the heterogeneous decomposition of ozone on atmospheric mineral dust surrogates at ambient...

The HeterogeneousDecomposition of Ozone onAtmospheric Mineral DustSurrogates at AmbientTemperatureFEDERICO KARAGULIAN, MICHEL J. ROSSI

Laboratoire de Pollution Atmospherique et Sol (LPAS), Ecole Polytechnique Federale de Lausanne (EPFL),Station 6, CH-1015 Lausanne, Switzerland

Received 6 October 2005; revised 9 December 2005; accepted 9 December 2005

DOI 10.1002/kin.20175Published online in Wiley InterScience (www.interscience.wiley.com).

ABSTRACT: The rate of uptake of ozone on various mineral dust surrogates, expressed as uptakecoefficient γ, has been studied employing a Knudsen flow reactor. Experiments were performedat T = 298 ± 2 K on substrates of kaolinite, CaCO3, natural limestone, Saharan dust, and Arizonatest dust. Initially, the uptake coefficients have been calculated on the basis of the geometricsurface area of the powder samples. Both initial and steady-state uptake coefficients γ0 and γss

were found very similar for all the examined substrates. In addition, additional uptake experi-ments on marble sample have shown that γ0 and γss may be overestimated between a factor of50 and 100, respectively. Based on these considerations, we proposed initial and steady-stateuptake values of the order of 10−4 and 10−5, respectively. On kaolinite, the uptake coefficientdecreased with increasing O3 residence time τg thus indicating a complex mechanism. In con-trast, γ decreased and became independent of τg at long residence time after long exposure toO3. For all uptake experiments the disappearance of O3 was accompanied by the formation ofO2. The different mineral dust surrogates may be more accurately distinguished by their time-dependent O2 yield r (t) rather than the magnitude of γ. The heterogeneous reaction of O3 onmineral dust has been found to be noncatalytic and of limited importance in the atmosphere.C© 2006 Wiley Periodicals, Inc. Int J Chem Kinet 38: 407–419, 2006

Correspondence to: Michel J. Rossi; e-mail: [email protected].

Contract grant sponsor: Office Federal de l’Enseignement et dela Science (OFES).

Additional chemical kinetic data in tabular form (Tables E.1to E.3) are available as “Supplementary Material” at http://www.interscience.wiley.com/jpages/0538-8066/suppmat/.c© 2006 Wiley Periodicals, Inc.

INTRODUCTION

Ozone decomposition on mineral dust is a reaction ofpotential atmospheric significance that has attractedconsiderable attention [1–6]. Soil derived mineral dustis a significant component of atmospheric aerosols. Ona global scale, the largest contributions originate fromarid areas such as large desert areas. It is estimatedthat from 1000 to 3000 Tg of such aerosols are emitted

408 KARAGULIAN AND ROSSI

annually into the atmosphere [7]. The term mineraldust collectively refers to aerosols of widely varyingcomposition that may include quartz, calcite, gypsum,hematite, and clays such as kaolinite, illite, and mont-morillonite. Although dust sources are unevenly dis-tributed across the earth, evidence that they have globaleffects comes from satellite data that provide opti-cal depth images of dust blowing for instance fromthe Saharan desert all the way to the southeasternUnited States [8] during spring and summer.

Mineral aerosol particles are characterized by a verylarge surface area that can strongly absorb the short-wave solar radiation and that may have an effect on theradiative forcing of the atmosphere [9], or may causea rate reduction in photolysis, thus inhibiting ozoneproduction. In fact, in the lower troposphere, mineralaerosol may strongly influence the balance of atmo-spheric trace gases, including ozone [9–11]. So far, themechanism for O3 destruction on mineral dust is un-clear. Recent work has shown that ozone loss can be dueto decomposition, catalytic destruction, or absorptionon mineral oxides [12].

Dentener and coworkers suggested [2] in theirmodeling studies that the ozone destruction on min-eral aerosol surfaces could lead to a 10% reduc-tion of O3 concentration in the dust source areas.For this study a reaction probability γ = 5.0 × 10−5

for O3 on mineral dust surfaces was assumed. Re-cently, Bauer and coworkers [13] have found a de-crease of 5% of the global tropospheric ozone massin a global modeling study employing an uptake coef-ficient γO3

= 1.0 × 10−5 as a best guess. On the otherhand, another modeling study which considered thecoupling of the photochemical and heterogeneous ef-fects of dust [14] led to a global ozone decrease of 0.7%assuming γO3

= 5.0 × 10−5.The present study deals with the kinetics of the het-

erogeneous reaction of O3 on mineral dust surrogatespresented as powders; its objective is to investigate themechanism of adsorption of ozone on surrogates ofmineral dust as well as the kinetics of the heterogeneousreaction including reaction products that are releasedinto the gas phase.

EXPERIMENTAL SETUP

All experiments were performed in a TEFLON® coatedKnudsen flow reactor operating in the molecular flowregime. This technique has been described in great de-tail in the literature [15]. Briefly, ozone was introducedinto the Knudsen reactor from the gas handling sys-tem using a fine needle valve as a flow control device.The gas leaves the Knudsen reactor through an escape

orifice whose diameter ranges from 1 to 14 mm and de-termines both the residence time τ g and the molecularconcentration of ozone [O3]. The rate constant for theeffusive loss kesc is given by the kinetic theory of gasesand was measured for every used orifice. The character-istic parameters and relevant kinetic expressions usedin this work are reported in Table I.

Steady-state experiments are performed by intro-ducing a constant flow F0 of O3 across a capillary intothe reactor. By analyzing the change of the mass spec-trometer (MS) signal levels of O3 upon opening andclosing the sample compartment, a value for the ob-served uptake coefficient γobs may be determined, pro-vided it is a first-order uptake process:

γobs(t) = kobs

ω= kesc

ω

(S0

S(t)− 1

)(E.1)

where ω and kesc are the gas-phase collision frequencyand escape rate constant, respectively, given in Table I.

