Enhancement of Air Filter with TiO2 Photocatalysis for Mycobacterium Tuberculosis Removal

HEPA filter is normally recommended for removal of Mycobacterium tuberculosis from the airstream. Due to its high cost, a cheaper air-filter substitution is proposed in this study. Both lowand high-grade glass fiber air filters were coated with 5% TiO2 using either polyethylene glycol (PEG), Silane (Si-69), or DURAMAX (D-3005) as a binder. The coated filter was placed in a test duct and irradiated with UVA at an intensity of 4.85 ± 0.41 mW/cm to investigate photocatalysis for M. tuberculosis. The effects of dark and light conditions as well as initial exposure to UVA on bacterial removal were studied. Silver-doped and iron-doped TiO2 at different concentrations were also tested using face velocities of 0.1 and 1 m/s and humidity levels of 50 ± 10% and 70 ± 10% RH. The most appropriate binder for coating TiO2 onto filters was 3% D3005. Using 0.1% metal-doping, 100% removal efficiencies were found. Dark/light conditions affected the hydrophobic/hydrophilic properties of TiO2-coated filters and the removal efficiencies. The removal rate could be increased by extending the warm up period of the lamps. When the face velocity was increased, the removal efficiency dropped in both filters. Similarly, high humidity adversely affected the removal efficiency, particularly with the coated high-grade air filter.


INTRODUCTION
Tuberculosis (TB) is an airborne-transmitted disease which has caused fatality in over 40% of infections worldwide since 1990.About 8.7 million new cases were reported in 2011, of which 13% were related to HIV infection.Death from TB in 2011 was around 1.4 million, including 1 million HIV-negative cases (WHO, 2013).Thailand is among the highest 22 countries in the world for TB incidence, with 93,000 new cases every year (WHO, 2010).A prevalence rate of 189 per 100,000 population and a mortality rate of 18 per 100,000 population were reported in 2009 (Bureau of Tuberculosis, 2012).To control the infectious environment, the US Centers for Disease Control and Prevention (CDC) has recommended good ventilation (rate ≥ 12 air changes/hr), together with HEPA filters to remove respirable particles at a rate of 300-800 ft 3 /min, and ultraviolet germicidal irradiation (UVGI) (CDC, 2003).
Many health-care facilities in Thailand, however, do not install either HEPA filters or UVGI.The main reason for not installing HEPA filter is the cost and that for UVGI is the harm to skin and eyes (ACGIH, 2012), which causes staff to avoid using it.This study therefore provides an alternative means for TB control by adding TiO 2 photocatalysis to two inexpensive glass fiber air filters: low grade (dust spot efficiency of 25-30%) and high grade (dust spot efficiency of 90-95% for particle ≥ 0.3 µm) air filters.Photocatalysis has become a widely recognized method of microbial disinfection.When electrons in TiO 2 absorb photon energy from the light, they will jump to the conduction band and leave holes behind.These holes can react with surrounding water to produce hydroxyl radicals (°OH), while electrons in the conduction band react with oxygen to produce superoxide radical anions (°O 2 -).These radicals can attack microorganism cell walls to release K + , RNA, proteins, and other important components, resulting in cell death (Wamer et al., 1997;Maness et al., 1999).
Although many studies have investigated the capability of photocatalysis on microorganism disinfection, information for M. tuberculosis is still scarce.For example, Lin and Li (2003) used commercial TiO 2 air-filters to disinfect E. coli, B. subtilis spores, C. famata var.flareri, and P. citrinum spores in a test chamber.Li et al. (2006) coated TiO 2 mixed with aluminum oxide and silver nitrate onto the outermost layer of a fabric mask to destroy E. coli and S. aureus.Vohra et al. (2006)  In TiO 2 coating procedures, a binder is normally added to enhance the binding strength and uniform dispersal of coating material on the substrate.This study thus tried different binders with different concentrations to optimize photocatalysis.Silver doping and iron doping were also applied to TiO 2 coating procedures for efficiency enhancement.The effect of UVA initiation in terms of the lamps' warming up period was investigated, as well as the environmental conditions of relative humidity and air velocity.

