Indoor Air Quality Impacts of Ventilation Ducts: Ozone Removal and ...

291 downloads 164 Views 815KB Size Report
can potentially improve indoor air quality. However, ozone loss on duct surfaces probably does not greatly reduce indoor concentrations. systems. For exposure ...
TECHNICAL PAPER

Morrison, Nazaroff, and Modera ISSNCano-Ruiz, 1047-3289 J. Air Hodgson, & Waste Manage. Assoc. 48:941-952 Copyright 1998 Air & Waste Management Association

Indoor Air Quality Impacts of Ventilation Ducts: Ozone Removal and Emissions of Volatile Organic Compounds Glenn C. Morrison and William W. Nazaroff Department of Civil and Environmental Engineering, University of California, Berkeley, California J. Alejandro Cano-Ruiz Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts Alfred T. Hodgson and Mark P. Modera Ernest Orlando Lawrence Berkeley National Laboratory, Berkeley, California

ABSTRACT The concentrations of contaminants in the supply air of mechanically ventilated buildings may be altered by pollutant emissions from and interactions with duct materials. We measured the emission rate of volatile organic compounds (VOCs) and aldehydes from materials typically found in ventilation ducts. The emission rate of VOCs per exposed surface area of materials was found to be low for some duct liners, but high for duct sealing caulk and a neoprene gasket. For a typical duct, the contribution to VOC concentrations is predicted to be only a few percent of common indoor levels. We exposed selected materials to ~100ppb ozone and measured VOC emissions. Exposure to ozone increased the emission rates of aldehydes from a duct liner, duct sealing caulk, and neoprene gasket. The emission of aldehydes from these materials could increase indoor air concentrations by amounts that are as much as 20% of odor thresholds. We also measured the rate of ozone uptake on duct liners and galvanized sheet metal to predict how much ozone might be removed by a typical duct in ventilation

IMPLICATIONS The quality of air entering a building may be affected by pollutant emissions from ventilation system materials and by pollutant reaction on duct surfaces. This paper shows that primary emissions of VOCs from materials in ventilation systems should not significantly increase the total indoor concentration. However, ozone interactions with these materials can increase the emission rate of irritating compounds, resulting in concentrations that approach the odor threshold. Ozone is also removed at duct surfaces, which can potentially improve indoor air quality. However, ozone loss on duct surfaces probably does not greatly reduce indoor concentrations.

Volume 48 October 1998

systems. For exposure to a constant ozone mol fraction of 37 ppb, a lined duct would initially remove ~9% of the ozone, but over a period of 10 days the ozone removal efficiency would diminish to less than 4%. In an unlined duct, in which only galvanized sheet metal is exposed to the airstream, the removal efficiency would be much lower, ~0.02%. Therefore, ducts in ventilation systems are unlikely to be a major sink for ozone. INTRODUCTION Air provided to mechanically ventilated buildings passes through heating, ventilation, and air conditioning (HVAC) systems that contain many materials. Ducts are typically fabricated of sheet metal and sealed with gaskets or duct sealant. Near vibrating machinery, ducts may be joined by sections made of flexible, polymer-coated fabric. Large amounts of fiberglass duct liners, with polymer resin coatings, are installed inside ducts to deaden sound and to increase thermal efficiency. Ventilation systems also contain particle filters made from materials such as glass fibers. The quality of the air passing through these systems can be altered by four classes of processes: (1) primary emission of compounds, particularly volatile organic compounds (VOCs) from materials; (2) sorption and desorption processes between pollutants and surfaces; (3) pollutant removal by deposition or chemical reaction at surfaces; and (4) reaction between air pollutants and surface materials that lead to the release of chemically transformed compounds. Of particular interest for (3) and (4) are ozone-surface interactions, which tend to reduce the ozone concentration in the supply air, but may generate carbonyls or organic acids that can be released into the air. Some studies have reported a higher incidence of nonspecific health symptoms (“sick-building syndrome”) Journal of the Air & Waste Management Association 941

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera among office workers in buildings with air conditioning, and possibly simple mechanical ventilation, than in buildings with natural ventilation.1,2 Attention in these cases has usually been focused on microbial contamination. However, another possible contributor to this observation is the emission of pollutants, such as VOCs, from HVAC systems. Ventilation systems have been identified as potentially significant sources of VOCs.3,4 For example, Mølhave and Thorsen determined that the materials in the HVAC system of one building were responsible for 80% of all direct indoor emissions of VOCs.4 Interactions of ozone with indoor surfaces has been quantified for rooms,5,6 but has not been studied for ducts. The rate at which ozone is removed at duct surfaces may be important since most of the outdoor air that enters mechanically ventilated buildings passes through ducts. In addition to ozone removal, compelling evidence from laboratory studies demonstrates the potential for ozone reactions at indoor surfaces to generate carbonyls and organic acids that are more irritating than their olefinic precursors. For example, exposure of carpet to ozone in a room-sized chamber reduced the gas-phase concentrations of some hydrocarbons while substantially increasing the concentrations of formaldehyde, acetaldehyde and C5-C10 aldehydes.7 Exposure of latex paint in a test system to ozone was observed to generate formaldehyde.8 Some evidence from field studies suggests that

such reactions might increase concentrations of aldehydes, ketones, and organic acids in houses.9,10 This study addresses the impact on indoor air quality of duct liners and other materials found in the ducts of ventilation systems. We measured the emissions of VOCs and aldehydes, with and without exposure to ozone. In combination with mathematical models based on the principle of material balance, these measurements allow us to estimate the increase in indoor VOC concentrations caused by these ventilation system materials. We also measured the uptake of ozone by duct liners and galvanized sheet metal to predict the ozone removal efficiency for air flow through a typical ventilation duct section. The experiments were performed in a small, stainless steel chamber under conditions of controlled temperature, humidity and air-exchange rate. METHODS Materials The study materials, listed in Table 1, included new and used duct liners (NDL and UDL, respectively), a neoprene gasket, a flexible duct connector, duct sealant, galvanized sheet metal, a flexible spiral-wound duct, and air filters (AFs). Upon collection, the samples were packaged in multiple layers of aluminum foil and stored for periods up to several weeks prior to the experiments. (Used materials were stored for as much as a year in a freezer.) The