S0 and S(t) represent the MS signal of O3 relatedto the gas flow entering and leaving the reactor withthe powdered sample either isolated or exposed, re-spectively. The gas residence time of a molecule in thereactor is defined as τ g = 1/kesc which increases withdecreasing size of the escape orifice. In the follow-ing, we define γobs either as γ0 or γss correspondingto the initial and steady-state uptake coefficients,respectively.

A series of uptake experiments were performed em-ploying a pulsed valve to admit O3 into the reactor.The pulsed-valve experiments were carried out by us-ing a solenoid valve, through which ozone was intro-duced in pulses with a duration of 5 ms and at a doseof 5 × 1015 molecules per pulse [15]. The “referencepulse” was fired when the sample was isolated fromthe gas. We determined the total number of injectedmolecules per pulse and the value of kesc, the latter ofwhich was obtained by fitting the decaying MS sig-nal to an exponential decay owing to in the absenceof reaction. The “reactive pulse” was obtained by re-peating the same operation with the plunger lifted. Thetotal observed exponential decay in the presence ofa reactive substrate was thus characterized by a newrate constant, kdec, given by kdec = keff + kesc that isused to determine keff and leads to the uptake coef-ficient γeff = keff/ω. The samples used in this study arethe following: kaolinite, poorly ordered (KGa-2, War-ren County, Georgia, USA), CaCO3 (Fluka), Arizonamedium test dust (Powder Technology Inc., BurnsvilleMN, USA), and Saharan dust (from the Cape VerdeIslands). In Table E.1, we report the composition of themain components of the analyzed mineral dust. Thetrue density ρt of all the examined powder samples

THE HETEROGENEOUS DECOMPOSITION OF OZONE 409

Table I Characteristic Parameters and Relevant Kinetic Expressions

Definition Value

Reactor volume, V 2000 cm3

Reactor surface area, Ar 1300 cm2

Sample surface area, As 19.5 cm2, (TEFLON holder), 4.9 cm2 (DELRIN holder)

Collision frequency ω for O3 with As 87.65 s−1 (TEFLON holder), 22 s−1 (DELRIN holder)

ω= 1.81(T/M)1/2 As

Typical concentration range 4–1000 × 1011 cm−3

N = F /Vkesc, F = flow into the reactor

Orifice diameters 1, 4, 8, and 14 mm

Rate constant for effusive loss (theoretical value), kesc = (8RT/πM)1/2 AH(1/4V )

AH = surface area of the orifice

kesc for O3 (experimental values at 300 K) 4.5 s−1 for 14 mm orifice

1.8 s−1 for 8 mm orifice



was taken from the literature, while the bulk den-sity ρb was determined from the weight and the vol-ume of the sample. The average particle diameter wasdetermined using SEM, and the Brunauer–Emmet–Teller (BET) surface area of every sample displayed inTable E.1 was measured using a Sorptomatic 1900Carlo Erba (Fison Instruments). Two kinds of sam-ple holders were used: one consisted of a TEFLON®

coated Pyrex holder of 19.6 cm2 of available samplesurface, the other consisted of an internal reductionpiece made out of DELRIN® leading to a sample sur-face of 4.9 cm2. Both TEFLON® and DELRIN® didnot show any reactivity with ozone under the presentexperimental conditions.

O3 was prepared in an ozone generator (Fischer 502)by means of a corona discharge using a flow of pureO2 at a pressure of 400 mbar. Subsequently, O3 wascondensed on silica gel cooled to 185 K and then letto desorb from silica gel into a darkened storage ves-sel. The concentration of ozone in the O2/O3 samplewas determined by UV absorption spectroscopy [3].A glass absorption cell of 7-cm optical path lengthand equipped with quartz windows was used for themeasurement of O3 absorption at 256.3 nm using across section σ= 1.15 × 10−17 cm2 for the calculationof [O3] at the measured total pressure of 2 mbar [16].According to the UV absorption, we estimated that theO2 impurity amounted to 15–28% of the total pressure.

Mass m/e 32 was both a marker for the potentialreaction product O2 as well as an important fragmentof O3. In addition, one has to consider that the MSsignal S32(t) at m/e 32 will contain the contributionsdue to the O2 impurity present in the O3 sample thatamounts to 15–28% of the total pressure according toUV absorption, and due to the small O2 backgroundalready present in the flow reactor.

Ozone has a measurable fragment and a parent peakat m/e 32 (O+

2 ) and m/e 48 (O+3 ), respectively. In

the absence of the sample, one has to consider thecontributions to the MS signal S0

32 at m/e 32 due tothe O2 impurity Simp

32 in the O3 sample and to theO2 background Sback

32 in the flow reactor. We defineSimp

32 = F impO2

/CO2, where F imp

O2is the O2 flow relative

to the total measured flow F totO2 + O3

and is expressed asF imp

O2= F tot

O2 + O3· x/100 with x(%) being the percent-

age O2 impurity in the total measured flow F totO2 + O3

.The calibration factor CO2

has been determined in sep-arate experiments where the MS signal intensity at m/e32 was measured as a function of the admitted pure O2

flow. The percentage x(%) of O2 in O3 is measured inseparate calibration experiments using UV absorptionas discussed above.

