lable at ScienceDirect

Building and Environment 75 (2014) 1e10

Contents lists avai

Building and Environment

journal homepage: www.elsevier .com/locate/bui ldenv

Gas-phase filters breakthrough models at low concentration e

Effect of relative humidity

Ali Khazraei Vizhemehr, Fariborz Haghighat*, Chang-Seo LeeDepartment of Building, Civil and Environmental Engineering, Concordia University, Montreal, Quebec H3G 1M8, Canada

a r t i c l e i n f o

Article history:Received 31 October 2013Received in revised form30 December 2013Accepted 17 January 2014

Keywords:Relative humidityBreakthroughVOCModelGas-phase filter

* Corresponding author. Tel.: þ1 514 848 2424 319E-mail addresses: Fariborz.Haghighat@Concordia

(F. Haghighat).

0360-1323/$ e see front matter � 2014 Elsevier Ltd.http://dx.doi.org/10.1016/j.buildenv.2014.01.010

a b s t r a c t

Indoor air quality (IAQ) is a major concern in non-industrial buildings since it influences occupants’health, comfort and productivity. Adsorption-based granular activated carbon (GAC) filters are commonlybeing used to purify indoor air by removing indoor air pollutants. Predicting the breakthrough time offilters is necessary for scheduling their maintenance and/or regeneration. However, the pollutants’concentrations that typically encounter in indoor environment are very low thus increasing the influenceof humidity on the filter performance. Also, the existing standard recommends the test to be carried outat 100 ppm, which is much higher than the actual volatile organic compounds (VOC) concentration inbuildings. This paper reports the development of a framework for predicting the breakthrough curve ofactivated carbon filters at low concentration and different levels of relative humidity applying acceler-ated test data. The procedure is based on two well-known empirical models: WheelereJonas and YooneNelson equations. The overall mass transfer coefficient in the WheelereJonas equation and the pro-portionality constant in the YooneNelson equation (both as a function of adsorption capacity) are afunction of humidity level. Results show that the proposed framework allows the breakthrough time athumid condition and low contaminant concentration to be estimated using the data obtained from theexisting standard test procedure.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Predicting the breakthrough time (BT) of filters is necessary forscheduling their maintenance, regenerating and/or replacement.One approach to predict the breakthrough time is to develop cor-relation among the influential parameters using the experimentaldata taking into consideration the effect of environmental param-eters. In this way, the developed model can be applied to providethe user with the information for estimating the filter performanceunder conditions of actual use. Recently, a framework was pro-posed for predicting the dynamic performance of granular acti-vated carbon (GAC) filters at dry conditions; see Fig. 1 [1]. Twoapproaches were proposed where in the first approach the value of50% breakthrough time (s) corresponding to the YooneNelsonequation was estimated from either the linear function of inletconcentration (Method 1) or the 10% breakthrough time (tb,10%)(Method 2); and in the second approach the value of adsorption

2; fax: þ1 514 848 7965..ca, haghi@bcee.concordia.ca

All rights reserved.

capacity corresponding to the WheelereJonas equation was ob-tained from the extrapolated value of validated adsorptionisotherm fitted to the experimental data (Method 3). Both param-eters were later used to predict the breakthrough curve at lowlevels of concentration. It was demonstrated that the proposedframework can predict with a good accuracy in the range of15e300 ppm concentration. However, further research is needed toverify the applicability of the proposed model in the lower range ofconcentrationwhere relative humidity could play an important roleon the filter performance.

Although the air containing the organic vapor is seldom free ofwater vapor, most of the earlier studies on the developmentof predictive breakthrough models focused on dry air conditions[1e6]. Therefore, in order to generalize the methodology there is aneed to study the effects of humidity on the influencing parame-ters. Previous experimental studies show that adsorption capacitydecreases with increasing relative humidity (RH) particularly athigher range of RH [7e10]. This can be explained by the capillarycondensation effect of water vapor, at the active sites, on the sur-face of the micropore in view of Kelvin equation [11]. Further, it wasreported that adsorption of organic vapors, water-soluble andinsoluble compounds, behave differently under humid conditions

Fig. 1. Different pathways for quantification of breakthrough time at low concentra-tions using WheelereJonas and YooneNelson equation.