Table 1. Emission rates of total volatile organic compounds (TVOC) and aldehydes from duct components. a

Material new duct liners NDL2 NDL3 used duct liners UDL2 UDL3 UDL4 UDL5 used fan-box insulation neoprene gasket duct connector duct sealant spiral-wound duct galvanized sheet metal air filters AF1 AF2 AF3 AF4

TVOC

Emission Rate (µg m-2 hr-1)b HCHO CH3CHO Acetone

C5-C10 Aldehydes

Most Common VOCs

Most Common Aldehydes

C6, C9, C10 C8, C9, C10

b b

b 40

b b

29 b

b b

950 1280 b b

b 37 29 b

b b 25 b

87 38 b b

220 260 b b

unresolvable TXIBc

1140 7200

d b

d b

b b

b b

CCl3F, chlorobenzene branched alkanes & alkenes, alkyl substituted aromatics

670 8800 b

b b b

b 67 b

b b b

b 760 b

b

b

b

b

b

b b 550 430

b b b 38

b b 20 b

b b 57 b

b b b b

unresolvable hydrocarbons

C6

a

Measured at the end of 24-hr test period, in the absence of ozone exposure. b = below quantification limit: 300 µg m-2 hr-1 for TVOC; 20 µg m-2 hr-1 for formaldehyde, acetaldehyde, and acetone; and 150 µg m-2 hr-1 for C5-C10 aldehydes. c TXIB = 2,2,4-trimethyl-1,3-pentanediol, diisobutyrate. d Sample lost or invalid. b

942 Journal of the Air & Waste Management Association

Volume 48 October 1998

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera Table 2. Ozone uptake coefficient parameters. a

chamber was then sealed and ventilated. Gas samples for the analyses of Material A [(cm2/mol)-B] B r2 γ(24 hr) Sample Mass (g) total VOCs (TVOC), individual VOCs, galvanized sheet steelb 2.5 × 10-24 -1.99 0.94 1.1 × 10-6 na formaldehyde, and acetaldehyde NDL1b 7.3 × 10-11 -0.66 0.97 7.9 × 10-6 na were collected from the chamber exNDL2c 5.2 × 10-10 -0.64 0.79 3.2 × 10-5 9.25 d -8 -5 haust stream for elapsed times cenNDL2 1.9 × 10 -0.39 0.85 1.5 × 10 8.70 NDL2e 1.0 × 10-8 -0.45 0.93 1.5 × 10-5 7.72 tered at 3, 6, and 24 hr after the chamNDL3 2.8 × 10-8 -0.36 0.80 1.5 × 10-5 na ber was first sealed. UDL 1.8 × 10-8 -0.46 0.81 4.8 × 10-5 na Samples for TVOC and VOC a 2 analyses were collected at 0.1 L min-1 The parameters A and B quantify the aging effect (see eq 7); r indicates the correlation between log(γ) and log(U); γ (24 hr) is the measured uptake coefficient following 24 hr of exposure to ozone at 100 ppb. for 20 min on multisorbent tubes and b Exposure level of ozone strayed up to 130 ppb because of low overall ozone removal rate. were analyzed by thermal desorption c Parameter values based on initial 24-hr, 100 ppb exposure experiment; data from second experiment shown in Figure 5. d gas chromatography/mass spectromSecond sample of NDL2, exposed to 100 ppb for 24 hr. e Third sample of NDL2, exposed to 200 ppb for 24 hr. etry.11 TVOC was calculated from the total ion-current response.12 An avernew duct liners were either purchased from the supplier age response factor was calculated based on the individual or obtained from a new roll at a sheet metal shop. relative response of characteristic compounds versus an Duct liners are used primarily to reduce noise transinternal standard (bromofluorobenzene). These commission from HVAC fans, but also for thermal insulation. pounds were n-hexane, n-octane, n-nonane, n-undecane, Each duct liner was coated by the manufacturer on one n-dodecane, 1,2,4-trimethylpentane, benzene, toluene, side with a black resin material that is used to reduce fiber ethyl benzene, and m-xylene. The lower limit of quantifierosion into the airstream and also to reduce airflow recation for the TVOC analysis was about 25 µg m-3. Indisistance. According to the manufacturers, NDL2 is coated vidual VOCs were quantified using pure standards. Formwith cured, cross-linked, phenol formaldehyde polymer aldehyde, acetaldehyde, and acetone samples were collected hexamethylene tetramine, and NDL3 is coated with cured for 60 min at 0.5 L min-1 on treated dinitrophenylhydrazine urea extended phenol-melamine-formaldehyde resin. The (DNPH) cartridges and analyzed by high-performance liquid coatings for NDL1 and UDL are unknown but these duct chromatography.13 The lower limit of quantification for these liners appear similar to the others. compounds was approximately 1 µg m-3. The lower limit The study materials for ozone loss measurements, of quantification for the TVOC emission rate from a malisted in Table 2, included galvanized sheet metal (GS), terial with an exposed surface area of 0.01 m2 was ~300 three new duct liners (NDL1, NDL2, and NDL3), and one µg m-2 hr -1. For combined C5-C10 aldehydes, the lower limit used duct liner (UDL) that had been removed from a duct in 1989 and stored in a sealed container until measurement (in 1996). Emissions of VOCs All experiments were performed using the apparatus shown in Figure 1. The reaction chamber is a 10.5-L electropolished stainless steel container with a Tefloncoated, silicone gasket. This chamber is placed inside a temperature-controlled cabinet along with humidifying spargers. The chamber was continuously ventilated at 1.0±0.05 L min-1 with nitrogen that was humidified to 50±5% relative humidity. The cabinet temperature was held at 23 °C. The temperature and humidity inside the chamber were continuously measured using a Vaisala temperature and humidity probe. Specimens of flat materials (typically 0.01 m2) were cut from larger pieces and placed in stainless steel holders. The duct sealant was applied to a metal plate and weighed, and the exposed surface area was estimated from direct measurements of length, width, and height. The specimen was placed on a wire rack in a chamber, and the Volume 48 October 1998