Before exposing the substrate to ozone the correctedMS signal at m/e 32, S0

32, is given by Eq. (E.2):

S032 = S0

32 − Sback32 − Simp

32 (E.2)

Upon lifting the plunger the flow FO2(t) of oxygen pro-

duced by ozone decomposition on the examined sub-strate has been calculated using Eq. (E.3):

FO2(t) = (

Sr32(t) − Sback

32 − Simp32 − R

3248 S48(t)

) · CO2

(E.3)where Sr

32(t) is the raw MS signal at m/e 32 during O3

uptake, R3248 = S0

32/S048 = 1.7 ± 0.5 represents the ratio

between the corrected MS signal at m/e 32 and 48 be-fore the exposure of the sample to ozone, S48(t) is theMS signal at m/e 48 during the uptake experiment andCO2

is the calibration factor for oxygen. The term withinthe bracket of Eq. (E.3) corresponds to the net O2 MSsignal S32 at m/e 32 resulting from heterogeneous O3


decomposition. Equation (E.3) assumes that molecu-lar oxygen does not react with the sample, a fact thathas been established is separate reference experiments.Therefore, �FO3

(t) = F inO3

(t) − FoutO3

(t) represents theflow or the rate of O3 lost during the uptake experi-ment that was calculated by calibrating the resultingMS signal at m/e 48.

The relative O2 yield per O3 molecule destroyed onthe substrate is given by the ratio r (t) = FO2

(t)/�FO3(t)

as a function of time during which the surface is ex-posed to O3. It is the relative yield of O2 generated.

RESULTS AND DISCUSSION

O3 Reaction on Poorly Ordered Kaolinite

The uptake of ozone on kaolinite samples taken as asurrogate for mineral dust monitored at m/e 48 wasmeasured at room temperature (RT) with the goal toobtain a quantitative measure of the reaction kineticsas well as the reaction products. The measurementswere performed at ozone concentrations in the rangefrom 4.0 × 1011 to 2.4 × 1012 cm−3.

Figure 1 shows a representative uptake experimentof ozone on 0.2 g of kaolinite spread out on a surfaceAs of 4.9 cm2. Curves (a) and (b) correspond to theraw MS signal monitored at 32 (O+

2 ) and m/e 48 (O+3 ),

respectively. A constant flow of O3 that was isolatedfrom the sample by the isolation plunger was initiallyestablished. When the flow of O3 reached a constantlevel, the isolation plunger was lifted and the initialuptake coefficient γ0 of ozone on the substrate was ob-tained. A decrease of the uptake of ozone with exposuretime was observed until steady-state uptake is reached

Figure 1 Representative Knudsen-Cell experiment for O3 uptake on 0.2 g sample of kaolinite. Curves (a) and (b) correspond

to raw MS signals monitored at m/e 32 and 48, respectively ([O3] = (2.7 ± 0.7) × 1012 cm−3; 14 mm orifice; surface sample

area As = 4.9 cm2).

at t ≥ 800 s. This steady-state level is presumably con-trolled by the competition between the rate of adsorp-tion, heterogeneous reaction, and desorption of ozoneor its decomposition products.

Uptake experiments of O3 on kaolinite powder weresystematically carried out by varying the initial con-centration of O3 and its residence time τ g. For agiven orifice size and thus τ g, we observed that kss

was independent of [O3] within experimental uncer-tainty. As a case in point, we present data for the14 mm orifice (τ g = 0.2 s) and for [O3] ranging between(4.0 ± 0.5) × 1011 and (2.4 ± 0.7) × 1012 cm−3, lead-ing to kss = (0.203 ± 0.033) s−1, and a steady-state up-take coefficient γss = (9.0 ± 1.5) × 10−3 (dashed linein Fig. 2, Table E.2) based on the geometric surfacearea of the sample support.

From this series of measurements, it is evident thatthe decomposition reaction of ozone on kaolinite maybe described by a pseudo-first-order rate constant kss.However, if we decrease the orifice size, thus increaseτ g at constant [O3] = (2.4 ± 0.7) × 1012 cm−3 the val-ues of kss decrease as displayed in Fig. 3 and Table E.3.The strong dependence of kss on τ g at a given [O3] sug-gests that the mechanism of ozone uptake is complexand does not correspond to a simple first-order rate lawfor uptake. These observations indicate that the reactiv-ity of O3 on kaolinite decreases for long residence timesτ g as the heterogeneous reaction rate not only dependson the gas-phase concentration but apparently also onintermediates whose surface concentration depends onthe extent of reaction that scales with τ g. Therefore, theconclusion that the rate law is first order in O3 basedon the results of Fig. 2 may be erroneous because it isbased on the apparent independence of kss on [O3] overa narrow concentration range.


Figure 2 Plot of the pseudo-first-order rate constant kss for steady-state uptake of O3 on 0.2 g of kaolinite versus O3 concentration

at ambient temperature (14 mm orifice; surface sample area As = 4.9 cm2).

In order to confirm this hypothesis, further experi-ments were carried out on samples of kaolinite by ex-posing them to an initial partial pressure of 4 Torr ofO3 during 16 h in a separate static vessel. In the fol-lowing, we used the 8 and 14 mm orifices in order toprobe the heterogeneous uptake of O3 and obtainedkss = (4.3 ± 3.5) × 10−2and (4.8 ± 3.2) × 10−2 s−1 forthe 8 and 14 mm orifices, respectively. A similar value,namely kss = (5.6 ± 3.0) × 10−2 s−1, was already foundfor the 4 mm orifice without preliminary exposure toO3 as shown in Fig. 3. It seems that a long exposureof the sample to O3 deactivates a large fraction of re-active surface sites and leads to values of γ that areup to a factor of five smaller than γss measured atlow τ g (Fig. 3). However, a constant “baseline” reac-tivity persists at steady-state independent of τ g, afterO3 pretreatment. This “baseline” or residual rate con-stant for O3 decomposition is thus independent of [O3]following the results displayed in Fig. 3 as well as

Figure 3 Plot of kss of O3 on 0.2 g of kaolinite versus orifice size (full circles) at surface sample area As = 4.9 cm2;

([O3] = (2.0 ± 0.5) × 1012 cm−3). Open triangles correspond to samples that have been exposed to 4 mbar of O3 for 16 h

offline.