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e102

[12,13]. Cooperative adsorption takes place between hydrophilicvolatile organic compounds (VOCs) and water vapor up to certainhumidity levels while there is always competitive adsorption forhydrophobic ones [14].

However, few studies considered the effect of VOCs concentra-tion level in their studies. Some studies mentioned that the RHeffect is more pronounced at the lower adsorbate concentrationsthan at the higher concentrations [15e18] while they did not sug-gest any practical and/or procedure to demonstrate the effect ofconcentration change at different RH levels.

This paper first gives a brief review of adsorption isotherms,breakthrough predictors, their parameters and the effect of envi-ronmental conditions on each term. It then, reports the outcomes ofa series of experiments which were carried out on a small-scale set-up for a large range of concentrations, and finally proposes a pro-cedure to estimate the filter breakthrough time/performance at lowconcentration using the experimental results from high concen-tration and different relative humidity levels.

2. Models

2.1. Isotherm models

Adsorption isotherm relates the sorbed-phase concentration(capacity) to the air-phase concentration. There are various models

Table 1Representative adsorption isotherm models.

Model Cse ¼ f ðCeÞ Mode

Linear Cse ¼ KpCe Kp is tLangmuir Ce

Cse¼ 1

Cs0KLþ�

1Cs0

�Ce Cs0 is

Freundlich ln Cse ¼ ln Kf þ 1n Ce n is Fr

DubinineRadushkevich (DeR)Cse ¼ C0

s0 exp�D

h�RT ln

�P0P

��2i

C0s0 is

constaBrunauer, Emmett, and Teller (BET) ðP=P0Þ

Cseð1�P=P0Þ ¼ 1cC00

s0þ c�1

cC00s0ðP=P0Þ c is a

a mon

a Ce is the equilibrium air-phase concentration within the pores (mg/m3air), Cse is the

phase concentration) (mg/gsolid), R is the universal gas constant (8.314 J/(mole K)); T is thtemperature T, and P is the partial pressure of the sorbate in the gas.

to describe this relation (type I to V based on Brunauer’s classifi-cation). Axley [19] stated that for sorption of air pollutant inbuilding materials, Langmuir and Linear models are the mostappropriate choices. For sorption of any gas phase pollutants, if itsconcentration is within one order of magnitude of its saturatedvalue, the BET model should be used. DeR and Freundlich modelsare used for industrial sorbents which show nonlinear equilibriumbehavior [1,19]. Table 1 summarizes the linearized form of afore-mentioned isotherms and their parameters.

2.2. Breakthrough models

The three methods mentioned in the framework integrate theapplication of two well-known empirical equations, see Fig. 1. Acareful examination of these methods shows that each equationmay be expressed as t ¼ X þ YZ. The corresponding terms arepresented in Table 2.

X and Z terms are the same in both YooneNelson andWheelereJones equations while Y term is defined as M/k0Q in the Yoon’sequation and as rb/kv in the Wheeler’s equation. Consistent withthe preceding discussion, the only difference observed in thetheoretical breakthrough models is attributed to the Y terms.

2.3. Investigation of parameters

From the earlier developed framework (Fig. 1), it can be seenthat there are four criteria which indicate the independency orweak dependency of the breakthrough time with the concentra-tion. Detailed information of these indicators is necessary in orderto generalize the application of the developed framework.

1) Mass transfer coefficient (Kv and k)

It was experimentally explored that Kv in the WheelereJonasequation and k (k ¼ k0s) in the YooneNelson equation, are weakfunction of inlet concentration [1]. This conclusion is in conformitywith Lodewyckx and Vansant’s developed equation for Kv inwhich it is a function of both adsorbent (size and capacity) andadsorbate (molecular weight and similarity coefficient) charac-teristics [20]:

Kv ¼ 800b0:33u0:75

d1:5p

ffiffiffiffiffiffiffiffiffiffiCseMW

r(1)

where dp is the mean adsorbent size, u is the linear velocity, and b isthe similarity coefficient in the DeR equation. As long as the samemedium is used as sorbent and the same compound is used asadsorbate, these properties remain constant. Also it is reportedthat [5]:

l parametersa

he partition or distribution coefficient or Henry’s constantthe maximum adsorption capacity (mg/g) and KL is the affinity constant (m3/mg)

eundlich exponent and Kf is Freundlich constant.

the maximum capacity available for the adsorbate (mg/g); D is the microporositynt (mL/J)dimensionless constant; C00

s0 is the amount of sorbent (capacity) required to formolayer of the adsorbate (mg/g)

equilibrium adsorbate concentration in solid phase (adsorption capacity or sorbed-e absolute temperature of the system; P0 is the sorbate saturation vapor pressure at

Table 2Empirical equations for breakthrough time.