Figure 1. Experimental apparatus. For emission experiments without ozone, humidified nitrogen is supplied to the chamber. For experiments involving ozone exposure, a controlled amount of ozone is generated by UV irradiation of air and blended with a humidified air stream that ventilates the reaction chamber. The ozone concentration is measured, recorded, and controlled by a data acquisition and control system. Journal of the Air & Waste Management Association 943

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera of quantification was ~150 µg m-2 hr-1, largely because of variability in the background concentrations of nonanal and decanal. For total measurable carbonyl compounds, the lower limit of quantification for emissions was ~200 µg m-2 hr-1. The lower quantification limit for formaldehyde, acetaldehyde, and acetone emissions was ~20 µg m-2 hr-1. For selected materials, emissions were measured in the presence of ozone (Table 3). In these experiments, the inlet ozone mol fraction was set to ~120 ppb. The average outlet ozone level ±1 standard deviation (variability) are reported in Table 3. Exposure was initiated immediately after the 24-hr unexposed emissions period, without removing the specimen from the chamber, and was maintained for an additional 24-hr period. The ozone concentration in the chamber exhaust was continuously measured with a Dasibi Model 1003-AH photometric ozone analyzer with a measurement accuracy of 1 ppb or 1%, whichever was greater. Gas samples for analysis of VOCs and aldehydes were collected as described above except that an ozone denuder, constructed of tubing internally coated with potassium iodide, was added to eliminate interferences caused by ozone in the sampling and analysis of formaldehyde and acetaldehyde.14 The emission rate of an analyte was calculated by means of the following equation, derived from material balance: E=

Q (C - Co) A

(1)

where E is the emission rate of the analyte per unit area of material (µg m-2 hr-1), Q is the volumetric flow rate of the gas stream (m3 hr-1), C is the concentration of the analyte in the chamber exhaust (µg m-3), Co is the chamber background concentration (µg m-3), and A is the exposed surface area of the material (m2). To put the emission results in perspective, we conducted simple model calculations of the impact of a duct system on indoor air contaminant concentrations. The model assumes that indoor air is well-mixed, including the air in the ducts. This assumption is supported by the high recirculation rate typical of conventional HVAC system design. Species are assumed to be nonreactive and emitted at a steady rate. The increment in contaminant concentration caused by the duct system, Cd, is then given by a steady-state material balance:

Edd Q

Cd =

(2)

where the emission rate from the duct system, Ed, is assumed to be entirely due to duct liner and duct sealant, found to be the dominant emission sources Ed = Adl Edl + Ads Eds 944 Journal of the Air & Waste Management Association

(3)

Here, Adl is the area of duct liner, Ads is the area of exposed sealant, Edl is the species emission factor for duct liner, and Eds is the species emission factor for duct sealant. Estimates for Adl and Ads are discussed in a later section. Ozone Deposition The ozone uptake rate at duct surfaces was parameterized by the deposition velocity, vd 6 F = vd C

(4)

where F is the deposition flux (mass or mol per area per time) and C is the airborne ozone concentration near the surface. In determining deposition velocity from chamber experiments and in predicting ozone loss in model duct systems, we consistently used the superficial surface area, given by a plane of the same dimensions as the exposed surface. As described by Cano-Ruiz et al.,15 ozone deposition onto indoor surfaces is controlled by two processes operating in series: mass transfer of ozone molecules to the surface followed by uptake at the surface. The mass-transfer rate is described by the “transport-limited deposition velocity,” vt, which represents the deposition velocity in the limit that pollutant-surface reactions are instantaneous. For gases, vt is influenced by the fluid flow field (advection and eddy diffusion), by the molecular diffusivity of the pollutant, and by the shape and texture of the surface. The rate of surface reaction is characterized by an uptake coefficient, γ, which represents the ratio of the rate at which molecules are removed at a surface to the rate at which they would strike a plane surface with the same superficial area. Ozone deposition velocities have been reported for individual materials commonly used indoors, and for entire rooms.6,15 Sabersky et al.16 noted that the deposition velocity decreased with time as a material such as plywood was exposed to ozone. This aging effect was reversible for plywood but not for some other materials. Other researchers have also noted this aging effect.17,18 Some evidence suggests a seasonal difference in roomaveraged ozone deposition velocity, possibly due to surface aging or regeneration.19 Cano-Ruiz et al.20 noted the absence of data on deposition velocity or uptake coefficients for materials that line ventilation ducts. We measured the uptake coefficient for several duct liners and for galvanized sheet metal. We also measured the rate at which the surface of the material ages, thus reducing its ability to scavenge ozone. This information allows us to predict ozone removal for air flow through a ventilation duct as a function of time. The ozone uptake experiments were conducted in the same apparatus as described above with the following Volume 48 October 1998

