Photocatalytic Treatment ofBioaerosols: Impact of the ReactorDesignS E B A S T I E N J O S S E T , † , ‡

J E R O M E T A R A N T O , † N I C O L A S K E L L E R , †

V A L E R I E K E L L E R , * , † A N DM A R I E - C L A I R E L E T T ‡

Laboratoire des Materiaux, Surfaces et Procedes pour laCatalyse (LMSPC), CNRS, Strasbourg University, 25 rueBecquerel, 67087 Strasbourg, France, and LaboratoireGenetique Moleculaire, Genomique, Microbiologie(GMGM), CNRS, Strasbourg University, 28 rue Goethe,67028 Strasbourg, France

Received October 9, 2009. Revised manuscript receivedFebruary 19, 2010. Accepted February 19, 2010.

Comparing the UV-A photocatalytic treatment of bioaerosolscontaminated with different airborne microorganisms such asL. pneumophila bacteria, T2 bacteriophage viruses and B.atrophaeus bacterial spores, pointed out a decontaminationsensitivity following the bacteria>virus>bacterial spore rankingorder, differing from that obtained for liquid-phase or surfaceUV-A photocatalytic disinfection. First-principles CFD investigationapplied to a model annular photoreactor evidenced thatlarger the microorganism size, higher the hit probability withthephotocatalyticsurfaces.Appliedtoacommercialphotocatalyticpurifier case-study, the CFD calculations showed that theperformances of the studied purifier could strongly benefit fromrational reactor design engineering. The results obtainedhighlighted the required necessity to specifically investigatethe removal of airborne microorganisms in terms of reactordesign, and not to simply transpose the results obtained fromstudies performed toward chemical pollutants, especially fora successful commercial implementation of air decontaminationphotoreactors. This illustrated the importance of the aerody-namics in air decontamination, directly resulting from themicroorganism morphology.

IntroductionThe regulation of volatile organic compounds (VOC) hasrecently created a strong incentive for innovative sustainableenvironmental research. As a result, the indoor air qualitycontrol is receiving a growing interest due to the publicconcern over human health. Targets are not only VOC ormore generally chemical pollutants, usually malodorous,toxic, or contributing to global warming, but include alsoairborne microorganisms such as bacteria, viruses, or spores.

The U.S. Environmental Protection Agency considers theindoor air pollution as one of the top five environmentalrisks to public health, since we spend 70-90% of our timeindoors, where pollutant contents are higher (1, 2). Biologicalpollutants are particularly threatening because of the con-

tinuously increasing resistance of microorganisms againstmedical treatments and their dissemination due to theintensification of human transports, as shown with worldwidedamages (SARS, avian, or porcine flu). If many airbornemicroorganisms (AMO) show no or a low virulence, animpressive variety of AMO are a real hazard to safety, suchas bacteria, viruses, fungi, with a huge societal impact interms of mortality and cost (3).

The removal of airborne chemical or biological pollutantsis therefore a challenging task for which photocatalysis hasattracted attention since decades for acting as an efficientair treatment technology because of the oxidizing power ofUVA-irradiated semiconductors (4). The analogy betweenchemical and biological targets results from the organicnature of the microorganism constituents that photocatalysiscan oxidize through oxidizing photoholes or •OH radicals,similarly to liquid and gas phase organics. The cell wallsbeing a complex assembly of high molecular weight organiccompounds (MW > 10 000), contact with TiO2 causes oxidativedamage to cell membrane, considered as the first barriermaintaining the vital cell functions and the first target forphotocatalysis, and leads to inactivate bacteria, viruses,spores, yeasts (5).

This photocatalysis/biology interface has been pioneeredby the photoelectrochemical sterilization of microbial cellsby platinized semiconductors, which opened the door to theapplication of photocatalysis to the life science and enlargedits potential applications (6). In contrast to photocatalysisapplied to chemicals, photocatalysis applied to biologicaltargets remained mainly focused on the treatment of liquidsand on self-decontaminating surfaces, mainly targetingbacteria (especially Escherichia coli bacteria), viruses, fungi,algae, and protozoa (7, 8). By contrast, despite the interestin terms of public health and a large spectrum of applications,the photocatalytic disinfection of contaminated air remainedscarcely studied, due to the complexity of working withbioaerosols, which combines difficulties inherent to micro-biology and to aerosol sciences. Works on bioaerosolsconcerned E. coli, Microbacterium sp., Bacillus subtilis,Bacillus cereus, Staphylococcus aureus, Aspergillus niger, aCandida famata yeast or the MS2 and λ phage viruses (9-13).

Our previous works were devoted to the UV-A photo-catalytic treatment of flowing air contaminated by E. coliand L. pneumophila (14, 15). This paper reports on the needof photoreactors specifically designed for biological applica-tions, since up to now, the chemical approach was the mainconcern. When targeting biological agents, the main researchsconcerned the increase in the biocidal properties mainly bymetallic promotion of the photocatalyst (5). In contrary toworks reporting on innovative designs proposed for removingchemical pollutants and on the corresponding tools devel-oped for their scaling-up, works for reducing the biologicalcontamination level by optimizing the reactor geometryremained scarce, with few articles focused on the photo-catalytic treatment of bioaerosols with reactor design(9-11, 14, 15).