Model t ¼ X þ YZ Model parametersa

YooneNelson t ¼ CseMCeQ

þ CseMk0CeQ

� ln CbCi�Cb

or

tb ¼ sþ sk ln

CbCi�Cb

k’ is a rate constant (min�1)s is the time required for 50%breakthroughk ¼ k0 s is a product constant

WheelereJonas t ¼ CseMCeQ

þ CserbkvCe

� ln CbCi�Cb

kv is the adsorption rateconstant (min�1)

a tb is the breakthrough time (min); M is the weight of carbon media bed (g); Ciand Cb are the upstream (inlet) and downstream (breakthrough) concentration (mg/m3), Q is the volumetric airflow rate (m3/min); rb is the bulk density of the packedbed (kg/m3).

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e10 3

k0 ¼ kCeQCse

¼ ks

(2)

where k is a dimensionless constant of proportionality, and it isindependent of the inlet concentration. However, the overall masstransfer coefficient, Kv, is likely to be influenced by humiditychanges.

The presence of water molecules, hinder the transport oforganic vapor molecules through the pore system and slow downthe kinetics. A simple model to demonstrate the moisture effect onthe adsorption rate coefficient was introduced by Wood andLodewycks [21]. The empirical correlation between the wet/dryrate coefficient ratio and the wet/dry breakthrough time ratio (oradsorption capacity ratio) is as follows [21]:

KvðwetÞ ¼�0:33þ 0:67

�tbðwetÞtbðdryÞ

orCseðwetÞCseðdryÞ

KvðdryÞ (3)

This equation shows the indirect effect of adsorbed water on anorganic vapor adsorption rate coefficient which has the upper limitof Kv (dry). The same form of Equation (3) can be written for k(dry)and k(wet) constant of YooneNelson equation.

2) Stoichiometric breakthrough time (s)

Penetration curves can be considered approximately symmet-rical. According to Fig. 2, the maximum amount of a contaminantcan be taken up by a filter (this corresponds to equilibrium condi-tionwhich is the value Cse of the adsorption isotherm at the influentconcentration) can be expressed by:

CseM ¼ QðXþ YÞ (4)

whereM is the total weight of adsorbent, and X and Y are the areasabove the breakthrough curve which are equal, see Fig. 2.Furthermore, the inflexion point of the penetration curve has theco-ordinates of (Ci/2, s). Yoon and Nelson [5] used this value of s intheir model development.

Fig. 2. Ideal symmetrical pene

The influence of the adsorption capacity on the breakthroughtime is more pronounced than the overall mass transfer coefficient[22]. Water vapor adsorption of the activated carbon specimenslowers the available adsorption volume and hence Cse. The value ofadsorption capacity in humid conditions can be obtained from theextended Langmuir isotherm equation. The equation is written foreach component of the mixture (VOC and water vapor) where irepresents the specific contaminant and j represents all the gaseouspollutants in the air. In this study, water is considered as a singlepollutant which adsorbed onto the filter. Correspondingly, the Kw

constant will be calculated from the linearization of waterisotherm.

Cse;iðwetÞ ¼ KL;i$Cs0;i$Ce;i1þPn

j¼1 KL;j$Ce;j¼ KLCs0Ce

1þ KLCe þ KwCw(5)

where Kw is the water equilibrium constant, Cw is the water vaporconcentration (RH). The ratio of adsorption capacity at wet condi-tion (Cse,wet) to that at the dry air condition (Cse,dry) can be definedas:

CseðwetÞ ¼ CseðdryÞh1þ KwCw;satðRHÞ

1þKLCe

i (6a)

Correction Factor ¼ 1þ KwCw1þ KLCe

(6b)

where Cw ¼ Cw,sat(RH), Cw,sat is the water vapor concentration atsaturation (100% RH) and RH is expressed as a decimal fraction. As sis a direct function of Cse (see Fig. 2), the same form of Equation (6)can relate the s (dry) and the s (wet).