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera exceptions. The stainless steel chamber lid was replaced with a Teflon lid. The sample material was placed in a Teflon frame so that only the upper surface was exposed and the sample itself was placed on a Teflon shelf. The compressed gas for these experiments was air (instead of N2), which was passed through an activated carbon trap to remove trace organic contaminants. Ozone was generated by exposing a fraction of the airflow to ultraviolet light. A portion of the vented exhaust was sampled by the ozone analyzer. An electromechanical three-way valve was used to control the airstream feeding the ozone analyzer so that either supply air or chamber air ozone concentrations could be measured. Each of the test materials was cut to a square 0.15 m on a side and was placed in a Teflon frame so that only the upper surface was exposed. (For duct liners, some excess fiberglass was removed from the bottom so that it would fit in the Teflon frame.) This assemblage was then placed in the chamber on a Teflon shelf and the chamber was sealed. One 24-hr experiment was performed on galvanized sheet metal, two of the three new duct liners, and the one used duct liner at 100 ppb. For one sample of new duct liner, NDL2, two 100-ppb experiments and one 200-ppb experiment were performed using three different samples. Also, the sample from the first 100-ppb NDL2 experiment was sealed in aluminum foil for a week after the end of the experiment, then subjected to a second 24-hr, 100-ppb experiment. This experiment was performed to measure any regeneration of the duct liner’s ability to remove ozone. Prior to each experiment, the chamber, Teflon frame, and shelf were washed in methanol and dried in an oven at 65 °C. The Teflon parts were then sealed in the reactor. Subsequently, the chamber was ventilated for 4 hr with air containing a high ozone level, >4000 ppb. This procedure quenched the reactor walls so that the baseline removal of ozone in the reactor was less than 2% under standard experimental conditions. A feedback loop controlled the concentration of ozone in the chamber. Before each experiment, the computer created a calibration curve of the inlet ozone level as a function of voltage supplied to the power controller. During each experiment, the analyzer measured the ozone level downstream of the exposure chamber and adjusted the power input to maintain this level at a programmed set point. Thus, the material was exposed to a constant ozone concentration throughout an experiment. The calibration curve served as a means of predicting the inlet ozone concentration so that the rate of surface reaction could be determined. In initial trials we found that the predicted and measured inlet ozone levels differed by less than 3%. In separate experiments, we measured the mass-transport-limited deposition velocity for two material geometries. Volume 48 October 1998

A copper plate was coated with a concentrated solution of potassium iodide and allowed to dry. This plate was placed in the reactor to measure vt for the galvanized sheet metal. A piece of NDL1 was soaked in a concentrated solution of potassium iodide and allowed to dry. It was placed in the reactor to measure vt for duct liners. Potassium iodide is considered to be a perfect sink for ozone because of the fast oxidation of 3 I - to I3-.21 Deposition velocity was determined from the experiments by modeling the chamber as an ideal continuously mixed flow reactor (CMFR). The governing equation for ozone concentration in the chamber, derived from material balance, is (5) where V is the chamber volume (10.5 L), t is time, C is the ozone concentration in the chamber air, Cin is the inlet ozone concentration, Q is the airflow rate through the chamber (1.2 L min-1), As is the superficial area of test material (232 cm2), and vd is the deposition velocity. Since C is measured continuously, the slope dC/dt is known. The parameters Q, V, As, and Cin are also known. Thus, vd can be evaluated as a function of time from eq 5. Given the measured deposition velocity, v d(t), and the mass-transport-limited deposition velocity, v t, the time-dependent uptake coefficient is estimated from the expression 15 (6) where is the Boltzmann velocity for ozone (3.62 × 104 cm sec-1 at 296 °C). The experimental uncertainty in measuring γ is estimated to be ±40%, ±10%, ±30%, respectively, for reaction probabilities 10 -4, 10 -5, and 10-6. The uncertainty is larger for γ > 10 -4 because, for our experimental configuration, ozone removal for γ > ~10-4 occurs at approximately the mass-transport-limited rate. Uncertainty is larger for γ < 10-6 because, at this low level of reactivity, little ozone is lost as air passes through the test chamber. Experimentally, we observed an aging effect in which the uptake coefficient for some materials changed by more than an order of magnitude. To model this phenomenon, we assume that the uptake coefficient is a function solely of the cumulative ozone removed by the surface. Empirically, we found that a power function provided a good fit to data: γ = A UB

(7)

where U is the cumulative ozone uptake (i.e., the integrated flux to surface) in mol cm-2, Journal of the Air & Waste Management Association 945