Novel designs of photoreactors should be engineered forovercoming restrictive efficiency limitations and for meetingthe requirements for achieving a commercial implementation(16). Paradoxically, many commercial photocatalytic devicesclaim their efficiency for inactivating AMOs at high flow rates,although a large part of them have only been designed andtested for chemical applications, only assuming the riskyhypothesis that a similar efficiency toward pathogens wasreached. Results usually obtained raise many questions aboutthe efficiencies of photoreactors for removing AMOs at high

* Corresponding author phone: +33(0)368852736; fax: +33(0)-368852761; e-mail: vkeller@chimie.u-strasbg.fr.

† Laboratoire des Materiaux, Surfaces et Procédés pour la Catalyse.‡ Laboratoire Genetique Moleculaire, Génomique, Microbiologie.

Environ. Sci. Technol. 2010, 44, 2605–2611

10.1021/es902997v 2010 American Chemical Society VOL. 44, NO. 7, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2605

Published on Web 03/10/2010

flow rates, since they were only assessed at low labscale rates.However, more than in the case of VOCs for which anyconcentration reduction is valuable, air treatment for reduc-ing infections due to AMOs only makes sense if the removalrate is high (see the dose-response function to pathogenicexposure fitted using a beta-Poisson law in SupportingInformation (SI) Figure S1). Commercial devices should thusabsolutely incorporate only highly efficient reactors. Theimpact of AMOs on the photoactive surface appeared to becritical, and the development of efficient photocatalyticsystems has to focus on that point. Hence, the design ofphotoreactors for decontaminating bioaerosols substantiallydiffers from that targeting the treatment of chemical pol-lutants and it appeared necessary to specifically developphotoreactors devoted to the reduction of the biologicalpollution under realistic conditions, like was done in Grin-shpun et al. (12).

Recently, first-principles computational fluid dynamics(CFD) has become a promising tool for the design ofphotoreactors devoted to the removal of VOCs, since it easilytakes into account the complex interactions between theflow field, the light activation, and the reactions kinetics(1, 17-21). CFD might be helpful for bioaerosol remediationtoo, because working with bioaerosols remains time- andmaterial-consuming. Although concepts like the adsorptionor the diffusion of chemicals cannot be easily transferred tothe microbial inactivation, this tool could help in designingefficient biocidal reactors by enhancing the impact prob-ability of the microorganisms on the surface.

Experimental SectionPhotocatalytic Pilots for AMO-Contaminated Air Treat-ment. The micropilot used for performing the single-passUV-A photocatalytic treatment of contaminated air has beenpreviously detailed (scheme as SI S2), as well as the procedurefor preparing and aerosolizing L. pneumophila aqueoussuspension for generating a reproducible contaminated airflow, and that for recovering and numerating bacteria fromthe outlet stream (14). The single-pass tests were performedat a 5 m3/h air flow in an 70 mm wide and 300 mm longannular photoreactor with a central 19 mm diameter 8WUV-A actinic lamp (Philips, actinic BL, TL8). In this config-uration, the reactor worked with an annular space of 25.5mm. The TiO2 coating was performed by evaporating to

dryness aqueous slurry of the TiO2 and the coated reactorwas dried at 110 °C for 1 h in air. Details on the reactor andon the coating method could be found in ref 22.

Recirculation tests were performed with a tangential flowreactor at a 140 m3/h air flow in a 0.8 m3 glovebox adaptedfor AMOs and used as reaction chamber, targeting L.pneumophila, T2 naked virus, and B. atrophaeus spore. Thisreactor has been commercialized by the Biowind company(France) under the DPA label, schematized in Figure 1 withthe lighting coming from the top of the fan (14). TiO2 wascoated by dipping using an aqueous suspension of TiO2 andfurther drying at ambient temperature. This photoreactorhas an inner volume of 1.2 L with a photoactive surface of1040 cm2 and a TiO2 coating density of 1 mg/cm2. Details onthe reactor and on the coating method could be found in ref14. In both cases, TiO2 P25 (Degussa - Evonik) was used.

Targeted AMOs. Extensive details on the preparation ofthe bacteria, virus and spore suspensions used for aero-solization are reported as SI S3. L. pneumophila of 1-2 µmsize is a well-known AMO, well adapted for bioaerosol tests.The contaminated bioaerosol was obtained by aerosolizinga 5 mL aliquot of a 1.0 × 107 L. pneumophila (strain GS3.11)bacteria/mL aqueous suspension into the reaction chamberas droplets in a high flow rate air stream using a peristalticpump (6.3 × 107 microorganisms per m3 of air).