3) (t50/t10) ratio

t10 and t50 have specific characteristics and can be used to verifythe accuracy of the measured data and measurement technique.Considering the experimental breakthrough curves, t10 is a point inwhich the displacement phenomenon starts while t50 is a pointwhere the concavity of the penetration curves changes. However,the t50/t10 ratio remains constant [1,2,23], thus can be used as anindicator to develop a predictive tool.

3. Method and materials

A series of adsorption tests was performed to collect therequired data using a small scale experimental setup according tothe ASHRAE Standard 145.1 [24]. Table 3 depicts the properties ofselected contaminants which are among the predominant VOCsfound in a typical indoor environment.

tration curve calculation.

Table 3Properties of selected VOCs.

Chemical (supplier, grade) Chemicalcategory

Molecularformula

Molar mass(g/mol)

Boilingpoint (�C)

Vapor pressure at20 �C (mm Hg)

Solubility in waterat 20 �C (g/L)

Polarity Density (g/mL)

n-Hexane (Fisher Scientific, 95%) Alkane C6H14 86.2 69 132 0.013 Non-Polar 0.655MEK (Fisher Scientific, Certified ACS) Ketone C4H8O 72.1 80 78 290 Polar 0.805

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e104

The small scale set-up was a 5 cm (2 inch) diameter cylinderfilled with 25 g of cylindrical GAC; see Fig. 3. The experiments wereconducted using MEK and n-hexane as challenge gas, and theywere conducted at eight different concentrations: MEK at 1, 5, 15,30, 50, 70, 100, 200 ppm, and n-hexane at 1, 5, 15, 30, 60, 100, 150,200 ppm. For dry air condition and/or adsorption isotherm tests, airpassed through desiccators to be dehumidified and it then wasmixed with selected VOC. For humid air condition experiments, airpassed through the humidifier. The challenged gases were thenintroduced to the clean dry/humid air (upstream line) at 30 lit/minflow rate at 23� 1 �C. A photo-acoustic multi-gas analyzer detector(B&K Air Tech Instrument 1302) was used to collect samples fromthe downstream of GAC filter. Some of the experimental data wereused to develop the framework and some were applied to validatethe proposed framework.

4. Results and discussion

4.1. Adsorption isotherm fitting

The extrapolated data from the selected adsorption isothermsare used to obtain the sorbed-phase concentration in order to findthe breakthrough time using WheelereJonas equation (see themethod 3 of Fig. 1). The experimental data from small scale test forMEK and n-Hexane adsorption on the GAC were used to obtain theadsorption isotherms for Langmuir, Freundlich, DeR and BETisotherm model using the least-square technique.

4.1.1Langmuir modelThis model is the simplest one and widely used for monolayer

adsorption. Fig. 4 shows the results of regression for the dry

Fig. 3. Schematic diagra

Fig. 4. Results of regression using Langmu

condition. It can be seen that Langmuir model did not fit well to theMEK data as it did for n-Hexane. The R2value are 0.965 and 0.994 toMEK and n-Hexane, respectively.

4.1.2Freundlich modelAs an empirical equation, this model has an exponent (n) which

indicates the nonlinearity of the isotherm. Fig. 5 shows the result ofregression analysis for the Freundlich equation. It can be seen thatFreundlich model did not fit the data as good as the Langmuirmodel, as indicated by the lower value of R2 for both compounds. Apoor fit for MEK is often found at low concentrations.

4.1.3DeR modelDeR model is widely used to describe the VOC adsorption on

GAC media: It is a semi-empirical equation, originally derived frommicropore filling theory and depicts pore filling adsorption ratherthan a layer-covering one. The layer-covering theory was assumedfor the development of Langmuir and Freundlich isotherms. How-ever, as Freundlichmodel, DeR does not conform to the Henry’s lawregion at low concentration range [25,26]. The regression resultsfor DeR gave relatively good fitness for both compounds with thesame regression error. The fitting data is plotted in Fig. 6.