METHODS
In this study, low grade (25-30% dust spot efficiency) and high grade (90-95% dust spot efficiency) glass fiber air filters were coated with TiO 2 , using different binder types and binder concentrations and later doped with metals.The coated filter was inserted into a duct chamber and irradiated with UVA lamps to obtain photocatalysis.Mycobacterium tuberculosis was injected into the chamber and their colonies were collected before and behind the filter to observe the photocatalysis efficiency under different face velocities and humidity levels.More details of the experiment are in the following.

TiO 2 -Coating Procedure
Low grade and high grade glass fiber air filters (Fig. 1) were cut to a size of 20 × 20 cm.Before coating, the high grade filter was pre-treated to remove water-resistance by washing with 0.08 M sodium dodecyl sulfate, drying at room temperature for 24 hr and desiccating for at least 2 hr (Chuaybamroong et al., 2010).
The concentration of TiO 2 (Degussa-P25) was fixed at 5% by weight or 5 g/100 mL sterile water.Polyethylene glycol of molecular weight of 1000 (PEG-1000), DURAMAX D-3005 ceramic dispersant (D-3005), or Silane coupling agent (bis-[3-(triethoxysilyl)propyl]tetra sulfide or Si-69) were used as binders.Initially, the concentrations of PEG-1000 and D-3005 were 3% by weight of TiO 2 (3/100 × 5 g TiO 2 = 0.15 g binder), while that of Si-69 was 2% by weight of TiO 2 .Later, D-3005 concentrations of 0.1%, 0.5% and 1% were studied.Each binder was dissolved in 100 mL sterile water using a magnetic stirrer for 30 min.Then 5-g TiO 2 was slowly added and the mixture was stirred for another 30 min.Molecules of TiO 2 suspension were separated using an ultrasonic probe (Sonics, model Vibracell) for 2 min.A filter was then dipped into the suspension for 10 min, dried, and heated in a hot air oven at 100°C for 1 hr to remove any residual organic compounds from the binder.Each coated filter was weighed before and after coating to determine TiO 2 loading.Elemental compositions of uncoated and coated filters were analyzed using the scanning electron microscope (SEM) (Hitachi, model S-3000N) equipped with the energy dispersive x-ray spectroscopy (EDX) (Horiba, model EMAX X-act).Pressure drop for each filter and binder combination was observed by using a manometer.
For metal-doped TiO 2 , silver nano-particle in a range of 5-40 nm with a concentration of 0.1%, 0.3%, 0.6%, or 1.2% by weight of TiO 2 was mixed with 5-g TiO 2 in 100 mL of distilled water.The suspension was stirred for 1 hr followed by ultrasonic dispersion for 30 min and heated at 100°C until dry.It was then ground into a powder.Each binder was dissolved in 100 mL distilled water at varying concentrations and stirred for 30 min.The ground Ag-doped TiO 2 powder was then slowly poured into a binder suspension and stirred continuously for 1 hr, then ultrasonically dispersed for another 5 min.Each filter was then dipped for 10 min and heated as before.In the case of iron-doping, Fe(NO 3 ) 3 •9H 2 O at concentrations of 0.1%, 0.3%, or 0.6% by weight of TiO 2 was used instead of silver nano-particle.Other conditions were held constant.

M. Tuberculosis Preparation
Mycobacterium tuberculosis H37Ra (ATCC 25177), a non human-pathogenic strain, was used in this study.One loop of dried culture was mixed with 15 mL Middlebrook 7H9 broth and incubated at 37°C for 7 days (Malaithao et al., 2009).The suspension was then streaked onto Lowenstein-Jensen medium and incubated for another 7 days at 37°C and stored as a stock culture.For the experiment, the colonies from the stock were transferred to 0.85% NaCl solution and the concentration adjusted to 10 6 -10 8 cells/mL.Twenty milliliter of culture suspension was filled in a six-jet nebulizer (BGI, Inc., model MRE CN25) for spraying in a test duct.

Test Duct
A test duct (Fig. 2) was adapted from Griffiths et al. (2005).It consisted of an inlet filter (efficiency of 40-45%), the nebulizer, a mixing unit, UVA light source, a TiO 2 -coated filter in a frame (Fig. 1(c)), a blower, an outlet filter (efficiency of 60%), while underneath were UVC lamps for destroying the remaining germs.Face velocities of 0.1 and 1.0 m/s were studied.Relative humidities of 50 ± 10% and 70 ± 10% were compared, adjusted by using an ultrasonic humidifier (Microfog, model T10).Temperature was controlled at 25°C by an air conditioner thermostat.