50 nm mean size T2 naked bacteriophage viruses are usedas models for airborne viruses. The contaminated bioaerosolwas obtained by aerosolizing a 5 mL aliquot of a 1.51 × 108

T2 bacteriophages/mL suspension into the reaction chamberas droplets in a high flow rate air stream using a peristalticpump (9.4 × 108 microorganisms per m3 of air).

Bacterial endospores (≈1 µm) are considered to be themost resistant living form and are produced by Gram-positivebacteria to survive extreme situations. Bacillus atrophaeusspores, former labeled as B. subtilis, were used, since theyare used in dry-heat disinfection norms (EN866-2), and werealready positively used as a surrogate of environmentallyresistant pathogenic microorganisms like anthrax (B. an-thracis spores). The contaminated bioaerosol was obtainedby aerosolizing a 5 mL aliquot (1.0 × 107 spores/mL) of acommercial suspension (CIP77.18, Pasteur Institute, Paris)into the reaction chamber as droplets in a high flow rate airstream using a peristaltic pump (6.4 × 107 microorganismsper m3 of air).

FIGURE 1. Side-view scheme of the self-driven tangential fan decontamination photoreactor with 2D simulation of the velocitymagnitude in the cross section of the reactor: (1) contaminated air inlet, (2) purified and decontaminated air outlet, (3) fan blades, (4)air duct housing.

2606 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 7, 2010

Microorganism Numeration. Direct quantitative analy-sis of bacteria was achieved by epifluorescence microscopyusing the “LIVE/DEAD Baclight Bacterial Viability kit”(Invitrogen) as appropriate viability indicator and stainingmethod based on a membrane integrity test for easilydistinguishing between live (with integrate membranes)and dead cells (with damaged ones), whatever theircultivability (5). Thus, the treatment efficiency can bedirectly derived from the viabilities (expressed in percents)at the inlet and the outlet of a photoreactor. This methodis not suitable to viability of bacterial spores and theinfectivity of viruses. However, since such microorganismsdo not enter a “viable but noncultivable state” (VBNC),their counts can be performed using classical heterotrophicplate count. Details can be found in a review dealing withnumeration methods (5) and as SI S4.

Efficiency Indicator. The efficiency of a biocidal treatmentfor disinfecting liquids or surfaces was usually describedwith the “logarithmic reduction” (LR). This indicator com-pared the concentrations before and after the treatment, sothat the treatment efficiency was defined by this ratio, andexpressed on a log10 scale for overcoming the low precisionof biological numerations. However, this calculation couldnot be used here, since the numeration should be performedover the totality of the bacteria.

Thus, the viability of the bacteria bioaerosol could bedefined as the ratio between live bacteria and both live anddead bacteria in an air flow sampling. The bacteria viabilityin the inlet air corresponded to the viability of the startingsuspension (µin), whereas the bacteria viability in the outletair corresponded to the viability of the collected suspension(µout). The absence of any influence of both aerosolizationand collection processes on the bacteria viability has beenchecked. The LR could be easily expressed using thispercentage viability, and thus be replaced by the logarithmicreduction in viability (LRV) which considers that a virtualquantity of microorganisms N0 entered the photoreactor witha µin viability and left it with a µout one (eq 1).

The survival probability P was the probability for a bacteriato come out of the reactor alive, given by P ) 10-LRV.

Results and DiscussionComputational Fluid Dynamics (CFD) Study Applied toAnnular Photoreactors. If annular photocatalytic reactorsare irreplaceable for theoretical studies in the case of chemicalapplications because they enable determining the kineticparameters and validating the mechanisms which can furtherbe incorporated in CFD models, their efficiency for theremoval of AMOs at high flow rates remains questionable.In order to illustrate that, the motion of spherical particlesinside an annular photoreactor has been simulated usingthe “Fluent” CFD software, as a function of the annular spaceand of the particle size ranging from 10 nm to 10 µm, at aconstant flow rate of 5 m3/h (Figure 2). In such a configu-ration, the photocatalytically active contact surface cor-responded to the internal side of the external tube, with theflow passing through the inner space between both coaxialtubes, the inner tube being in our model the external surfaceof the UV-A actinic lamp tube.

For this computation, no reactions have been incorpo-rated in the model, the particles hitting the internal surfaceof the external wall being simply “trapped” on the photo-catalytic surface. This hypothesis corresponded to a 100%inactivation efficiency of the biocidal surface, so that thecomputation described the motion of particles, but not thephotocatalytic aspects of the reactors (i.e., 100% inactivationefficiency toward a microorganism if it impacts on theilluminated surface, at a given and constant UV-A irradiationrate). A Lagrangian approach was used, it focuses on particletracks, rather than on a control volume like the Eulerianmethod. The system was considered as isothermal, the flowgoverned by the ideal gas law and the AMOs were assumednot to significantly influence the flow characteristics due tovery low AMO concentrations. The velocity of the flow at thephotoreactor inlet and the initial velocity of the particleswere normal to the inlet boundary, and the outlet boundarywas set at atmospheric pressure. Hence, this simplified modeldid not incorporate the effects of parameters related to themotion of the air flow, such as the rotation or the shape offan blades when they are present. However, Sahle-Demessieet al. (23) detailed that these parameters have a significantinfluence on the process efficiency and should be includedin the investigation of more realistic reactors. Moreover,thanks to the axisymetry of this geometry, this 3D problem

FIGURE 2. Probability for spherical particles to hit the photocatalytic surface of an annular photoreactor, as a function of theparticle diameter on a normal scale (A), and on a log-scale (B), and of the annular space, at a constant input of 5 m3/h.