4.1.4BET modelMonolayermolecular adsorption occurs in the solids micropores

which have pore size not greater than the adsorbate molecule size.Thus, the adsorption limit is governed by the accessible microporevolume [27]. The BET model can be used for the adsorbents with awide range of pore sizes in which the monolayer adsorption isextended to multilayer adsorption and then to capillary conden-sation by increasing the load of adsorbate [11,19]. Fig. 7 shows that

m of the test duct.

ir isotherm for (a) MEK (b) n-Hexane.

Fig. 5. Results of regression using Freundlich isotherm for (a) MEK and (b) n-Hexane.

Fig. 6. Results of regression using DeR isotherm for (a) MEK (b) n-Hexane.

Fig. 7. Results of regression using BET isotherm for (a) MEK (b) n-Hexane.

Table 4Determination of the isotherm constants for four models.

Langmuirmodel

Maximum adsorptioncapacity Cs0 (mg/g)

Affinity KL (m3/mg) R2

MEK n-Hexane MEK n-Hexane MEK n-Hexane136.99 232.56 0.015 0.012 0.965 0.994

Freundlichmodel

constant Kf exponent n R2

MEK n-Hexane MEK n-Hexane MEK n-Hexane12.203 19.846 2.558 2.799 0.971 0.982

DeR model Maximum adsorptioncapacity Cs0 (mg/g)

constant D R2

MEK n-Hexane MEK n-Hexane MEK n-Hexane312.62 433.8 4E-09 4E-09 0.987 0.987

BET model Monolayer adsorptioncapacity Cs0 (mg/g)

Constant c R2

MEK n-Hexane MEK n-Hexane MEK n-Hexane144.89 232.50 3451 4301 0.961 0.993

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e10 5

the BET model gives the worst fit for MEK while for n-Hexane itprovides a better fit than Freundlich and DeR models.

Table 4 gives the fitted isotherm parameters for the four selectedcases, and it shows that the Langmuir and DeR isotherm providesthe best fit for n-Hexane and MEK, respectively.

4.2. Breakthrough curve investigation

Breakthrough curve data at high concentration are used todevelop method 1 and 2 of Fig. 1 in order to estimate the 50%breakthrough time needed in the YooneNelson equation at dry aircondition. Figs. 8 and 9 give the breakthrough times of GAC filtersfor MEK and n-Hexane at various concentrations. All curvesexhibited asymmetrically sigmoid shapes with steeper break-through curves at higher concentrations, meaning that equilibriumwas attained faster for higher gas concentrations. Over time, the gasconcentration (MEK or n-Hexane) in the bed and the breakthrough

Fig. 8. Experimental breakthrough curves of n-Hexane adsorption on GAC at various concentrations and dry air conditions.

Fig. 9. Experimental breakthrough curves of MEK adsorption on GAC at various concentrations and dry air conditions.

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e106

time increase and the outlet concentration eventually reach theinlet concentration. At this point (tb¼ Cout/Cin¼ 1), themedia bed issaturated and there is no more adsorption. However, at very lowconcentration, the breakthrough curve is smoother. At the higherconcentrations, the inflection point (where the breakthrough graphswitches from concave down to concave up or vice versa) is the

Fig. 10. Different pathways for quantification of BT at very low concentration and diff

stoichiometric breakthrough time while at 1 ppm and/or less, thisvalue is very close to the saturation point.

As expected, the ratios of 10% and 50% (stoichiometric) break-through time for any two randomly selected inlet concentrationsgive a consistent value (proportionality constant). Referring to themethod 2 of Fig. 1, this number is used as an indicator to verify the

erent relative humidity levels using YooneNelson and WheelereJonas equations.

Fig. 11. Validation of the proposed methods for 1 and 5 ppm of MEK at dry condition.

Fig. 12. Validation of the proposed methods for 1 and 5 ppm of n-Hexane at dryconditions.

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e10 7

measurement accuracy and to rapidly quantify the breakthroughtime at low concentration levels using YooneNelson equation.However, the rest of ratios (e.g. t30/t10 and t80/t10) do not follow thesame pattern.