of τ g used to probe the O3/kaolinite interaction (opentriangles in Fig. 3, Table E.3). Theoretically, the valueof r (t) should tend toward 1.5 for the decompositionof O3 according to the following pseudo-elementaryreactions:

O3 + SS → O(SS) + O2 (1)

O3 + O(SS) → O2 + O2(SS) (2)

where SS and O(SS) are reactive surface sites for O3

and the atomic oxygen intermediate after O3 decompo-sition, respectively. O2(SS) indicates an adsorbed per-oxy species formed in reaction (2) through oxidationof the atomic oxygen intermediate O(SS) by O3. Thepresent measurements do not allow the positive iden-tification of the formation of O(SS) or O2(SS). How-ever, spectroscopic studies on the decay of O3 on amanganese oxide substrate by Li and Oyama [17,18]


showed evidence for the formation of a peroxy speciesakin to O2(SS) on the substrate surface. The final step inthe ozone decomposition reaction may be the thermaldecomposition of the peroxide species to form gaseousO2 according to

O2(SS) → O2 + SS (3)

Reaction (1) is much faster than reactions (2) and(3), and does not involve the most-abundant reactionintermediate, the peroxide species O2(SS). Therefore,reactions (2) and (3) will be rate limiting and thusmeasurable. Reactions (1) and (2) are presumed to beirreversible because we have never detected O3 in des-orption using mass spectrometry.

The net is the sum of reactions (1), (2), and (3) anddescribes the catalytic decomposition of O3 on mineraldust:

2O3 → 3O2 (4)

This mechanism may explain why transition metal ox-ides, such as manganese oxide, are good catalysts forozone decomposition.

In the O3 uptake experiment reported in Fig. 4, r (t)was always significantly less than 1.5 so that catalyticdecomposition of O3 may be ruled out. At steady-stater (t) ranges from 1.0 for a short residence time τ g = 0.2 sfor O3 (curve (a), Fig. 4), to essentially 0.2 for largeresidence time τ g = 2.0 s (curve (c), Fig. 4). For theO3-pretreated kaolinite discussed above, the ratio r (t)hovers at small values consistent with r (t) < 0.5 duringthe whole duration of the uptake (curve (d), Fig. 4).

The full mechanism is not as simple as just shown inreactions (1), (2), (3), and (4). There is evidence fromthe inverse dependence of the kinetics on the residence

Figure 4 The relative yield r (t) during the uptake experiment of O3 on 0.2 g of kaolinite. Curves (a), (b), and (c) represent

plots of r (t) at 14 mm (τg = 0.2 s), 8 mm (τg = 0.5 s), and 4 mm (τg = 2.0 s) orifices, respectively. Curve (d) represents a plot

of r (t) at 1 mm for a saturated sample exposed to 4 Torr of O3 during 16 h. ([O3] = (2.0 ± 0.5) × 1012 cm−3, As = 4.9 cm2).

time τ g (displayed in Fig. 3) that suggests that O3 mayalso bind to the substrate without O2 release. In orderto be able to describe this large variation in r (t), thepresent results suggest a mechanism consisting of fourreactions for the heterogeneous interaction of O3 withkaolinite. Initially, O3 collides with a reactive surfacesite SS on the surface and deposits an O atom (reaction(1)). This reaction initiates the heterogeneous decom-position of O3 on chemically reactive surface sites (SS)on the substrate and plays an important role at short res-idence time τ g of O3 leading to r (t) = 1.0 (curve (a),Fig. 4). Two other types of reactions may be proposedwhere O3directly reacts with the surface to form a com-plex without O2 formation according to reactions (5)and (6):

O3 + SS → adduct (5)

O3 + O(SS) → adduct (6)

These reactions are introduced in order to explain thedisappearance of O3 without the formation of gas phaseO2. Without additional and detailed knowledge of thesurface, the nature and molecular structure of the com-plex resulting from reaction (5) remains obscure.

The larger rate of O3 disappearance compared to O2

formation is expected to hold for short residence timesof O3 at the beginning of the interaction between O3 andkaolinite just after having lifted the plunger and leadsto r (t) = 1.0. The combination of reactions (1) and (5)or (6) enables values of r ≤ 1.0 as all measured valuesof r (t) decrease from 1.0 to a low value of r (t) ∼ 0.2at large values of τ g and large extents of reaction asshown in Fig. 4.

Considering the data displayed in Fig. 4, it appearsin conclusion that adduct formation, reaction (5) or(6), is favored over reaction (1) at increasing extents


Figure 5 Dependence of the uptake coefficient γ0 (open circles) on sample mass at [O3] = (4.5 ± 1.0) × 1011 cm−3 at an

orifice diameter of 14 mm for the uptake of O3 on kaolinite powder (As = 19.6 cm2). Full circles display γeff obtained in pulsed

valve experiments carried out under the same experimental conditions.

of reaction at constant [O3]. At the beginning of re-action, r (t) is close to one and drops to 0.5 at longerreaction times which corresponds to the increasing im-portance of reaction (6) compared to reaction (1) (curve(a), Fig. 4). However, the fact that r (t) drops below 0.5as displayed in Fig. 4 (curves (c) and (d)) points tothe fact that reaction (5) is also occurring at the begin-ning of the O3 uptake experiment. If reaction (5) wouldexclusively be occurring, r (t) = 0 would be obtained.

In order to establish whether the effective availablesurface area is influenced by the internal surfacearea formed by interstitial voids between individualdust particles, the mass dependence of the O3 up-take on kaolinite was measured in the Knudsen flowreactor at ambient temperature and at [O3] = (4.5±1.0) × 1011 cm−3. The mass of kaolinite ranged from0.3 to 4 g, and the results are shown in Figs. 5and 6.

Figure 6 Dependence of the steady-state uptake coefficient γss on sample mass at [O3] = (4.5 ± 1.0) × 1011 cm−3 at an orifice

diameter of 14 mm for the uptake of O3 on kaolinite powder (As = 19.6 cm2). The numerical fit of the data using the pore

diffusion model is represented by dashed lines.