LRV ) log( N0 µin

N0 µout) ) log( µin

µout) (1)

VOL. 44, NO. 7, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2607

in the Cartesian frame could be replaced by a 2D one usinga cylindrical frame. Details were reported in Table 1.

The mass and the momentum conservation equationswere approximated using the Reynolds Averaged Navier-Stokes equations, the most widely used model since it canmodel all turbulence scales (24). The Reynolds numbersreported in Table 1 were calculated by considering thehydraulic diameter for an annular duct as being the differencebetween both inner and outer diameters of the annularreactor. Used for the simulations, they indicated a flowglobally ranged in a laminar to transitional regime. However,in order to accurately model the photocatalytic treatment ofa stream, it is necessary to emphasize on the phenomenaoccurring in the boundary layer, near the photocatalyticsurface. This near-wall zone is particularly complex becauseof the no-slip condition and the high shear stress at the wallwhich result in large gradients. Thus, the shear stress transportk-ω turbulence model (SSTKW) was used to model thecontinuous phase, since it is an empirical model well adaptedfor wall-bounded flows. The solution control was based onthe coupled pressure velocity mode, and used explicitrelaxation factors set at 0.75 for both momentum andpressure, second order for the pressure, and also third ordermuscle for the momentum, the turbulent kinetic energy andthe specific dissipation rate. The motion of the particles wassolved by integrating the force balance equation (written inthe x direction), which equates the particle inertia with theforces acting on the particle, taking into account the dragand the Archimedes forces as well as an additional accelera-tion (e.g., Brownian forces). By contrast to particles havingan aerodynamic diameter larger than 1 µm, it should be notedthat working with particles smaller than 1 µm required toincorporate the Cunningham’s correction factor to the dragforces per unit particle mass expression derived from theStockes law. The Cunningham correction factor rapidly andasymptotically decreases down to 1 with increasing theparticle diameter under normal pressure. Details relative tothat section are reported in SI S5 and in ref (24). Microor-ganisms have various morphologies and sizes, ranging fromsome 10s of nanometers for small viruses, to a few mi-crometers for bacteria or even larger for droplets containinga high number of germs, so that, both laws have to be em-ployed depending on the studied microorganisms.

Hence, Table 1 reported the computation results for theflow field characteristics, whereas Figure 2 evidenced theprobability pattern of spherical particles with diametersranging from 10 nm to 10 µm to hit the photocatalytic surface,in photoreactors with annular spaces ranging from 6 to 46mm, at a constant flow input (5 m3/h). Thus, this describesthe behavior of most of the microorganisms, classified indifferent aerodynamic groups, from viruses usually with sizesin the 10-100 nm range to bacteria and bacterial spores inthe 0.5-3 µm range. Dust or droplets composed of more

than one AMO should be considered as larger particles. Thedecreases of the residuals for the continuity and the velocityfield components to values lower than 10-3 are usuallyconsidered as a good convergence indicator for the simula-tions (25). One should add that the reactor could beconsidered in a first approximation as being close to a plugflow reactor, due to residence times of microorganisms witha narrow distribution.

Figure 2A shows that the probability for microorganismsto hit the photoactive surface of an annular photoreactorremains very low, mostly under 10%, whatever the annularspace. This was especially true for submicronic particles, forexample, viruses or small bacteria, which actually follow themain stream. This hit probability increased when increasingthe passage time in the photoreactor, which is proportionalto the open section, and thus to the annular space (Table 1).This observation was in agreement with the fact that largerthe reactor, slower the particles, so that the probability theyhit the photocatalytic boundary increased. One should alsonote that larger the microorganism, higher the hit probabilitybecause of the higher inertia which allows more importantdirection changes. This behavior resulted from the fact thatsmall size particles (e.g., nanosize viruses compared tobacteria or bacterial spores) will follow the mean air flowand thus will be less impacted on the photocatalytic surface.For a given annular space, the hit probability increased whenincreasing the particle diameter, so that a hit probability ofabout 35% could be obtained for 10 µm diameter large-sizemicroorganisms at low passage times. However, the il-lumination of the photoactive surface quickly decreases withthe increase in the external radius of the annulus, so that,for an equal hit probability, the shortest annular space shouldbe preferred, since the illuminated surface remains moreactive. Another possibility for enhancing the hit probabilitythrough the use of a large photoreactor with very low internalspeed would consist in the use of an external illumination.Unfortunately, even if the efficiency of this configurationwas already reported, it would require the use of many lightsources for surrounding the reactor, leading to very restrictiveextra costs.