Fig. 13. Predicted breakthrough curves with various inlet

4.3. Development of a framework

Fig. 10 demonstrates a framework that can be used to predictthe breakthrough time of a gas-phase air-cleaning filter at lowconcentration using high concentration experimental data atdifferent level of relative humidity. Two different approaches areproposed:

4.3.1. Applying the YooneNelson equationIn this approach, the breakthrough time (BT) value is calculated

according to the following steps:

1) Find the value of s from either predicted 50% BT (tb,50%) (method1 in Fig. 1) or 10% BT (tb,10%) (method 2 in Fig. 1). If there isadequate data approaching to 50% saturation, s can be directlyobtained from the predicted breakthrough time correlation atany level of inlet concentration. Otherwise, s can be obtainedfrom the 10% BT as an indicator number to be substituted in themain equation. Using the tb,10% instead of tb,50% remarkably de-creases the experimentation time.

2) Find the correction factor (Equation (6b)) using water adsorp-tion isotherm,

3) Calculate the corrected value of s (swet),4) Calculate the corrected value of k (kwet) from Equation (3) using

either the correction factor or from the ratio of any percentile ofbreakthrough times if the experimental data is available, and

5) By substituting the modified value of k and s (the only RH-dependent parameters in the model) the penetration profile(Cb/Ci) versus the elapsed time (tb) can be plotted.

4.3.2. Applying the WheelereJonas equation

1) Find Cse value from extrapolated adsorption isotherm (method 3of Fig. 1),

2) Find the correction factor (Equation (6b)) fromwater adsorptionisotherm equation at the given humidity level,

3) Calculate the corrected value of Cse for the given humidity level(Cse,wet) (Equation (6a)),

4) Calculate corrected value kv (kv,wet) from Equation (3) usingeither the correction factor or from the ratio of breakthroughtimes if the experimental data is available at required RH level.Themass transfer coefficient at dry air condition, kv, is a constantwhich can be obtained either from one adsorption test at anyconcentration or more accurately as mean value of some highlevel concentration adsorption tests’ results, and

MEK and n-Hexane concentrations at dry condition.

Fig. 14. Water isotherm on GAC at 23 �C.

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e108

5) After substituting the modified value of kv and Cse as the onlyRH-dependent parameters in the model, the breakthroughprofile can be obtained.

4.4. Framework validation and prediction

Using the proposed methodology for estimating the break-through time, a series of simulated breakthrough curves for 1 ppmand 5 ppm concentrations were generated at dry air condition, asshown in Figs. 11 and 12. These Figures show that there is goodagreement between the three method’s prediction and the mea-surement. The agreement between the experimental and predictedcurve was determined by estimating the relative error of break-through time for 5 ppm MEK and n-Hexane. The relative error fortb10%, tb50% and tb80% were �3.2%, 1.1%, 5.8% and 0%, 2.1%, 1.3%,respectively.

Given that the realistic operating conditions for indoor HVACunits typically require that the adsorptive filter to be operated atlow adsorbate concentration, a series of MEK and n-Hexanebreakthrough profiles were also generated (using YooneNelsonmethod 1) to predict the lifetime of filter as shown in Fig. 13. Asexpected, 10 ppb test has longer breakthrough time, almost oneorder of magnitude more than 100 ppb test. The results are in therange of the field experience data reported by Graham and Bayati[29].

To validate the proposed framework (Fig. 10) at humid condi-tions, the predicted breakthrough curve for 100 ppm of MEK at dryand 50% RH as well as 1 ppm of MEK at dry and 40% RH werecompared with the experimental results (using YooneNelsonmethod 1). The water vapor isotherm constant (Fig. 14) was used toidentify the modified parameters (mass transfer coefficients, stoi-chiometric breakthrough time and adsorption capacity)

Fig. 15. Validation of the proposed methods for 100 ppm of MEK.

(Kw ¼ 0.015). As demonstrated in Figs. 15 and 16, there is a goodagreement between the prediction made by the framework and themeasurements for both scenarios.

The validated framework was used to predict the breakthroughtime at relative humidity levels of 0%, 20%, 50%, 80% and 100%, asshown in Fig. 17. The presence of water vapor caused the shape ofMEK and n-Hexane breakthrough curve to be narrower at higherlevels of RH.

