evaluation of overall deposition ef® ciency and local patterns, as well as provide insight into accuracy of modeled deposition mechanisms in CFD.Detailed comparisons of experimental local particle deposition patterns with CFD predicted deposition patterns are needed to evaluate and provide guidance to CFD efforts.
Several investigators have utilized CFD to study particle deposition in various three-dimensional human tracheobronchial ( airway bifurcation models Gradon and Or-Â licki 1990; Balashazy and Hofmann 1993a,b; Â Â Kinsara et al. 1993;Asgharian and Anjilvel 1994;Heistracher andHofmann 1995, ) 1997 .Each investigation has utilized a unique geometry, especially in the transi-( ) tion zone proximal to the bifurcation and bifurcation-connectin g airway structure , so some of the anatomical models used are be tter approximation s of actual lung anatomy than are others.Kinsara et  CFD bifurcation model that has yet been described.were removed from the ® lter by sonication in distilled water and quanti® ed using a ce ll-counting hemocytometer me thod ( ) Bhalla 1997 .Particle removal from the ® lter was veri® ed as complete by scanning the entire ® lter using ¯uorescence microscopy.For 10 m m particles, the particles on the ® lter were less numerous, and they were counted using ¯uorescence microscopy.

Hollow Tracheobronchial Model
After deposition of ¯uorescent monodisperse particles, a line in the plane of the airway center lines was marked on the outside of the hollow models.The models were subsequently sliced into approximately equal top and bottom halves along this line.Fluorescent particle counts in photographi c microscope ® elds were determined using ¯uorescence microscopy.In order to quantify particle deposition, a coordinate system ( was superimposed upon each half top and ) bottom of the hollow tracheobronchial model.Each rectangular ® eld was divided into 4 equal quadrants, and particles in each quadrant were counted by 2 individuals.Particle counts were accepted and photographe d after complete agreement.The large number of particles required to de® ne local deposition patterns precluded exact determination of each particle's location.Additionally, a lack of 1 to 1 correspondence be tween microscopically-counted particles vs. particles counted from pho-( tographs including confocal microscope ) images prevented a photographi c determination of particle counts and locations.Depth of ® eld, surface texture, and light re¯ections from the hollow tracheobronchial models provided artifacts on photographs that were not problematic during counts at the microscope.Due to the mag-ni® cation required and length of time required for analysis for the 1 m m particle run, each half of the hollow model was cut into 3 pieces by bisecting the second generation airway.( ) Heistracher and Hofmann 1997 used the conjugate-gradient method with preconditioning and convergence acceleration for solution of the linear equation systems.The boundary conditions used in the CFD predictions included a parabolic velocity entrance pro® le with constant pressure boundary conditions at the model exits ( 5) ambient pressure of 1.0 = 10 Pa .With Reynolds numbers ranging from 1882 to 733, a k-epsilon turbulence model, with k being the turbulent kinetic energy and epsilon the turbulence dissipation rate, was used.Particle-wall interactions were modeled using a ``prismatic'' approach.When a spherical particle is moving on a straight line, the enveloping surface is a cylinder.In general, the pathways will be curved lines, which can be subdivide d into straight sections.The cylindrical pieces of the pathway ( were approximated by prisms e.g., with 8 ) or more edges , where the number of edges were selected via input parameters.This ``prismatic'' approac h allowed for strict consideration of interception, which is important near signi® cantly curved regions of ( the surface in the vicinity of the carinal ) ridge .For the experimental work, entry ¯ow, exit ¯ow, and boundary conditions were similar, except that the entrance ¯ow pro® le was probably not fully parabolic.From Figure 3 of Heistracher and Hof-( ) mann 1997 , the particle de position locations were reduced to the number of particles that would have been seen in the microscope ® elds used to examine the hollow tracheobronchial models.

RESULTS
The overall deposition ef® ciencies for 1 m m and 10 m m aerodynamic diameter particles in hollow tracheobronchial models were 0.01% and 81% , respectively.As expected, the larger particles exhibited less uniform deposition patterns, which are typical of inertial effects.A total of 4,614 1 m m particles were deposited, counted, and mapped to determine the deposition ( ) ( ) of the top 2a and bottom 2b halves ( ) Figure 2 .The shading scheme is shown in Figure 2c.A total of 2,316 10 m m particle s were deposited, counted, and mapped to ( determine the deposition pattern Figure ) 3 .The CFD simulation had a total of 2000 particles entering the ® rst airway with 1,616 particles depositing, resulting in a deposition ef® ciency of 81% .Because of the need to use connectors to attach the hollow model to aerosol generation and ¯ow control tubing, a small length of entrance and exit airways modeled in CFD was not available for analysis experimentally.A total of 16 particles that deposited in the top of the entrance airway in the CFD simulation were excluded from this analysis since this area was covered with a connector, thus resulting in a revised overall deposition ef® ciency for the CFD prediction of 80% .
In both the 1 m m and 10 m m experimental results, a slight left to right asymmetry in particle deposition was detected.In both runs, slightly more particles were on the ( left side than the right side the difference ) was about 2.5% of the total .Part of this difference may be a counting artifact due to the way the grid system was superimposed on the hollow tracheobronchial models.The difference in number of particles on the left versus right sides, which was judged to be accurate enough to compare with the CFD prediction, indicates that air-¯ow distribution through the model was acceptably symmetric.