Bactericidal Efficiency of a Annular Photoreactor. Thesimulations showed that the hit probability of microorgan-isms increased with the size of the annular space, and reacheda plateau for annular spaces larger than 25-30 mm (mainlyfor particles with diameter lower than 5 µm, that is, usualviruses, bacteria and bacterial spores). Indeed, large annularspaces lowered the flow velocity inside the reactor andincreased the residence time; on the other hand, too largeannular spaces decreased the hit probability on the biocidalsurface since the surface-to-volume ratio decreased. Sincetoo large annular spaces would lead to a decrease in theillumination of the photocatalytic surface and thus wouldnegatively influence the decontamination efficiency, we

TABLE 1. Computational Details of the Axisymetric Meshes Used to Simulate Annular Photoreactorsa

Rint ) Rlamp ) 0.009 m scaled residualsRext-Rint

(m)annular

section (m2)velocity

(m/s) Renumber ofparticles

maximal cellsizes (m) continuity x velocity y velocity k ω

0.006 4.52 × 10-4 3.07× 100 2420 5.50 × 102 1.00 × 10-4/9,10 × 10-10 9.90 × 10-7 2.50 × 10-10 3.70 × 10-12 4.60 × 10-8 1.90 × 10-8

0.016 1.71 × 10-3 8.13 × 10-1 1708 4.65 × 102 1.80 × 10-4/4.40 × 10-9 8.40 × 10-4 2.20 × 10-7 5.70 × 10-9 8.40 × 10-5 2.50 × 10-4

0.026 3.59 × 10-3 3.86 × 10-1 1320 5.08 × 102 2.00 × 10-4/8,68 × 10-9 5.60 × 10-4 8.00 × 10-8 3.60 × 10-9 6.20 × 10-6 3.60 × 10-4

0.036 6.11 × 10-3 2.27 × 10-1 1075 5.31 × 102 2.00 × 10-4/1,12 × 10-8 3.00 × 10-4 3.40 × 10-8 2.00 × 10-9 6.80 × 10-6 1.40 × 10-4

0.046 9.25 × 10-3 1.50 × 10-1 907 5.46 × 102 4.07 × 10-4/2,14 × 10-8 4.50 × 10-4 4.00 × 10-8 1.90 × 10-9 1.40 × 10-6 2.90 × 10-4

a The scaled residuals for the calculation of the flow characteristics indicate a very good convergence at the end of theiterations. (Constant input: 5 m3.h-1). A fine mesh was used for discretizing the boundary layers and the flow, without toohigh computational costs. For each annular space, a first mesh was produced with trigonal cells in the size range of 1-4µm. The mesh, viz. the maximal cell size, was refined in the high velocity gradient zones after that a first solution leadingto a convergence limit was first reached, so that the value of the maximal velocity gradient was finally divided by a factor10. Both initial and final maximal cell sizes (initial/final) are shown here.

2608 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 7, 2010

carried out the photocatalytic treatment of air contaminatedby L. pneumophila inside the photoreactor previouslydescribed, with an annular space of 25.5 mm.

The decontamination efficiency toward L. pneumophilaobtained for two experiments series, with and without UV-A, was expressed in terms of LRV and single pass survivalprobability, obtained through the TiO2-coated annular pho-toreactor working in the seep-flow mode (inlet air flow of 5m3/h). Two grams of TiO2 was evenly coated onto the innersurface of the tube, corresponding to a TiO2 surface densityof 3.5 mg/cm2). L. pneumophila was not sensitive to theaerosolization process, and therefore could be consideredas a valuable microorganism for investigating the decon-tamination efficiency of the photocatalytic treatment.

The TiO2-coated reactor did not show any bactericidalproperties in the dark, with LRVs of -0.2 and 0.0, corre-sponding to single-pass survival probabilities of 118% and100%, respectively (988 and 1584 bacteria observed, respec-tively). Obviously, obtaining a slightly negative LRV valueand a survival probability slightly greater that 100% was notsurprisingly in the microbiology field and has obviously noreal meaning. This resulted from the use of biological agents(and AMO especially) and from the accuracy of the epi-fluorescence numeration method. Such levels correspondedto the absence of any decontamination activity. Thus thedecontamination efficiency obtained under UV-A, if any,could totally be attributed to the UV-A photocatalysis onTiO2. Actually, even under UV-A, there was no statisticallysignificant abatement, with a LRV of 0.1 corresponding to asingle-pass survival probability of 79% (1931 bacteria ob-served). This LRV and this survival probability remainedwithin the measurement error, greatly higher when workingwith AMOs than with VOCs, and therefore this reactorconfiguration could be considered as exhibiting no airdecontamination, or air decontamination efficiency close tozero.

This was in agreement with the simulation results andconfirmed that the major part of the AMOs passes throughthe reactor without impacting on the biocidal surface.