The steady-state and initial uptake coefficient γss

and γ0, respectively, of O3 were found to linearly in-crease at low masses of kaolinite as displayed in Figs. 5and 6 for γ0. Samples below 0.4 g may be considered tobe part of this linear mass-dependent regime. Increas-ing the sample mass further had a negligible effect onthe amount of O3 adsorbed because not the entire sam-ple surface is apparently available for O3 adsorption.This maximum value is attributed to the inability of O3

to penetrate through several layers of the sample duringthe residence time of O3 in the gas phase, thus resultingin a constant number of molecules taken up despite theincreasing sample mass. The fact that we observe an“apparent” mass dependence of γ0 from 0 to 0.4 g isinterpreted by the fact that we are “filling” a coherentsample layer.

In order to better define the saturation behavior ofO3 on the substrate, a series of uptake experiments


were performed employing a pulsed valve to admit O3

into the reactor. The increase of γeff with the samplesaturates at large sample mass and stays constant at(8.0 ± 0.7) × 10−2 as displayed by the full circles inFig. 5. The same behavior is observed for γ0 whichstays constant at (9.0 ± 0.7) × 10−2 at large samplemasses. As shown in Fig. 5, γeff and γ0 are identi-cal within experimental uncertainty for all performedexperiments. The results displayed in Figs. 5 and 6have been plotted as a function of the mass and of thenumber of layers of kaolinite. In order to determine thenumber of layers, the total volume of kaolinite powderwas calculated from its true density ρt = 2.3 g/cm3 andthe mass of the sample spread out across the geometricarea of the sample holder. From the average particle sizeand the thickness of the sample, we obtain the numberof layers. A typical grain diameter of 1.0 �m is givenaccording to the manufacturer’s undocumented spec-ifications for the used kaolinite powder (Table E.1).However, most mineral dust powders are porous ma-terials and the microstructure of the dust substrate iscomposed of clusters of random distribution with in-terstitial voids between them. Therefore, it is more rea-sonable to take into account a grain size diameter thatis much larger than 1.0 �m as suggested by electronmicroscopy (SEM) of similar material where character-istic grain diameters are in the tens of micrometers ( c©OMNI Laboratories, Inc, http://www.omnilabs.com/).

The plot in Fig. 5 shows that a mass of 0.4 g corre-sponds to one nominal layer of approximately 90 �mdiameter particles of kaolinite spread out over the geo-metric surface of the sample holder (19.6 cm2). Thus,one nominal layer of kaolinite will contain closelypacked spheres of “effective” grain diameter of 90 �m.Therefore, the linear mass-dependent portion of γss vs.mass in the 0–0.4 g range corresponds to a sampleholder partially covered with 90-�m-diameter kaoli-nite particles which is the structural model we adoptin this work. In a recent study on the reaction of NO3

on mineral dust [19], we have obtained similar resultsfor the mass dependence of γ including an “effective”grain diameter of 50 �m for kaolinite powder.

Figure 6 displays steady-state uptake experimentsperformed on the same amount of kaolinite as in theexperiments reported in Fig. 5. The trend of these mea-surements is identical to that observed in pulsed valveexperiments. Beyond a mass of 0.4 g, the steady-statevalue of γss stabilizes at (1.6 ± 0.3) × 10−2, a factor at5 less than γ0. The use of the BET surface area andthe application of the pore diffusion theory [20] yieldsγpd,ss = (2.7 ± 0.3) × 10−6. The curve in Fig. 6 has beenfitted to the observed steady-state uptake coefficient γss

as a function of sample mass using a grain diameter forkaolinite of 1 �m, a surface sample area As = 19.6 cm2,

as fitting parameters at a fixed orifice diameter of 14 mmand [O3] = (4.5 ± 1.0) × 1011 cm−3.

The use of the BET surface area and the pore diffu-sion theory substantially underestimate the true uptakecoefficient for kaolinite approximately by a factor 103.Therefore, we propose to regard this result as a lowerlimit to the true value of γss. We want to point out thatthe results obtained from the pulsed valve experiments(γeff) are virtually identical to the initial value of γ0

after the start of the uptake reaction. These results re-flect the fact that O3 explores the external surface of thekaolinite sample at t = 0 and that it is improbable for itto explore the BET surface area of the sample duringa time period representative of a typical pulse decay,typically a fraction of a second.

An additional series of experiments concerns thethermal cycling behavior of kaolinite that had previ-ously been exposed to several Torr of O3 in order to dis-cover stable reaction products that desorb upon heatingof the poisoned substrate. For a 1.5 g kaolinite sample,kss = 6.3 × 10−2 s−1 corresponding to γss = 1.4 × 10−2

was measured in the 4-mm-diameter flow reactor afterpumping the sample for 30 min. After exposing kaolin-ite to 8 Torr of O3 for 12 h, γss decreased to 9.4 × 10−4

in agreement with results discussed in Fig. 3. Afterheating the exposed sample to 80◦C and letting it cooldown to ambient temperature, γss increased by roughlyan order of magnitude to 8.8 × 10−3. During the ther-mal treatment, only desorption of H2O but no O2 wasobserved. This rejuvenation of the kaolinite sampleis consistent with the thermal decomposition of thesurface peroxide species observed by Li and Oyama[17,18] according to reaction (3) which implies an in-crease of the number of surface sites SS for adsorptionof O3 upon thermal decomposition of O2(SS). How-ever, this regeneration of SS only affects the value of γss

and leaves γ0 unchanged in contrast to the O3 exposurewhich also decreases γ0 for O3 uptake. We certainly ex-pected to find O2 among the desorbing products but didnot observe significant amounts, probably owing to therelatively large background of O2 in the flow reactor.This type of desorption experiments will be performedin the future in a suitably modified Knudsen flow reac-tor in conjunction with surface-sensitive spectroscopicmethods that will contribute toward the understandingof the underlying molecular mechanisms.