Photocatalysis Study
M. tuberculosis was injected into the chamber using compressed air at a pressure of 20-25 psi applied to a nebulizer.Droplets passed through a mixing unit which had a serpentine vane to guide the flow to turn many times so that the droplets could be dried out and only dry bacterial cells sent to the filter.At 10 cm before the filter, five fluorescent UVA black-light lamps (Sylvania, model 36WT8) were inserted to irradiate at a UVA intensity of 4.85 ± 0.41 mW/cm 2 onto the filter.The intensity was measured at the filter surface with a radiometer (Cole Palmer, model LW09811-50).The lamps were turned on 20 min after the M. tuberculosis was injected in order to achieve stable concentration, then microbe sampling was started.M. tuberculosis was sampled at a distance of 60-cm before and 20-cm behind the filter, using single stage Andersen impactors (SKC, Inc., model Standard Biostage) equipped with Middlebrook 7H10 agar every 20 min for a total of 180 min.During the sampling period, microbes were continuously injected into the test duct.Bacterial removal efficiencies upon switching from dark (without UVA irradiation) to light (with UVA irradiation) were also investigated.Triplicate runs were carried out using coated filter with 3% D-3005 as a binder.The effect of UVA light initiation was measured at different irradiation exposure times: 0, 30 and 60 min after the lamps were switched on.The sampling time was 30 seconds with a flow rate of 28.3 L/min.After sampling, the agars were incubated at 37°C for 7-14 days.The colony count began on the 8 th day.Removal efficiency was obtained by comparing the colonies before the filter to those behind.
After finishing each run, the chamber was sterilized using UVC and spraying 3% glutaraldehyde thoroughly for several hours to eliminate the remaining germs.The air sampling was re-performed using the impactor before starting a new run.An inside surface swab was also done to confirm sterilization.All statistical analyses (ANOVA and t-test) were performed using SPSS version 16.

Filter Efficiency without Photocatalysis
The study started by testing the effect of UVA photolysis on M. tuberculosis using the uncoated filters.Then the effect of microbial adsorption onto the coating material (5% TiO 2 + 3% D-3005) was tested without UVA irradiation.Relative humidity was 46-48% and face velocity was 0.1 m/s.The results are shown in Fig. 3.
It can be seen that photolysis has no effect on M. tuberculosis since no difference was observed with or without UVA irradiation on uncoated filters.Meanwhile, microbial adsorption onto the coating materials increased removal by around 15-16%.Generally, UVA damages microbial DNA when cellular nucleobases absorb the light, resulting in oxidative damage from photosensitization reactions.The quality and quantity of DNA damage depend on types and amount of photosensitizers (Cadet et al., 2005), the molecular structures of which determine the required light energy (Wainwright and Crossley, 2004).Thus it is likely that M. tuberculosis (H37Ra strain) does not contain the photosensitizer that absorbs light at the wavelength of UVA (365 nm).
In addition, M. tuberculosis have negative charges on their surfaces (Hellio et al., 1988;Falkinham, 2002;Stokes et al., 2004) while D3005 or polyacrylic acid ammonium salt is considered as a cationic dispersant (Ganesh, 2011).Counter-charge attraction could play a role in bacterial removal.Moreover, TiO 2 in the dark is reported as hydrophobic (Watanabe et al., 1999).High lipid content in M. tuberculosis cell wall also makes the microbe hydrophobic (Best, 1990).Hydrophobic interaction thus could be another cause of the microbial adsorption onto the coating materials.
However, it should be noted that pre-treatment with sodium dodecyl sulfate can affect filtration efficiency, thus the removal percentage of high grade filter in Fig. 3 is much lower than its reported dust-spot efficiency.