A Case-Study: A Commercial Tangential Flow Photo-reactor. Hence, when targeting commercial implementation,the mass transfer of the AMO and the contact probability ofthe AMO with the photoactive surfaces have to be optimized,without generating too high pressure drops, even at veryhigh flow rates. To afford this challenge, a so-called “tan-gential fan photocatalytic reactor” (DPA) has been engineered(Figure 1) and this configuration previously led to a LRV of1.2 and to a 6% survival probability for the L. pneumophilabioaerosol decontamination in a “single-pass” mode at 5m3/h (14).

However, in order to simulate the decontamination ofsmall rooms at high flow rates and to maintain the AMOs inaerosol, the recirculation rate has to be drastically increased(up to 140 m3/h in the present study), leading of course toa strong decrease in the single-pass efficiency. Therefore,the decontamination tests have to be performed in arecirculation mode inside a test chamber. At high air flowrates, this photoreactor uses the direction change and thevortex produced between the blades for generating strongshear forces and turbulences that enhance the separation ofthe AMO from the air flow (Figure 1) (14). The decontami-nation efficiency toward bioaerosols as a function of timefor the DPA device was evidenced in a recirculation con-figuration at a 140 m3/h flow rate for L. pneumophila, T2virus, and B. atrophaeus endospores (Figure 3). In such aconfiguration, the decontamination tests were probably alsolimited by mass transfer phenomena inside the reactor. Threephotocatalytic tests and three blank tests were performedfor the T2 virus aerosols, whereas four photocatalytic testsand four blank tests, and three photocatalytic tests and three

blank tests were performed over B. atrophaeus and L.pneumophila, respectively.

Nonlinear fits of classical apparent first order decon-tamination kinetics were obtained for the AMOs, so thatapparent time constants were used for characterizing thedecontamination efficiencies. A natural decay was observedduring blank tests whatever the microorganisms, resulting(i) from the high flow rate used during the experiments at

FIGURE 3. Results of the decontamination tests with (A) T2bacteriophage viruses (three tests), (B) B. atrophaeus spores(four tests) and (C) L. pneumophila bacteria (three tests) withUV-A (empty squares) and without UV-A (filled squares)photocatalytic activation, at 140 m3 ·h-1 (spores: 37-51% RH at22-24 °C; bacteriophages: 31-39% RH at 21-25 °C; bacteria:34-55% RH at 20-25 °C).

VOL. 44, NO. 7, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2609

140 m3/h and leading to impact some microorganisms ontothe chamber walls, and (ii) from the possible instability ofthe bioaerosols during the test.

Therefore, the natural decay observed during blank testswas considered by defining the τoff apparent time constant,whereas the τon apparent time constant was derived fromthe photocatalytic tests, also including the natural decay ofthe bioaerosols. The results of the photocatalytic tests (%Con)have thus to be corrected with those of the blank tests (%Coff)for isolating the photocatalytic abatement from the abate-ment due to the experimental settings. Thus, assuming forall cases first order kinetics, and starting with a 100% initialvalue, the %Ccorr photocatalytic abatement and the corre-sponding corrected time constant, that is, τcorr, were derivedfrom eqs 8-9.

The apparent time constants obtained for the AMOs andcharacterizing the decontamination efficiency toward theAMO bioaerosols are reported in Table 2 with the 95%confidence intervals of the parameters.

We showed that working with different kinds of AMOssuch as viruses, bacteria and spores is of high interest whenstudying air decontamination. The nature of the microor-ganisms is very important for assessing their sensitivitiestoward UV-A photocatalysis, and Huang et al. reviewed thatthe sensitivity of microorganisms to TiO2 photocatalysis forwater treatment was likely in the viruses > bacterial cell >bacterial spores order (25). They suggested that the differentmicroorganisms respond differently to photocatalysis dueto structural differences, particularly in terms of complexityand of thickness of the cell envelop. For instance, the Gramnegative bacteria are more affected than Gram positive one,as a result of structural differences in the outer membrane(5). In a first approximation, the sensitivities of microorgan-isms toward TiO2 photocatalysis were usually in agreementwith those obtained for classical disinfection chemicals,viruses being the most sensitive microorganisms, followedby Gram (-) bacteria, then by Gram (+) bacteria, and bybacterial endospores, which are known to be one of the mostresistant living forms (5). However, the observed sensitivityorder was almost only related and assigned to differences interms of chemical compositions of the microorganisms, butneither the influence of the microorganism morphology northat of their aerodynamic characteristics were pointed out,because the disinfection results were usually obtained inliquid phase using suspensions of photocatalysts mixed withmicroorganisms.

In the present work, decontamination apparent timeconstants of 24 min, 17 min and 57 min were obtained forT2 viruses, L. pneumophila and B. atrophaeus respectively,evidencing that the sensitivity would here differ from theexpected order, and thus would follow a bacteria > viruses

> bacterial spores order, significantly differing from thatobtained for liquid-phase or surface UV-A photocatalyticdisinfection. Hence, the results obtained with AMOs il-lustrated the importance of the aerodynamics in air decon-tamination at the biological level, especially at high flow rates,when compared to liquid-phase decontamination or self-decontaminating surface applications, for which the aero-dynamic aspects resulting from the microorganism mor-phology have logically not been taken into account. As aresult, the efficiency shown by photocatalytic reactors towardgaseous and liquid pollutants cannot be directly transposedto the photocatalytic removal or airborne microbiologicaltargets.