Reaction of O3 on CaCO3 and NaturalLimestone

We have performed uptake experiments of O3 onCaCO3 using a sample of 1 g spread out on a sur-face area As of 4.9 cm2 and a residence time τ g = 0.2 s


Table II Summary of the Reactive Uptake Coefficients for Various Mineral Dust Surrogates Obtained in the PresentWork Based on the Geometric Surface Area

Sample Orifice Size (mm) [O3] (cm−3) γ0 γss

Kaolinite 14 (2.4 ± 0.7) × 1012 (6.3 ± 0.2) × 10−2 (1.0 ± 0.2) × 10−2

CaCO3 14 (5.3 ± 0.7) × 1012 (1.2 ± 0.3) × 10−2 (3.6 ± 0.2) × 10−3

CaCO3 (marble) 1 (2.3 ± 0.4) × 1012 (2.3 ± 0.4) × 10−4 (3.5 ± 1.6) × 10−5

Natural limestone 4 (3.0 ± 0.7) × 1012 (1.3 ± 0.2) × 10−2 (1.6 ± 0.5) × 10−3

Saharan dust 4 (3.5 ± 0.7) × 1012 (9.3 ± 2.6) × 10−2 (6.7 ± 1.3) × 10−3

Arizona test dust 4 (3.3 ± 0.7) × 1012 (1.3 ± 0.6) × 10−2 (2.2 ± 1.2) × 10−3

Natural limestone 4 (2.0 ± 1.0) × 1013 (2.1 ± 0.3) × 10−3 (2.4 ± 0.7) × 10−4

Saharan dust 4 (1.0 ± 0.4) × 1013 (3.7 ± 1.8) × 10−2 (3.3 ± 2.5) × 10−3

Arizona test dust 4 (8.0 ± 1.5) × 1012 (1.3 ± 0.7) × 10−2 (2.5 ± 1.2) × 10−3

(orifice diameter of 14 mm) corresponding to an orificeof 14 mm. At [O3] = (5.3 ± 0.5) × 1012cm−3, the ratiobetween the O2 product yield and the total O3 loss re-sults in r (t) = 1.4 and in γss = (3.6 ± 1.8) × 10−3 andγ0 = (1.2 ± 0.3) × 10−2 (Table II and Fig. 7, curve (a)).At the beginning of the reaction, we have obtainedr (t) ≤ 1.0 and at steady-state conditions r (t) tends to-ward 1.5 which corresponds to heterogeneous decom-position using the substrate as a catalyst. An importantpoint is the “inverse” behavior of r (t) of O3 on CaCO3

with the extent of reaction (curve (a), Fig. 7) as op-posed to the decrease of r (t) with reaction time forO3 on kaolinite (curves (a)–(c), Fig. 4). For compari-son purposes, we have carried out a systematic studyof γss as a function of sample mass in order to de-termine a corrected uptake coefficient γpd of O3 onCaCO3 powder using the pore diffusion model [20]. Wehave used a large sample surface area of As = 19.6 cm2

and an expanded range of sample masses. At steady-state γss,pd = (7.8 ± 0.7) × 10−7 with a tortuosity factorτ = 2.0 was found for [O3] = (5.3 ± 0.7) × 1012 cm−3.

Figure 7 The relative yield r (t) during the uptake experiment of O3 on 1 g of CaCO3 (curve (a)), 0.3 g of Saharan dust (curve

(b), 1 g of natural limestone (curve (c), and 0.3 g of Arizona medium test dust (curve (d)). The measurements were performed

at [O3] = (3.5 ± 0.5) × 1012 cm−3, As = 4.9 cm2 and 4 mm orifice (τg = 2.0 s).

The fit to the measured values of γss resulted in a valueof γpd that is smaller by a factor of 4.6 × 103 with re-spect to the measured value γss reported above.

In order to assess whether the pore diffusion modelmay yield a lower limit for γ, we have performeda few uptake experiments on rough marble sample(solid crystalline CaCO3). In this case, we do notexpect interstitial diffusion to occur and the pro-jected surface area of the sample that is relevantfor uptake is the geometrical surface area of 19.6cm2. For [O3] = (2.0 ± 1.2) × 1012 cm−3 and a res-idence time τ g = 28 s (orifice diameter of 1 mm),we found γ0,marble = (2.3 ± 0.4) × 10−4 and γss,marble =(3.5 ± 1.6) × 10−5 as displayed in Fig. 8.

The uptake values γ0 and γss calculated on the basisof the geometric surface area AS of the powder sampleare overestimated by a factor of 50 and 100, respec-tively, compared to the uptake values calculated fromthe reaction of O3 on marble sample. On the other hand,the uptake value γss,pd calculated applying the pore dif-fusion theory (KLM) is underestimated by a factor of


Figure 8 Knudsen-Cell experiment for O3 uptake on a 2 g marble sample of CaCO3 whose surface has been roughened up.

The curve shows the raw MS signal monitored at m/e 48 ([O3] = (2.0 ± 1.2) × 1012 cm−3; orifice diameter of 1 mm; surface

sample area As = 19.6 cm2).

50 compared to γ0 and γss calculated on the basis ofthe geometric surface area.

In conclusion, for O3 reacting with CaCO3 pow-der the use of pore diffusion theory (KLM) as wellthe geometric surface area As are low and high limits,respectively, in the estimation of the true uptakecoefficient.

Two series of experiments were performed on 1 g ofnatural limestone composed of CaCO3 (97%) that con-tained a small percentage of metal oxides, especiallyiron and aluminum oxides (see Table E.1). As reportedin Table II, the values of γ0 and γss measured at low[O3] were larger by a factor of 6 with respect to those athigh [O3] which is a general trend observed for manymineral dust materials (see Table II).