DISCUSSION
Although particles smaller than 10 m m in aerodynami c diameter are most relevant to environmental exposures, this test of CFD predictions used 10 m m particles because of the availability of predictions in an anatomical model that was relatively realistic and constructabl e using the steroliography process.Furthermore, 10 m m particles have small but ® nite deposition ef® ciencies in tracheobronchial airways of people.Using the particle deposition model of the National Council on Radiation Protection ( ) and Measurement NCRP 1997 , at nearresting ventilation an average nose-breathing adult will deposit about 6% of the total ( number of 10 m m particles inhaled taking ) into account the inhalability, which is 71% .For mouth-breathing, the tracheobronchial deposition of 10 m m particles will be greater, due to lower deposition ef® ciencies for the mouth in comparison to the nose.
The NCRP model predicts -3% tracheo- bronchial deposition during nose breathing for 1 m m aerodynamic diameter particles, even though the nasal deposition is only 17% and the inhalability is 97% .If and when CFD predictions are availabl e for 1 m m particles in the PRB model, our experiments for such particles can be used to test the predictions.
Figures 2 and 3 show that particle deposition patterns for 1 m m and 10 m m particles were consistent with deposition mecha-( nisms impaction, sedimentation, and ) diffusion which are used in CFD predicted deposition.As would be expected for both particle sizes, enhanced deposition at bifurcations occurred, presumably due to impaction.Figure 4 shows a plot of cumula- tive percent deposition from the entrance to the exits of the hollow model for both experimental results and CFD prediction ( ) by Heistracher and Hofmann 1997 for 10 m m particles deposited at a ¯ow rate of 7.5 L rmin.This graph demonstrates the close agreement between experimental results and CFD predictions.Similar heavy deposition occurred at both proximal and distal bifurcations.The differences seen in this graph are more apparent when the local deposition pattern is viewed in the shading scheme used for the experimental data.Figure 5a shows the Heistracher and ( ) Hofmann 1997 CFD prediction for particle deposition pattern for 10 m m particles deposited at a ¯ow rate of 7.5 L rmin.In Figure 5b, this pattern has been converted to the identical shading scheme used for experimental data.The number of particles deposited were normalized to the number experimentally deposited.Similarities and differences in local deposition patterns can be seen from a comparison of Figures 3  and 5b.In addition to predicting heavy deposition at both proximal and distal bifurcations, both CFD predictions and experimental results showed a gradient of particle deposition from the outer airway wall to the inner airway wall in airways between proximal and distal bifurcations.Also, both CFD predicted and experimental results showed similar patterns of he avy deposition on the inside of one of the daughter airways and very little deposition on the corresponding side of the other daughter airway.In general, the experimental deposition pattern is not as concentrated in speci® c areas as CFD predictions.Additionally, CFD predicted a signi® cant amount of deposition in the entrance region that was not observed in our experiment.This could be due to use of a brass connector in experiments to smoothly attach the model to the aerosol system.Additionally, the upstream effect of the carinal ridge to an initially parabolic ¯ow pro® le might cause this deposition in the CFD prediction.
The agreement in overall deposition ef- ® ciency and the similarity between local deposition patterns predicted by CFD and our experimental results indicates that CFD may be a viable tool for predicting particle deposition in hollow tracheobronchial models.The discrepancies in local deposition patterns point to the need to carefully ex-( amine both experimental artifacts effects ) of charge, model surface roughness, etc. and assumptions used in CFD predictions.Additional work will be required on more realistic airway geometries before particle deposition predictions from CFD can be con® dently applied to human airways.However, the agreement between experimental and CFD results in this study is encouraging.