CFD Investigation Applied to the DPA PhotocatalyticPurifier Case-Study. In order to understand the flow fieldinside the tangential reactor, a 2D CFD model has beendeveloped starting with a trigonal mesh with cells in themillimeter range. Figure 1 shows the velocity magnitudecomputed inside the 2D cross section of the tangential fanreactor working at a flow rate of 140 m3/h, corresponding toan input velocity of 5 m/s and to a rotor radial velocity of3800 rpm (398 rad/s). Again, the k-ω model was used for theviscosity, and the mesh was refined in the high velocitygradient zones after that a first solution was first reached, sothat the value of the maximal velocity gradient was finallydivided by a factor 10. The absolute convergence criterionson the residuals for the continuity and for the componentsof the velocity field, reached values lower than 10-8, indicatingthat a stable solution was reached.

The CFD calculation showed that a nonnegligible part ofthe air flow was directly sucked to the reactor output (zoneindicated by a red circle), so that the flowsand the AMOtargets especiallyscould not impact on the photocatalyst,which was coated on the rotating blades of the fan and onthe internal surface of the duct housing (dashed line).

Tiny AMOs such as viruses are more likely to follow theair pathways than larger particles such as vegetative bacteriaor bacterial spores, which have more stochastic moves dueto their inertia, leading to a higher probability to hit one ofthe biocidal surfaces. This could explain the unexpectedresistance of the 40 nm diameter T2 bacteriophage virus tothe photocatalytic treatment when compared to the mi-crometer-sized L. pneumophila. This pointed out the prom-ising and beneficial role that CFD computation could playfor improving the efficiency of photoreactors, which stronglydepends on the kind of targets, especially when material-and time-consuming AMO inactivation experiments shouldbe necessarily conducted. So, decontamination tests per-formed in recirculation showed that the DPA photocatalyticdevice was well active for removing AMOs from air, but CFDcalculations evidenced that further improvements could beobtained by reducing the contact-free bypass inside thereactor, especially when small size AMOs like naked virusesshould be inactivated. Specific designs have thus to berationally engineered for taking into account the particularityof AMOs compared to chemical targets, for looking for acommercial implementation of photoreactors with high airdecontamination efficiency.

TABLE 2. Decontamination Time Constants for the T2 Bacteriophage Viruses, Bacillus atrophaeus Spores and Legionellapneumophila Bacteria, with (On), Without (Off) and after Correction (corr.)

time constant T2 bacteriophage viruses B. atrophaeus spores L. pneumophila bacteria

τon 22 min (95%CI: 21-23 min)a 48 min (95%CI: 44-52 min) 13 min (95%CI: 9-17 min)τoff 295 min (95%CI: 241-374 min) 302 min (95%CI: 270-337 min) 59 min (95%CI: 50-73 min)τcorr 24 min (95%CI: 22-25 min) 57 min (95%CI: 51-64 min) 17 min (95%CI: 10-26 min)

a CI: Confidence Interval.

%Ccorr(t) ) %Con(t)

%Coff(t)) 100%.e-t/τon

100%.e-t/τoff) 100%.(e-t/τon+t/τoff

)

(8)

τ corr. ) τ onτ off

τ off - τ on(9)

2610 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 7, 2010

AcknowledgmentsThe Alsace Regional Council, France, is deeply thanked forfinancial support through the projects N° 653/04 and 1096/07 and the Ph.D. Grant of Dr. Sébastien Josset. The BiowindCompany, France, is acknowledged for supplying the DPAphotocatalytic air purifier.

Supporting Information AvailableS1, Dose-response function to pathogenic exposure. S2,Single-pass experimental setup scheme. S3, Preparation ofthe microorganism suspensions. S4, Microorganism nu-meration. S5, Reynolds numbers, force balance equation,Stockes law, and Cunningham’s correction factor. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited(1) Romero-Vargas Castrillon, S.; de Lasa, H. I. Performance

evaluation of photocatalytic reactors for air purification usingcomputational fluid dynamics (CFD). Ind. Eng. Chem. Res. 2007,46, 5867–5880.

(2) Lomborj, B. The Skeptical Environmentalist; Cambridge Uni-versity Press: Cambridge, U.K., 2001.

(3) WHO. Fact Sheet No. 104; World Health Organization: Geneva,2007.

(4) Pichat, P. Chemical Degradation Methods for Wastes andPollutants; Marcel Dekker Inc.: New York, 2003; pp 77-119.

(5) Josset, S.; Keller, N.; Lett, M.-C.; Ledoux, M. J.; Keller, V.Numeration methods for targeting photoactive materials in theUV-A photocatalytic removal of microorganisms. Chem. Soc.Rev. 2008, 37, 744–755.