Effect of Switching from Dark to Light Conditions
Bacterial removal from coated filters switching immediately from dark (no UVA irradiation) to light (UVA = 4.9 mW/cm 2 ) was tested over triplicate runs.Face velocity was 0.11 m/s.Temperature and humidity were 24-26.5°Cand 50-56%, respectively.The results are shown in Fig. 4.
It was found that under dark conditions (t 0-60 min ), the removal efficiencies of coated filters were quite constant, showing the same trend as those in Fig. 3.However, as soon as the lamps were turned on after t 60 min the removal efficiencies dropped to 5-14% within 20 min.After that the removal efficiency started to increase again due to.photocatalysis.This could explain by the work of Watanabe et al. (1999) who studied the wettability change of anatase crystalline sample by irradiating with UV light.They found that the water contact angle decreases with irradiation time  (hydrophilic), and increases with no UV irradiation (hydrophobic).With the irradiation of 1 mW/cm 2 , highly hydrophilic surfaces are produced in three hours and the water contact angle is reduced to almost zero.They explained that Ti 4+ sites are reduced to Ti 3+ by UV light with oxygen deficiency.Dissociated water then adsorbs onto the vacant site allowing photocatalytic oxidation to proceed immediately.In our study, turning on the light could reduce the hydrophobic interaction between M. tuberculosis and the TiO 2 -coated filter.While a completely hydrophilic surface has not been produced yet, the removal efficiencies are hence lower than those without photocatalysis.

Effect of UVA Initiation Period
This study also investigated the effect of the UVA initiation period by warming up the UVA lamps for different intervals of time.The test conditions were the same as those reported above.The results are shown in Fig. 5.
It is clear that the UVA initiation had an effect on bacterial removal at t 0 .The longer the lamps had been warmed up, the higher the initial efficiency found.