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera (8) Given the uptake coefficient, γ, the ozone removal efficiency in a ventilation duct can be predicted. As described in the appendix, we used an analogy with heat transfer to estimate ozone removal efficiency for ducts in which γ is constant. A numerical approach was applied to incorporate experimental information on aging. RESULTS AND DISCUSSION VOC Emissions Measured VOC emission rates at elapsed times of 24 hr are reported in Table 1. The neoprene gasket and the duct sealant had the highest TVOC emission rates of 7,200 and 8,800 µg m-2 hr-1, respectively. Two of the used duct liners (UDL2 and UDL3) also had relatively high TVOC emission rates, ~1,000 µg m-2 hr-1. The highest emissions of C5-C10 aldehydes came from UDL2, UDL3, and the duct sealant. The fact that the used duct liners were significantly stronger emitters than new liners suggests the possibility that contamination, in the form of dust deposition and/or VOC sorption, may have an important influence on VOC dynamics in ventilation systems. Many of the materials exhibited low emission rates. For example, the two new duct liners (NDL2 and NDL3), two used duct liners (UDL4 and UDL5), the spiral-wound duct, the galvanized sheet metal, and two air filters (AF1 and AF2) all had TVOC and combined C5-C10 aldehyde emission rates that were below the lower limits of quantification at 24 hr. For all of the materials, the concentrations of the analytes in the chambers generally declined with time during the 24-hr test period. The specific compound most strongly emitted from UDL3 was 2,2,4-trimethyl-1,3-pentanediol diisobutyrate or TXIB, a commonly used plasticizer. The used fan-box insulation emitted mainly chlorinated compounds, likely from the manufacture of the stiff foam. The neoprene gasket emitted many compounds that were tentatively identified as branched alkanes and alkenes, and alkyl substituted aromatics. The aldehyde emissions from the duct sealant were dominated by a single compound, hexanal. The used duct liners produced a number of n-aldehydes above the

quantification threshold. Quantifiable emissions of formaldehyde were observed for NDL3, UDL3, UDL4, and AF4. Quantifiable emissions of acetaldehyde were found for UDL4, the duct sealant, and AF3. Few specific compounds, other than aldehydes, were resolvable in the VOC emission spectra from UDL2, or the duct sealant. Because of their elevated rates of TVOC emission, the neoprene gasket and the duct sealant were selected for investigating the effects of ozone exposure on emissions. Two new duct liners, NDL2 and NDL3, were also selected because the exposed surface area of duct liner in HVAC systems is typically large. The emission rates of TVOC and aldehydes from these four materials at the end of a 24-hr period during which they were exposed to ozone, and the average downstream ozone concentration, are shown in Table 3. The emissions of both TVOC and C5-C10 aldehydes from NDL2 increased with exposure; the increase in the TVOC value is largely explained by the increase in the production of aldehydes. For both the neoprene gasket and the duct sealant, there was a decline in the emission rate of TVOC. This decline may, in part, be attributed to a natural decay in the emission rate over time. The emissions of C5C10 aldehydes from the neoprene gasket increased with exposure to ozone. For the duct sealant, there was a small decrease in the emission rate of C5-C10 aldehydes and a substantial increase in the emission rate of acetaldehyde. The emissions from NDL3 were relatively unaffected by exposure to ozone. The relatively large standard deviation in outlet ozone concentration for these materials reflects the steady increase in outlet concentrations due to material aging. Because of their relatively low emissions or the low exposed surface area in ducts, the study materials are not expected to be dominant contributors to the indoor TVOC concentrations, in the absence of ozone exposure. The increase in TVOC concentration associated with duct material use was estimated for a relatively new building (Soda Hall; volume = 1.4 × 104 m3) on the campus of the University of California at Berkeley. Based on discussions with the building manager and examination of building plans, we assumed that the supply duct was lined with 34 m2 of UDL2 and that the area of exposed duct sealant was 0.7 m2. From eq 3 and the emissions data in Table 1, the

Table 3. Emission rates of total volatile organic compounds (TVOC) and aldehydes from duct components after exposure to ozone for 24 hr.

Material

TVOC

NDL2 NDL3 neoprene gasket duct sealant

550 b 6400 4000

a

Emission Rate (µg m-2 hr-1) a HCHO CH3CHO Acetone b 60 b 24

20 b b 290

166 b 120 b

C5-C10 Aldehydes

Inlet

380 b 330 660

110 140 140 100

Ozone (ppb) Chamber/Outlet 72 ± 12 31 ± 8 65 ± 9 27 ± 15

b = below quantification limit: 300 µg m-2 hr-1 for TVOC; 20 µg m-2 hr-1 for formaldehyde, acetaldehyde, and acetone;and 150 µg m-2 hr-1 for C5-C10 aldehydes.