From these measurements it is evident that both γ0

and γss do not follow a rate law pseudo-first order inO3 and suggests that the mechanism of O3 uptake maybe complex. Interestingly, on this substrate the ratior (t) for ozone uptake and decomposition remains be-low 0.5 even at a high concentration of ozone and longresidence time (Fig. 7, curve (c)) compared to CaCO3

(Fig. 7, curve (a)). This kind of sample seems to bean excellent absorber of O3 without extensive decom-position at ambient temperature. Despite the high per-centage of CaCO3 in natural limestone, the presence ofsmall amounts of metal oxides such as Fe2O3 in excessCaCO3 could be responsible for the suppression of O2

formation.When we compare γ for CaCO3 and natural

limestone at roughly the same ozone concentration([O3] = (4.0 ± 1.0) × 1012 cm−3), we find that the val-ues for γ0 are the same whereas γss is a factor of 2.5larger for CaCO3 according to Table II, presumablybecause of a faster saturation process in natural lime-stone compared to pure CaCO3. However, a complete

understanding of the surface reactivity is only possi-ble if the surface rather than the bulk composition isconsidered.

Reaction of O3 on Saharan Dust andArizona Medium Test Dust

Two series of experiments were performed on 0.3 g ofSaharan dust using two different ozone concentrations.As reported in Table II, γ0 and γss measured at low[O3] were larger by a factor of 2.5 and 2, respectively,compared to the values at high [O3].

As already observed for natural limestone, γ0 andγss do not follow a rate law pseudo-first order in O3

and suggests that the mechanism of O3 uptake is com-plex. A similar dependence has been observed beforeby Hanisch and Crowley [12] and Sullivan and cowork-ers [21] in their work on ozone decomposition onSaharan dust and on fresh alumina films. The reasonfor this behavior may be related to the finite numberof available surface sites of the substrate that are notcompletely saturated at low [O3] resulting therefore ina larger value of γ compared to high [O3].

For steady-state uptake performed at [O3] =(3.5 ± 0.5) × 1012 cm−3 the ratio r (t) has been foundto be approximately 0.8 (Fig. 7, curve (b)), whereas for[O3] = (1.0 ± 0.4) × 1013 cm−3 it stabilizes around 1.0.

On 0.3 g Arizona medium test dust, we found iden-tical values of γ0 and γss for both values of [O3] asreported in Table II. The ratio r (t) at steady-state con-ditions tends toward a value of approximately 0.4 forboth ozone concentrations (Fig. 7, curve (d)) indicatinga significant reactivity for O3 uptake on this substrateand lack of ozone decomposition leading to O2 forma-tion. Both Arizona test dust and natural limestone resultin a similar r (t) dependence for ozone decomposition,


Table III Typical Initial and Steady-State Reactive Uptake Coefficients for Various Mineral Dust Surrogates Obtainedin Previous Work

Sample γ0,BET γss,BET γpd,ss γpd,ss (This Work)

Kaolinite (3.0 ± 1.0) × 10−5 [22] – – (2.7 ± 0.3) × 10−6

Saharan sand (6.0 ± 2.0) × 10−5 [22] 6.0 × 10−6 [22]

2.3 × 10−7 [23] –

Saharan dust – – (2.2+1.4−1.2) × 10−6 [12]

CaCO3 4.3 × 10−7 [23] – – (7.8 ± 0.7) × 10−7

distinctly different from results for Saharan dust.Saharan dust contains a significant percentage of metaloxides, with r (t) tending toward values between 0.8and 1.0 at steady-state conditions. If we analyze thechemical composition of the mineral dust examined inthis work more carefully, we note that iron oxides suchas Fe2O3, FeO, and Fe3O4 are present in different per-centages: Saharan dust contains more iron than naturallimestone and Arizona test dust, a fact that may be res-ponsible for the difference in γ and r (t) on the differ-ent substrates. Specifically, we note that r (t) increaseswith the amount of iron oxides in the sample whichleads to dominant O3 decomposition resulting in O2

in contrast to O3 uptake. This aspect is supported byrecent work of Michel et al. who found that the initialreactivity of oxide powders toward ozone followed theorder �-Fe2O3 > �-Al2O3 > SiO2 > kaolinite with theSaharan sand and China loess having lower reac-tivities more similar to SiO2 and kaolinite [22]. Inthis investigation, the initial reactive uptake coeffi-cient γ0 of O3 on several mineral oxide powders wasmeasured using a Knudsen reactor leading to γ0 =(2.0 ± 0.3) × 10−4 for �-Fe2O3, (1.2 ± 0.4) × 10−4

for �-Al2O3, (6.3 ± 0.9) × 10−5 for SiO2, and(3 ± 1) × 10−5 for kaolinite.The greater reactivity ofthe iron oxides is also in agreement with the obser-vations by Suzuki who carried out a laboratory studyof O3 reactivity on various mineral oxides [6]. Usinga UV absorption monitor, the authors report the rel-ative O3 reactivity of silica, �-Fe2O3, Fe3O4, and �-Al2O3 in comparison with natural sea sand collected inJapan, which was further separated into an “iron sand”and a “remainder sand” component. The “iron sand”component had a reactivity similar to Fe3O4, a ma-jor component in the “iron sand,” that decomposed O3

at a faster rate than the natural sand and the “remain-der sand.” This suggests that iron oxide such as Fe3O4

more effectively destroys ozone than the other solidphases present in the sand. It was concluded that iron-containing compounds had a superior ozone decompo-sition rate than alumina and silica model compoundsin agreement with the present kinetic results.