(6) Matsunaga, T.; Tomoda, R.; Nakajima, T.; Wake, H. Photoelec-trochemical sterilization of microbial cells by semiconductorpowders. FEMS Microbiol. Lett. 1985, 29, 211–214.

(7) Blake, D. M.; Maness, P. C.; Huang, Z.; Wolfrum, E. J.; Huang,J.; Jacoby, W. A. Application of the photocatalytic chemistry oftitanium dioxide to disinfection and the killing of cancer cells.Sep. Purif. Methods 1999, 28, 1–50.

(8) Rincon, A.-G.; Pulgarin, C. Bactericidal action of illuminatedTiO2 on pure Escherichia coli and natural bacterial consortia:post-irradiation events in the dark and assessment of theeffective disinfection time. Appl. Catal., B 2004, 49, 99–112.

(9) Goswami, D. Y.; Trivedi, D. M.; Block, S. S. Photocatalyticdisinfection of indoor air. J. Sol. Energy Eng. 1997, 119, 92–96.

(10) Pal, A.; Pehkonen, S. O.; Yu, L. E.; Ray, M. B. Photocatalyticinactivation of airborne bacteria in a continuous-flow reactor.Ind. Eng. Chem. Res. 2008, 47, 7580–7585.

(11) Yu, K.-P.; Lee, G.W.-M.; Lin, S.-Y.; Huang, C. P. Removal ofbioaerosols by the combination of a photocatalytic filter andnegative air ions. J. Aerosol Sci. 2008, 39, 377–392.

(12) Grinshpun, S. A.; Adhikari, A.; Honda, T.; Kim, K. Y.; Toivola,M.; Rao, K. S. R.; Reponen, T. Control of aerosol contaminantsin indoor air: Combining the particle concentration reductionwith microbial inactivation. Environ. Sci. Technol. 2007, 41, 606–612.

(13) Vohra, A.; Goswami, D. Y.; Deshpande, D. A.; Block, S. S.Enhanced photocatalytic disinfection of indoor air. Appl. Catal.,B 2006, 64, 57–65.

(14) Josset, S.; Taranto, J.; Keller, N.; Keller, V.; Lett, M.-C.; Bonnet,V. UV-A photocatalytic treatment of high flow rate air con-taminated with Legionella pneumophila. Catal. Today 2007,129, 215–222.

(15) Keller, V.; Keller, N.; Ledoux, M. J.; Lett, M.-C. Biological agentinactivation in a flowing air stream by photocatalysis. Chem.Commun. 2005, 2918–2920.

(16) de Lasa, H.; Serrano, B.; Salaices, M. Novel photocatalyticreactors for water and air treatment. In Photocatalytic ReactionEngineering; Springer Science+Business Media, Inc.: New York,2005.

(17) Taghipour, F.; Mohseni, M. CFD simulation of UV photocatalyticreactors for air treatment. AIChE J. 2005, 51, 3039–3047.

(18) Romero-Vargas Castrillon, S.; Ibrahim, H.; de Lasa, H. Flow fieldinvestigation in a photocatalytic reactor for air treatment (Photo-CREC-air). Chem. Eng. Sci. 2006, 61, 3343–3361.

(19) Mohseni, M.; Taghipour, F. Experimental and CFD analysis ofphotocatalytic gas phase vinyl chloride (VC) oxidation. Chem.Eng. Sci. 2004, 59, 1601–1609.

(20) Denny, F.; Scott, J.; Pareek, V.; Ding Peng, G.; Amal, R. CFDmodelling for a TiO2-coated glass-bead photoreactor irradiatedby optical fibres: Photocatalytic degradation of oxalic acid. Chem.Eng. Sci. 2009, 64, 1695–1706.

(21) Fariborz, T.; Madjid, M. CFD simulation of UV photocatalyticreactors for air treatment. AIChE J. 2005, 51, 3039–3047.

(22) Josset, S.; Hajiesmaili, S.; Begin, D.; Edouard, D.; Pham-Huu,C.; Lett, M.-C.; Keller, N.; Keller, V. UV-A photocatalytic treatmentof Legionella pneumophila bacteria contaminated airflowsthrough three-dimensional solid foam structured photocatalyticreactors. J. Hazard. Mater. 2010, 175, 372–381.

(23) Sahle-Demessie, E.; Bekele, S.; Pillai, U. R. Residence timedistribution of fluids in stirred annular photoreactor. Catal.Today 2003, 88, 61–72.

(24) Fluent 6.3 User’s Guide, 2006.(25) Huang, Z.; Maness, P.-C.; Blake, D. M.; Wolfrum, E. J.; Smolinski,

S. L.; Jacoby, W. A. Bactericidal mode of titanium dioxidephotocatalysis. J. Photochem. Photobiol., A 2000, 130, 163–170.

ES902997V

VOL. 44, NO. 7, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 2611