The predicted breakthrough curves also compared for 1 ppm ofMEK and n-Hexane, as shown in Fig.18. Although it shows the samedeclining pattern in the reduction of breakthrough time, theadsorbent performance at low concentrations of adsorbate isaffected by the relative humidity level to a greater extent than bythe higher concentration. As an example, the breakthrough curvefor 1 ppm concentration of MEK shows a steeper decrease at RH of20% or higher resulted in a shift of filter service life from 300 h to120 h for dry and wet conditions, respectively. The results are ingood agreement with those reported earlier [16,28,30,31,32].

5. Conclusion

This paper reports the development of a methodology to esti-mate the service life of gaseous filters in a real built environmentconditions: An environment with low concentration of VOC andhumid air. The adsorption isotherm determined for MEK and n-Hexane was used to extrapolate the adsorption capacity data fromhigh to low range of concentration. The effect of RH on four effec-tive criteria (Kv, Cse in the WheelereJonas equation and s, k in theYooneNelson equation) was investigated. The parameters requiredfor the application of extended Langmuir isotherm including theVOC and water concentration as well as their equilibrium constantswere obtained from individual adsorption isotherm tests. Also, the

Fig. 16. Validation of the proposed methods for 1 ppm of MEK.

Fig. 18. Predicted breakthrough curves for 1 ppm MEK and n-Hexane adsorption at various relative humidity levels.

Fig. 17. Predicted breakthrough curves for 200 ppm MEK and n-Hexane adsorption at various relative humidity levels.

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e10 9

correction factor calculated from water adsorption isotherm wasintegrated with the dry air conditions’ parameters to evaluate theinfluence of humidity on the breakthrough time of GAC. The val-idity of modified framework was verified with measured data for 1and 5 ppm of MEK (as a polar compound) and n-Hexane (as a non-polar compound) at dry condition as well as 1 and 100 ppm of MEKat two specified level of RH. Using this procedure, one canextrapolate data to untested vapor concentrations and relativehumidity conditions.

Acknowledgments

The authors would like to express their gratitude to the NaturalSciences and Engineering Research Council of Canada for sup-porting this research project through a CRD grant.

References

[1] Khazraei Vizhemehr A, Haghighat F, Lee CS. Predicting gas-phase air-cleaningsystem efficiency at low concentration using high concentration results:development of a framework. Build Environ 2013;68:12e21.

[2] Shiue A, Den W, Kang YH, Hu SC, Jou Gt, Hu V, et al. Validation and applicationof adsorption breakthrough models for the chemical filters used in the make-up air unit of a cleanroom. Build Environ 2011;46:468e77.

[3] Wu J, Claesson O, Fangmark I, Hammarstrom LG. A systematic investigation ofthe overall rate coefficient in the WheelereJonas equation for adsorption ondry activated carbons. Carbon 2005;43:481e90.

[4] Tsai WT, Chang CY, Ho CY, Chen LY. Simplified description of adsorptionbreakthrough curves of 1, 1-Dichloro-1-fluoroethane (HCFC-141b) on acti-vated carbon with temperature effect. J Colloid Interface Sci 1999;214:455e8.

[5] Yoon YH, Nelson JH. Application of gas adsorption kinetics I. A theoreticalmodel for respirator cartridge service life. AIHA J 1984;45(8):509e16.

[6] Jonas LA, Rehrmann JA. Predictive equations in gas adsorption kinetics. Carbon1973;11(1):59e64.

[7] Shin HC, Park JW, Park K, Song HC. Removal characteristics of tracecompoundsof landfill gas by activated carbon adsorption. Environ Pollut 2002;119:227e36.

[8] OwenMK, VanOsdell EW, Jaffe LB, Sparks LE. The effect of relative humidity ongaseous air cleaner media performances: toluene adsorption by activatedcarbon. In: Proceedings’ of engineering solutions to indoor air quality prob-lems. Research Triangle Park, NC: Air and Waste Management Association;1996. pp. 551e62. VIP-51.

[9] CalMP, RoodMJ, Larson SM. Gas phase adsorption of volatile organic compoundsand water vapor on activated carbon cloth. Energy Fuels 1997;11(2):311e5.

[10] Pei J, Zhang J. Determination of adsorption isotherm and diffusion coefficient oftolueneonactivated carbonat lowconcentrations. BuildEnviron2012;46:66e76.

[11] Ruthven DM. Principles of adsorption and adsorption process. John Wiley &Sons, Inc.; 1984.