946 Journal of the Air & Waste Management Association

Volume 48 October 1998

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera TVOC emission rate from duct materials was estimated to be Ed = 38 mg hr-1. Assuming a ventilation rate of 1 × 104 m3 hr-1 (corresponding to an air-exchange rate of 0.7 hr-1) for this building, the estimated increase in indoor TVOC concentration caused by emissions from duct materials would be approximately 4 µg m-3. This increment is small compared with the reported weighted-average geometric mean in established office buildings (60 buildings, 384 measurements) of 180 µg m-3. 22 The study results suggest that exposures of some HVAC system materials to atmospheric ozone may result in increases in the concentrations of C5-C10 aldehydes, a group of odorous chemical irritants. For example, the exposure of NDL2 to ozone produced a combined C5-C10 aldehyde emission rate of ~400 µg m-2 hr-1 (Table 3). For the scenario outlined above, this emission rate from 34 m2 of material would contribute ~1.4 µg m-3 to the combined C5-C10 aldehyde concentration of the building. This contribution represents a nonnegligible fraction of the odor thresholds for some of these compounds (e.g., compare with the odor thresholds of 13 µg m-3 for nonanal and 6 µg m-3 for decanal). 23 Ozone Deposition For duct liners NDL2 and UDL, the initial uptake coefficient was so high as to be indistinguishable from unity. The initial value of the uptake coefficient for NDL1 and NDL3 was ~10-4. The final values of γ for duct liners cluster around 10-5 (Table 2). Interestingly, the final uptake coefficient for the used duct liner, UDL, was about four times higher than for NDL1 and NDL3. However, since we lack information about the history of ozone exposure, particle deposition, or manufacturing details for UDL, we cannot draw conclusions about the source or significance of this difference. The 24-hr value of γ for galvanized sheet metal was ~10-6, about 10 times less than those for the duct liners. All of the materials exhibited aging when continuously exposed to ozone, as illustrated by Figure 2. To maintain 100 ppb in the chamber air, the inlet ozone level begins above 300 ppb, then, as ozone reacts less rapidly, is reduced over the period of the experiment to less than 250 ppb. The mass-transport-limited deposition velocity, vt, was measured to be 0.19 cm sec -1 for galvanized sheet metal (represented by a copper plate) and 0.16 cm sec-1 for the coated duct liner. It is surprising that v t for the flat plate is higher than that for the duct liner, since the duct liner appears to have a higher intrinsic surface area and the increased roughness would be expected to enhance mass transfer. This unexpected result could occur if the fluid dynamics between the two systems were sufficiently different. The only apparent difference in conducting the experiments was that the top surface of the copper plate (as well as the galvanized Volume 48 October 1998

Figure 2. Inlet and outlet ozone levels as functions of time for ozone uptake experiment on a sample of new duct liner, NDL2. During the first 1.5 hr, the ozone generator was calibrated and the sample was not exposed to ozone (data not shown).

sheet metal) was approximately 1.5 cm lower than the top surface of the duct liner. The evolution of the measured deposition velocity, vd, is illustrated in Figure 3 for sample NDL2. Note that the initial deposition velocity value, vd, is 0.23 cm sec-1, which is higher than the measured transport-limited deposition velocity, vt. One possible explanation for this discrepancy is that the fluid dynamics were somewhat different between the two experiments. For example, the surface of the duct liner that was not coated with KI might have been rougher or had more intrinsic surface area. There are some differences in the pattern of fibers and the roughness of the surface of the duct liner, but the roughness “scale” appears to be similar. A second possibility is some ozone loss occurs by homogeneous gas-phase reactions, not accounted for in this calculation. Since the mechanism causing the discrepancy is unknown, we used the measured value of 0.16 cm sec-1 for vt and, in determining the uptake coefficient, discarded measurements for which vd > vt. We also discarded

Figure 3. Deposition velocity (vd) for a sample of new duct liner, NDL2, as a function of time. These results represent the interpretation of data from Figure 2. Journal of the Air & Waste Management Association 947

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera values from the first 20 min of exposure because the rapidly changing ozone concentrations produce large errors in estimating γ. The parameters A and B, used to describe aging according to eq 7, were obtained by linear regression of log(γ) versus log(U) using all other measurements (Table 2). This relationship is illustrated for one sample in Figure 4; similar curves were obtained for the other materials. The result of the repeated 24-hr experiment on the same sample of NDL2 is shown in Figure 5 which shows that following a brief increase in uptake coefficient associated with regeneration, the ozone scavenging rate returns to a profile consistent with predictions from the first 24-hr exposure. The assumption that the uptake coefficient is purely a function of cumulative ozone uptake suggests that the results of three NDL2 experiments should yield identical estimates of the parameters A and B in eq 7. The predicted uptake coefficient based on the 200-ppb NDL2 experiment is, on average, about 35% lower than that predicted by the first 100 ppb experiment. However, the initial deposition velocity varied among experiments: 0.23±0.02 cm sec-1 for the first 100-ppb experiment, 0.15±0.02 cm sec-1 for the second, and 0.17±0.02 cm sec-1 for the 200-ppb experiment. Such a large difference in the initial deposition velocity, and in the estimates of A and B, may be due to intrinsic differences in the three samples themselves. Note that the mass of the first sample exposed to 100 ppb is 20% greater than that of the sample exposed to 200 ppb. We also observed that the apparent bulk density of the fiberglass mat varied. On the assumption that the initial deposition velocity was a better measure of the masstransport-limited deposition velocity, we recalculated A and B for these experiments and found that the curves for the first 100-ppb experiment and the 200-ppb experiment largely overlapped. This result substantiates

the assumption that the cumulative uptake of ozone is a key controlling variable influencing aging. However, the repeated 100-ppb experiment does not match well the first experiment using this same approach. The test results for ozone scavenging were used to predict the ozone removal efficiency for an air-supply duct. Input parameters were based on the same building, Soda Hall, used for the assessment of VOC impact. A rectangular duct with cross-sectional dimensions of 1.2 × 1.5 m was considered. The volumetric airflow rate through the duct was 8.25 m3 sec-1. Calculations were conducted for 30-m lengths of lined duct and for 30-m and 150-m lengths of galvanized sheet metal. For each simulation, a period of 240 hr was considered with the inlet ozone level fixed at 37 ppb, corresponding to the middle of the reported range of annual average values for Los Angeles, 20–54 ppb. 24 The predicted ozone removal efficiency, η, is plotted as a function of time in the presence of aging for all lining materials in Figure 6. The curve for NDL2 is based on values of A and B determined from the first 100-ppb experiment. The ozone removal efficiency in this simulated duct is small. Note that NDL2 was the most “active” duct liner, in that η changed the most over the simulated exposure period. According to predictions, NDL2 initially removes as much ozone as UDL, but quickly loses the ability to remove ozone. After an extended exposure, the used duct liner is the most efficient ozone scavenger, with a long-term removal efficiency of about 3%. Figure 6 also shows that ducts lined only with galvanized sheet metal are unlikely to remove significant amounts of ozone, even where the length-to-hydraulic diameter (L/Dh) ratio is large. Note that these predictions are only valid for the duct portion of an HVAC system, and do not account for