(
The airway anatom y used in this study PRB ) model of Heistracher and Hofmann 1995 is shown in Figure 1.This 3 generation anatomical model was based upon the ( ) symmetric Weibel 1963 adult human tracheobronchial airway descriptions for generations 3] 5.The airway dimensions that were used are shown in Table 1.A com-( plete description of the airway surface in-) cluding the important bifurcation region ( FIGURE 1. Hollow silicone rubber SylgardW, Dow, ) Midland , MI PRB model with brass connectors used in this study.
can be found in the appendix to the Heis-( ) tracher and Hofmann 1995 publication.Solid epoxy airway models were produced from the CFD airway anatom y data ® le using sterolithography by a commercial ( ® rm Scion Technologies, Santa Clarita, ) CA .The photo-curing epoxy selected for TM ( the solid models was Somos 6110 E. I. du Pont de Nemours and Company, New ) Castle, DE because of its potential for resolving ® ne detail.The solid epoxy models, having all airway centers in a plane, ( were potted in silicone rubber Dow, Mid-) land, MI .After curing of the silicone, solid epoxy models were removed from the silicone rubber through slits.The slits were carefully se aled with the same type of silicone rubber.Hollow tracheobronchial airway model outlets were ® tted with brass connector tubes having internal diameters closely matching the airway outside diameters.Particle deposition experiments were conducted at a steady inspiratory air¯ow rate of 7.5 L rmin, which simulated a tra- cheal ¯ow rate of 60 L rmin and repre- sented a state of he avy exertion.Prior to each run, air¯ow was calibrated with a ( 14.5 liter bell-type spirometer Warren E. ) Collins Inc., Braintree, MA .For each particle de position run, hollow tracheobronchial models were oriented horizon-( ) tally 908 to the gravity vector .Particle Generation and DepositionTwo sizes of monodisperse ¯uorescent polystyrene latex particles were used: ( 0.96 m m and 9.7 m m nominal geometrical ) diameters .The 0.96 m m particles were purchased from a commercial supplier ( ) Duke Scienti® c, Palo Alto, CA and had a density of 1.05 g rcm 3 .The 9.7 m m particle s were purchase d from a another supplier (Interfacial Dynamics Corporation, Port-) 3 land, OR and had a density of 1.06 g rcm .The mass median aerodynami c diameter 32:1 January 2000 Aerosol Science and Technology 64 ( ) MMAD was calculated using data pro-( vided by the suppliers sizes were veri® ed in ) our laboratory using light microscopy .Thus, the MMADs of particles used in this ( ) study were 0.98 m m ; 1 m m and 9.97 ( ) ; 10 m m .A Lovelace-type compressed air nebu-( lizer Raabe 1972; In-Tox Products, Albu-) querque, NM was used to generate the ¯uorescent aerosols.The nebulizer was modi® ed for use with larger particles by removing the baf¯e and enlarging the air ( jet in the nebulizer stem Oldham et al.) 1997 .For deposition of 1 m m particles, hollow tracheobronchial models were ® tted into a port on a rodent nose-only exposure ( manifold In-Tox Products, Albuquerque, ) NM .In order to avoid large losses associated with 10 m m particles in the nose-only exposure manifold, a separate aerosol gen-( eration and delivery system a copper en-) closure was used for these experiments ( )Oldham et al. 1997  .Aerosols were generated at a nebulizer pressure of 1.4 = 10 y4 y2 0.1% by volume aqueous suspensions of particles in distilled water.Both particle suspensions were prepared so that -10% of the particles would ( ) be generated as multipletsRaabe 1968 .Aerosols were dried and diluted using a ( ) radial diluter that injected dry air 5% RH into the air stream.The quantity of dry dilution air was 10 times the nebulizer output.The aerosol was discharge d to Boltzmann equilibrium by passing through a 85 Kr ( ) discharge r Liu and Pui 1974 .Aerosols were pulled by a vacuum pump through the hollow tracheobronchial models.Particles that did not deposit in the models were collected on a 25 mm diameter polycarbon-( ate ® lte r Nucle opore Corporation, ) Pleasanton, CA, acquired by Corning and were quanti® ed by ¯uorescence microscopy.Because of the large quantity of 1 m m particles on the ® lter, direct microscopic counting was impossible.Particles

FIGURE 3 .
FIGURE 3. Deposition pattern for 10 m m MMAD monodisperse p articles displayed as particle density per microscopic ® eld for the entire hollow PRB model.Each rectangle represents a microscopic rectangular ® eld of 2.05 mm = 1.4 mm.The sh adin g scheme used is shown in Figure 2c.

FIGURE 4 .
FIGURE 4. Cumulative percent deposition for 10 m m MMAD monodisperse particles plotted from the hollow model entrance to exits for both experimental data v and CFD prediction B .
a 10 m m particle deposition pattern predicted by CFD; adapted from Heistracher and Hofmann ( ) ( ) 1997 .b The same pattern converted to sh aded ® elds as used in the microscopic an alysis of experimental deposition in the hollow model.
This research was supported by the National Heart, Lung, ( ) and Blood Institute R01 HL39682 and the Tobacco-Re-( ) lated Disease Research Program 6LT-066 .Dr. Phalen is a member of the University of California, Irvine, Center for Occupational and Environmental Health.

TABLE 1 . Airway morphometry of hollow tracheo- bronchial models.
Note: Details regarding complete mathematical de®nition of these bifurcation zones can be found in ( )Heistracher and Hofmann 1995, 1997 .