[12] Biron E, Evans MJB. Dynamic adsorption of water-soluble and insoluble vaporson activated carbon. Carbon 1998;36(7e8):1191e7.

[13] Qi N, Appel S, Levan MD, Finn JE. Adsorption dynamics of organic compoundsand water vapor in activated carbon beds. Ind Eng Chem Res 2006;45:2303e14.

[14] Haghighat F, Lee CS, Pant B, Bolourani G, Lakdawala N, Bastani A. Evaluation ofvarious activated carbons for air cleaning e toward design of immune andsustainable buildings. Atmos Environ 2008;42:8176e84.

[15] Nelson GO, Harder CA. Respirator cartridge efficiency studies: VI. Effect ofconcentration. AIHA J 1976;37(4):205e16.

[16] Werner MD. The effects of relative humidity on the vapor phase adsorption oftrichloroethylene by activated carbon. Am Ind Hyg Assoc J 1985;46(10):585e90.

[17] Chiang PC, Lee MD, Lin TF, Chiang HL. Effects of relative humidity on VOCsadsorption. Helsinki, Finland: Indoor Air; 1993. pp. 399e404.

[18] Scahill J, Wolfrum EJ, Michener WE, Bergmann M, Blake DM, Watt AS. A newmethod for the rapid determination of volatile organic compound break-through times for adsorbent at concentrations relevant to indoor air quality.J Air Waste Manag Assoc 2004;54:105e10.

[19] Axley JW. Tools for the analysis of gas-phase air-cleaning systems in buildings.In: Proceedings of the ASHRAE annual meeting 1994. pp. 1130e46.

[20] Lodewyckx P, Vansant EF. Estimating the overall mass transfer coefficient kvof the WheelereJonas equation: a new and simple model. AIHAJ Am Ind HygAssoc 2000;61(4):501e5.

[21] Wood GO, Lodewyckx P. Correlations for high humidity corrections of ratecoefficient for adsorption of organic vapors and gases on activated carbons in

A. Khazraei Vizhemehr et al. / Building and Environment 75 (2014) 1e1010

air-purifying respirator cartridges. J Int Soc Resp Prot 2002;(Spring/Summer):58e64.

[22] Lodewyckx P, Vansant EF. The influence of humidity on the overall masstransfer coefficient of the WheelereJonas equation. AIHAJ Am Ind Hyg Assoc2000;61(4):461e8.

[23] Agranovski IE, Moustafa S, Braddock RD. Performance of activated carbonloaded fibrous filters on simulataneous removal of particulate and gaseouspollutants. Environ Technol 2005;26(7):757e66.

[24] ASHRAE Stand. 145.1. Laboratory test method for assessing the performanceof gas-phase air-cleaning systems: loose granular media. Atlanta: AmericanSociety of Heating, Refrigerating and Air-Conditioning Engineers, Inc.; 2008.

[25] Yao M, Zheng Q, Hand DW, Perram L. Investigation of the treatability of theprimary indoor volatile organic compounds on activated carbon fiber cloths attypical indoor concentrations. J Air Waste Manag Assoc 2009;59(7):882e90.

[26] Cal MP, Larson SM, Rood MJ. Experimental and modeled results describing theadsorption of acetone and benzene onto activated carbon fibers. Environ Prog1994;13(1):26e30.

[27] Noll KE, Gounaris V, HouW. Adsorption technology for air and water pollutioncontrol. Lewis Publishers; 1992.

[28] Lodewyckx P, Wood GO, Ryu SK. The WheelereJonas equation: a versatile toolfor the prediction of carbon bed breakthrough times. Carbon 2004;42(7):1351e5.

[29] Graham JR, Bayati MA. The use of activated carbon for the removal of traceorganic in the control of indoor air quality. Proc IAQ Clim 1990;3:133e8.

[30] Wood G. A model for adsorption capacities of charcoal beds: II. Challengeconcentration effects. Am Ind Hyg Assoc J 1987;48(8):703e9.

[31] Wood G. Estimating service lives of organic vapor cartridges II. a single vaporat all humidities. J Occup Environ Hyg 2004;1(7):472e92.

[32] Abiko H, Furuse M, Takano T. Reduction of adsorption capacity of coconutshell activated carbon for organic vapors due to moisture contents. Ind Health2010;48:427e37.