Figure 4. Uptake coefficient, γ, as a function of the cumulative ozone uptake, U, for a sample of new duct liner, NDL2. The results correspond to data in Figure 2. Seven values of γ for U ~ 10-9 mol cm-2 exceed 10-3 and are not plotted. The line represents a linear regression of log(γ) vs. log(U).

Figure 5. Uptake coefficient, γ, as a function of the cumulative ozone uptake, U, for a sample of new duct liner, NDL2. The sample was exposed to 100 ppb ozone for 24 hr, stored for one week without exposure, then subjected to a second exposure for 24 hr at 100 ppb.

948 Journal of the Air & Waste Management Association

Volume 48 October 1998

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera the differential equation dU(x) = - Q dCm(x)

(A1)

where Q is the airflow rate through the duct. In terms of deposition velocity, the uptake rate is given by dU(x) = Cm(x) vd(x) P dx

(A2)

where vd is the ozone deposition velocity, and P is the perimeter of the duct. Eqs A1 and A2 may be combined and cast in dimensionless form using the following definitions: (A3)

Figure 6. Ozone removal efficiency for tested materials as a function of time in a simulated duct. All duct lengths are assumed to be 30 m except for one GS case at 150 m, as noted. NDL = new duct liner; UDL = used duct liner; and GS = galvanized sheet steel. Calculations assume that the duct surface is hydrodynamically smooth.

(A4)

(A5)

ozone loss on fans and filters. There is some indication that soiling may increase the ozone uptake coefficient. However, we cannot do a meaningful analysis of this phenomenon with the results from only one used duct liner. It is unknown how much of an effect soiling of the inner surfaces of the duct, or differences in humidity, might have on the overall ozone removal efficiency and reaction byproduct emissions. CONCLUSION The results of this study indicate that materials used for ventilation system ducts can have a small but discernible influence on the concentrations of ozone and VOCs in indoor air. Among the materials studied, duct liners appear most important as they are used in large quantity in duct systems, exhibit substantial reactivity with ozone, and, in some cases, emit VOCs at substantial rates. Oxidation reactions between ozone and duct materials can produce aldehydes at sufficiently high rates that predicted indoor concentration increments may be a significant fraction of the odor threshold. Additional work would be needed to understand how the presence of dust that accumulates in ducts over the long term influences indoor air pollutant levels. APPENDIX Modeling Ozone Removal in Ducts A mass balance may be applied to determine how the mean ozone concentration Cm(x) varies with distance, x, along a duct and therefore how the ozone removal efficiency is related to the uptake by the duct surfaces U. Under locally steady conditions (i.e. assuming that the local ozone concentration is constant), and assuming no gas-phase reactions and no mixing in the axial direction, the mean ozone concentration in a duct is governed by Volume 48 October 1998

(A6)

(A7)

In these expressions, Dh is the hydraulic diameter, Ac is the cross-sectional area of the duct, Re is the Reynolds number, ν is the kinematic viscosity of air, Sh is the Sherwood number, D is the diffusivity of ozone in air (0.167 cm2 sec-1 for ozone),25 Sc is the Schmidt number, η is the removal efficiency (between inlet and position x), and Co is the ozone concentration at the duct inlet. Substituting yields this differential equation and initial condition for η(x):

(A8)

Integrating between the duct inlet (x = 0) and outlet (x = L) yields the ozone total removal efficiency for the duct (A9) where the dimensionless duct length L* is defined by (A10) and the mean Sherwood number is defined to be (A11) Journal of the Air & Waste Management Association 949

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera Evaluation of eq A9 requires an estimate of Sh m. Here, the analogy between heat and mass transfer can be exploited, since a large number of correlations are available for the Nusselt number, Nu, which is analogous to Sh.26 Selection of an appropriate correlation depends on characterizing the flow conditions inside the duct and choosing a set of heat-transfer boundary conditions that are analogous to the boundary conditions of the mass-transfer problem under consideration. For flow, we will assume that the duct has a uniform cross section, that the flow is turbulent, and that the duct surfaces are either hydrodynamically smooth or rough. The treatment of boundary conditions is considered in the next few paragraphs. The flux of ozone to the duct wall must be equal to the rate at which ozone reacts with the surfaces. The surface uptake rate can be expressed in terms of an uptake coefficient, γ, using the molecular theory of gases: 15

The overall Nusselt number, accounting for heat transfer from the fluid to the exterior of the duct is likewise defined: (A17) Combining expressions A15–A17 yields an equation for the overall Nusselt number in terms of the Nusselt number for convective transport from air to the inner surface of the duct wall and the dimensionless wall resistance, expressed as 1/Bi: (x)

(A18)

For the case of fully-developed turbulent flow in smooth circular ducts, the Gnielinski correlation for Nu(x) has been found to be in overall accord with experimental data: 28,29 (A19)