Binder Effect on Photocatalysis
When TiO 2 was coated onto the filters using three different binders and irradiated with UVA, the photocatalysis efficiency increased, as shown in Fig. 6.Face velocity was 0.1 m/s and relative humidity was 49-56%.
The order of their removal efficiencies is consistent with their TiO 2 (plus binder) loadings (Table 1), except that of 3% PEG-1000.Although PEG-1000 yielded maximum loading, its re-weight after the three-hour test showed the greatest loss of 3.41 mg/cm 2 for low grade filter and 1.07 mg/cm 2 for high grade filter (their initial coated filter weights were 8.57 and 13.59 mg/cm 2 , respectively).The least losses (0.03-0.09 mg/cm 2 ) were found from D-3005 while 0.06-0.27mg/cm 2 losses were recorded from Si-69.Between two filter types, TiO 2 plus binder loadings on the high grade filter were about 4-6 times higher than those on the low grade filter.This is because of more dense fiber mat and hence more TiO 2 plus binder attaching on the high grade air filter (Fig. 7).However, the light may not penetrate through the fiber depth, thus removal efficiency from the high grade filter was not dramatically higher than the low grade filter as it should be.
In addition to more loading of TiO 2 + D-3005 than loading of TiO 2 + Si-69, D-3005 has positive charge (Ganesh, 2011) while Si-69 has neutral charge (Zhu and van Ooij, 2004).The charge attraction between M. tuberculosis and D-3005 might promote removal at the beginning.However, after 100-120 min, more M. tuberculosis colonies accumulated onto the filter surface, resulting in the decrease of the attractive force.The removal rates between D-3005 and Si-69 therefore converged.
Si-69 is a hybrid of organic and inorganic materials.One end of its molecule is a hydrolysable alkoxy group and the other end is an organofunctional group.When alkoxy is hydrolyzed, it converts to hydrophilic silanols (Si-OH) and absorbs onto a metal surface via hydrogen bonds between Si-OH and hydroxyl groups in the metal (M-OH).After drying, these bonds convert to metallo-silosane bonds (M-O-Si) at the interface, by releasing water.Excessive Si-OH groups condense to a siloxane (Si-O-Si) on the metal surface (Zhu and van Ooij, 2004).The low grade filter in this study consists of 3.3% by wt. of silica and lower quantities of metals (0.18% Al, 1.05% Na, 0.16% Mg, 0.38% Ca, 0.56% Ti, and 27.13% O, analyzed by the EDX).Meanwhile, the high grade filter consists of 11.9% by wt. of silica and a larger group of metals (1.33% Al, 3.23% Na, 0.3% Mg, 2.14% Ca, 1.04% K, 1.46% Zn, 2.48% Ba, and 33.97% O).This might imply that more metallo-silosane bonding (and more removal efficiency) occurs on high grade air filters, in addition to more TiO 2 + Si-69 loading.Consequently, less discrepancy between Si-69 and D-3005 was observed with high grade filter than low grade air filter (Figs. 6(a) and 6(b)).However, this study found that filters became more fragile as percentages of Si-69 increased, thus 2% of Si-69 was used instead of 3%.As the tensile strength increases, the toughness decreases (Pothan et al., 2002) and this may account for the brittleness in our study.
In the case of PEG-1000, the removal efficiency was lowest.PEG is a neutral polymer that easily dissolves in water or polar solvent.Water molecules orient to surround the polymer chain to create a hydration shell with a maximum of one water molecule per ethylene glycol unit (-CH 2 -CH 2 -O-), making PEG's solubility unlimited (Allen et al., 2002).Tirosh et al. (1998) used a differential scanning calorimeter to measure water in each molecule of PEG-2000 and found 136 ± 4 molecules of water binding with it.Thus, it is possible that there is less space for TiO 2 bonding which leads to the greatest loss of coating materials, resulting in lower photocatalysis than with other binders.Besides, Allen et al. (2002) reported that PEG has no attraction to lipid bi-layer in solution.Unfortunately, the cell wall core of M. tuberculosis is composed of a lipid complex called mycolyl arabinogalactan-peptidoglycan while its upper cell wall segment is composed of free lipids (Brennan, 2003).The unusually high lipid content in cell walls results in M. tuberculosis' hydrophobicity (Best, 1990), thus use of PEG is not appropriate.
In terms of testing time, the removal efficiency from photocatalysis increases with time in every set of the experiment.Since the attack of free radicals is mainly on microbial cell walls, it takes time to produce a puncture and release the internal organelles to death (Maness et al., 1999).As mentioned above, highly hydrophilic surfaces of TiO 2 are produced at three hours with irradiation of 1 mW/cm 2 (Watanabe et al., 1999).Thus full photocatalysis on the coated filters needs some period of time to develop.It should be noted that the removal efficiencies in tests with various binders (Fig. 6) at the beginning (t 0 ) were higher than those shown in Figs.8-11.This could be attributed to the initiation conditions of the UVA light source.For the data shown in Fig. 6, the light was left on continuously from the previous tests (≥ 3 hr), while in the other tests the light was turned on just before the microbial sampling started.Previously UVA-exposed TiO 2 surfaces might have been completely changed from hydrophobic to hydrophilic, allowing photocatalysis to begin immediately as evidenced by the tests in Fig. 5.

Binder Concentrations on Photocatalysis
After D-3005 was selected as the most suitable binder, concentrations of 0.1%, 0.5%, 1%, and 3% by weight were studied.The results are shown in Fig. 8.
It can be seen that the differences among these concentrations were found to be insignificant (p = 0.77-0.79),thus 3% D-3005 was used in the following tests.