(A12) where is the pollutant’s mean molecular speed (3.6 × 104 cm sec-1 for O3 at 293 K), and the net flux of species to the surface Js(x) is related to the total uptake rate U by the expression

where Pr = ν/α is the Prandtl number (α is the thermal diffusivity of the fluid), and f is the fully developed turbulent flow friction factor given by

(A20) (A13) The analogous heat-transfer problem is the case of convective heat transfer with finite thermal resistance normal to the wall and axially constant temperature at the exterior of the duct wall.27 In this case, the heat flux to the wall is given by (A14) where k is the thermal conductivity of the fluid, n is the coordinate perpendicular to the duct wall, hw is the wall heat-transfer coefficient, Tw(x) is the fluid temperature at the interior duct surface, and Tw0 is the temperature at the exterior of the duct. This expression can be rewritten in terms of the Biot number defined as Bi = hwDh/k: Bi(x)

(A15)

Heat transfer from the fluid to the walls of the duct is described in terms of the Nusselt number. The circumferentially averaged but axially local Nusselt number is defined as (A16) where Tm(x) is the flow-weighted average temperature at distance x along the duct. 950 Journal of the Air & Waste Management Association

The Gnielinski correlation (eq A19) is valid for the ranges 2100 ≤ Re ≤ 5 × 106 and 0.5 ≤ Pr ≤ 2000, and is applicable to both the case of a constant wall temperature boundary condition and the case of a constant heat flux boundary condition. Since these cases are the limiting cases of the finite thermal wall resistance boundary condition (i.e., the limits Bi = ∞ and Bi = 0, respectively), it follows that the Gnielinski correlation should also apply for the finite wall resistance case. To apply these results for ozone deposition in a duct, it is only necessary to transform the heat transfer solution (eqs A18–A20) to the mass-transfer problem under consideration. Eq A19 can be used by substituting Sc for Pr and Sh(x) for Nu(x):

(A21) Since the Schmidt number for O3 in air is Sc = 0.92, the transformed Gnielinski correlation will apply for ozone mass-transfer to duct surfaces provided Re is within the valid range, the duct walls are smooth, and flow in the duct is fully turbulent. A mass-transfer analog of the Biot number can be developed by rewriting eq A12 in a form that resembles eq A14: (A22) Volume 48 October 1998

Morrison, Nazaroff, Cano-Ruiz, Hodgson, and Modera where the (fictitious) ozone concentration at the outer surface of the duct, Cw0, is set to 0. The resulting analog to Bi is the wall Sherwood number Shw:

this case, eqs A20 and A21 must be replaced with analogous expressions for rough surfaces 28 (A25)

(A23) Since both Shw and Sh from the Gnielinski correlation are, in fact, independent of x, they can be combined using the analog to equation (A18) to give the mean Sherwood number, Shm, directly:

(A26)

(A24)

(A27)

Given Shm, the overall ozone removal efficiency in a duct is determined from eq A9. Predictions are shown in Figure 7 for the case of perfect removal at the walls (γ = 1). This figure shows that in the central supply duct of a typical commercial ventilation system (circular duct with D = Dh = 1.5 m, L = 30 m, and Q = 8.25 m3 sec-1, so Re = 5 × 105), about 13% of the ozone entering the duct would be removed by smooth walls which act as perfect sinks. Because of higher values of L/Dh, removal could be much higher in the secondary supply ducts which branch off the main duct. For example, in a circular duct with Dh = D = 0.5 m, L = 50 m, and Q= 0.5 m3 sec-1, so that Re = 8.5 × 104, the ozone removal efficiency could be as high as 63%. Ozone removal also depends on the surface uptake coefficient, γ, of the lining material as illustrated by Figure 8 (solid lines). For the example just considered, ozone removal efficiency diminishes from 63% for perfect uptake at the smooth surface (γ = 1) to 12% for γ = 10-5 (a typical uptake coefficient for duct liners) and to just 1.4% for γ = 10-6 (a representative uptake coefficient for galvanized sheet metal). The same general approach can be applied to predict ozone loss on hydrodynamically rough duct surfaces. In

where ε is the characteristic roughness length and Reε is the Reynolds number based on the surface roughness length scale and the friction velocity . Figure 8 (dashed and dotted lines) shows that the influence of roughness on ozone removal efficiency is small when γ < ~ 10-5 and large for γ > 10-3. For the examples previously considered, if Re = 8.5 × 104, γ = 10-5, and ε/Dh = 10-2 (as might be expected for duct liner), the ozone removal efficiency remains 12%, as predicted for a smooth surface. Likewise, for γ = 10 -6 and ε/Dh = 10 -3 (as might be expected for galvanized sheet metal), the removal efficiency remains 1.4%. Eqs A9, A10, and A21–A27 provide a means of computing the ozone scavenging efficiency for fixed values of the uptake coefficient, γ. To incorporate the effects of aging, we used a numerical method that incorporates experimental information on the reduction in γ with exposure, as described by eq 7.

Figure 7. Ozone removal efficiency in smooth ducts in which the surface acts as a perfect sink, plotted as a function of Reynolds number and length-to-hydraulic diameter ratio (L/Dh). Volume 48 October 1998

Figure 8. Effect of surface uptake coefficient (γ) and roughness (ε/ Dh) on the ozone removal efficiency as a function of Reynolds number in a duct with a length of 50 m and a hydraulic diameter of 1 m. Note that eq A25 does not strictly apply for ε/Dh = 0.001 where Re