Metal-Doped TiO 2 Photocatalysis
One way of enhancing TiO 2 photocatalysis is by doping metal into TiO 2 .Metal can trap electrons from the conduction band and transfer them to the adsorbed O 2 at the surface of the TiO 2 , decreasing electron-hole recombination in the titanium atom (Vamathevan et al., 2002).In this study, silver and iron were used as electron traps with concentrations of 0.1%-1.2%by weight of TiO 2 .Photocatalysis was studied at a face velocity of 0.1 m/s and relative humidity of 42-55%.The results can be seen in Fig. 9.
In the case of low grade filters (Fig. 9(a)), 0.1% Ag gave the best removal rate (100% removal at minute of 100), while 1.2% Ag gave the slowest rate (100% removal at minute of 140).However, for high grade filters (Fig. 9(b)), there was no difference between 0.1% and 1.2% Ag (p > 0.05).As can be seen in Table 1, the maximum loading was found with 1.2% Ag, but weight loss after a 3-hr test increased as the silver content increased too.After deducting the loss and TiO 2 + D-3005 content from their loadings, the silver contents on the low grade filter were 0.36, 0.22, 0.17, 0.55 mg/cm 2 , while those for high grade filter were 0.39, 0.16, 0.20 and 0.12 mg/cm 2 for 0.1%, 0.3%, 0.6% and 1.2% Ag, respectively.Thus with 1.2% Ag, excess silver remained only on the low grade filter.Excess coverage of silver on TiO 2 could decrease the amount of light irradiating onto the TiO 2 surface, hence reducing electron-hole pair generation (Vamathevan et al., 2002).Excess deposition could also occupy the active sites of TiO 2 thus reducing photocatalytic activity.Since metal acts as a site for electron accumulation on TiO 2 surfaces, when optimum loading is exceeded negative-charge silver sites start to attract positivecharge holes, and become centers of recombination.Then fewer holes are left to react with adsorbed oxygen at the TiO 2 surface (Sobana et al., 2006;Behnajady et al., 2008).Therefore optimum metal loading is very important for enhancing disinfection capability.In our study, 0.1% was the optimum concentration.With 0.1% Ag, similar silver content on low-grade filter (0.36 mg/cm 2 ) and high-grade filter (0.39 mg/cm 2 ) yielded similar removal efficiencies.
Iron-doped TiO 2 photocatalysis was also performed but no significant differences (p > 0.05) were found among the various concentrations although 0.1% of iron showed the best removal rate (data not shown).However, 0.1% Ag was superior to 0.1% Fe (Fig. 10).In addition to more silver loading than iron loading on the filters as shown in Table 1, silver can destroy lipopolysaccharide molecules in cell membranes and even destroy the DNA of microorganisms (van Grieken et al., 2009).

Effects of Face Velocity and Humidity
At the next stage, the coated filters doped with 0.1% Ag were tested for M. tuberculosis removal using two face velocities (0.1 and 1 m/s) and humidity of 50 ± 10% and 80 ± 10%.The results are shown in Fig. 11.In the case of low grade air filters (Figs. 11(a) and 11(c)), face velocity had more influence on M. tuberculosis removal than humidity.When face velocity was fixed at 0.1 m/s, removal was higher at humidity of 50% than at 80%.However, when velocity increased to 1 m/s, the efficiency was the same at both humidity levels.In the cases of high grade air filters (Figs. 11(b) and 11(d)), the trend was different.At the higher humidity level (80% RH), efficiency was lower at both face velocities.As the active sites of TiO 2 were filled with water, fewer were available for photocatalysis (Goswami et al., 1997).In the case of low grade filters, the fiber was probably more hydrophobic thus less water was taken up.Air velocity, in contrast, had more influence on the low grade air filter since it has looser fibers than those of the high grade filter, so that microorganisms can penetrate it more easily.A similar trend was observed with both silver doping and iron doping.

Pressure Drop
As can be seen in Table 2, pressure drop increased when the low grade air filter was coated, but did not exceed that of the uncoated HEPA filter.Thus, there is no performance decline if the HEPA filter is replaced by the coated low grade air filter.However, with coated high grade air filters, the pressure drop exceeded that of the uncoated HEPA filter.Thus, the costs associated with higher pressure drop need to be balanced against the costs from lower price filters.

CONCLUSIONS
An inexpensive alternative to HEPA filtration and UVGI could be the use of metal coated glass fiber filters with UVA irradiation.M. tuberculosis removal is most efficient when filters are coated with TiO 2 , using 3% D3005 as a binder.Immediately switching from dark (no UVA irradiation) to light (with UVA irradiation) had a complex effect on M. tuberculosis removal due to the hydrophobichydrophilic properties of TiO 2 .Initiation of UVA light source also affected the initial efficiency.With metal-doping with either silver or iron, a concentration of 0.1% was best for 100% removal.However, silver-doped TiO 2 yielded better efficiency than iron-doped TiO 2 due to silver's antibacterial effect.High humidity and high face velocity both had adverse effects on photocatalytic disinfection.

Fig. 4 .
Fig. 4. Removal efficiency of filters when switching from dark to light conditions.

Fig. 11 .
Fig. 11.Effects of relative humidity and face velocity on M. tuberculosis removals.

Table 2 .
Pressure drop (Pa) across the filter (ave.± SD) using different face velocities.