Comparison with Literature Values

Previous results on the uptake coefficients for O3 up-take on mineral dust obtained by other authors are dis-played in Table III for comparison purposes. O3 uptakeon Saharan sand and calcite has been investigated us-ing a fluidized-bed reactor [23]. Although uptake coef-ficients have not been derived, they may be estimatedfrom the data presented employing the BET surfacearea of the substrate which will lead to a lower limitingvalue. For Saharan sand, the signal reduction obtainedin that work corresponds to an initial uptake coefficientof γ0,BET = 2.3 × 10−7 at [O3] = 2.5 × 1012 cm−3. In asimilar manner, the initial uptake coefficient for calciteresulted in a value of γ0,BET = 4.3 × 10−7.

In recent work performed by Hanisch and Crowley[12] the heterogeneous reaction of O3 on Saharandust has been investigated at ambient temperature(296 K). The conversion efficiencies r for O2 formedpresented in that work were 1.0 and 1.3 mol of O2

per O3 destroyed for unheated and heated samples(T = 450 K), respectively. In the same work [12],it was shown that γ for the irreversible destructionof O3 on Saharan dust surfaces depended on theO3 concentration leading to �

pd0 = (5.5+ 4.0

−3.0 ) × 10−5

and �pdss = (2.2+1.4

−1.2) × 10−6 using the pore-diffusionmodel at [O3] = (8.4 ± 3.4) × 1012 cm−3. Experimentsat [O3] = 6.0 × 1012 cm−3 using Arizona test dust andkaolinite [24] were performed as well. Only measuredvalues of γ0 were reported; for unheated kaoliniteγ0 = 1.0 × 10−4 which was close to or below the de-tection limit, whereas γ0 = 2.0 × 10−3 was found forArizona test dust.

In two other recent studies that employed aKnudsen reactor for O3 uptake on ground Saharansand at [O3] = 1.9 × 1011 cm−3 Michel and coworkers[22,25] obtained γ0,BET = (6.0 ± 2.0) × 10−5 whereasat steady-state conditions γss,BET = 6.0 × 10−6 wasfound. In comparison, we have found γpd,ss = 2.4 ×10−6, which is lower by a factor of 2.5 (Table III).

The uptake coefficient, γBET, calculated from thepresent values of γss using the BET surface area


resulted in γss,BET = (2.0 ± 0.6) × 10−6 for kaoliniteat [O3] = (2.4 ± 0.7) × 1011cm−3, whereas γss,BET =(8.3 ± 0.4) × 10−7 for CaCO3 at [O3] = (5.3 ± 0.7) ×1012 cm−3. These values are identical to those foundfrom the pore diffusion model within experimental un-certainty. However, we would like to reiterate that theseγ values are likely lower limits to the true values.Finally, measurements of O3 deposition velocities(�dep) have been performed on various substrates in-cluding kaolinite sand and CaCO3 [26]. Using thesedata and the expression γ = 4�dep/c, Dentener et al. [2]estimated γ values for O3 in the range 10−4–10−5.

CONCLUSIONS

The present data indicate that the reactivity of O3 onkaolinite decreases with residence time τ g and that theheterogeneous reaction rate not only depends on thegas-phase concentration of ozone but apparently alsoon intermediates whose surface concentration dependson the extent of reaction, i.e on τ g.

Despite the difficulty to detect O2 as a productspecies from O3 decomposition, we report the relativeproduct formation rate as a function of time resultingfrom the decomposition of O3 on the different sub-strates by monitoring the ratio r (t) = FO2

(t)/�FO3(t).

There is no simple common mechanism that may ex-plain ozone decomposition on each type of mineraldust substrate which means that no single surrogatematerial cited above can be representative of “mineraldust” as far as the heterogeneous reaction with ozoneis concerned. We claim that neither γ0 nor γss is a goodindicator for O3 reactivity that is able to distinguish be-tween the different substrates because the resulting γvalues are all very similar for all the examined sub-strates within a range of from (1.2 ± 0.3) × 10−2 to(9.3 ± 2.6) × 10−2 for γ0 and from (1.6 ± 0.5) × 10−3

to (1.0 ± 0.2) × 10−2 for γss. The use of values derivedfrom the pore diffusion theory [20] and from the BETsurface area may be interpreted as a lower limit for O3

uptake that perhaps severely underestimates the inter-action of O3 with mineral dust.

The uptake values derived from the pore diffusiontheory found for kaolinite and CaCO3 are lower by afactor 3 × 103 with respect to those calculated usingthe geometric surface area of the sample. We assumethat Arizona test dust, Saharan dust, and natural lime-stone will follow the same trend. In addition, the uptakeexperiment performed on crystalline CaCO3 (marble)suggests that the pore diffusion correction of the mea-sured γ value will lead to an underestimation of thetrue value by a factor of 50–100. Therefore, we mayconclude that for all mineral dust samples used in this

work the uptake value γ is of the order of 10−5. On theother hand, r (t) seems to be a better indicator for thereaction of these surrogates with O3 because it discrim-inates the different samples to a larger extent comparedto the absolute values of γ0 or γss of the various typesof substrates.

In order to estimate the effect of ozone uptake onmineral dust we used field measurements of a dust eventreported in recent work [27]. At an altitude of 4 km, thetotal aerosol surface area was 1.5 × 10−6 cm2 cm−3 and[O3] was 31 ppb (5.0 × 1011 cm−3) at 279 K. If we takea γ value of 3.5 × 10−5 as lower limit for the uptakeof O3 on mineral dust, the lifetime of O3 is 10 days.This lifetime quite closer to that one found in the recentwork of Hanisch and Crowley [12] and Sullivan et al.[21], who used a lower limit for γ of about 10−5 andestimated a lifetime ranging between 33 and 55 daysfor the same aerosol surface area.

In conclusion, the present experimental showed thelow reactivity of O3 on mineral dust substrate. Theuptake values proposed for mineral dust and supportedby experiments on crystalline CaCO3 resulted in a valueof 10−5. This value is larger by a factor of 10 from thoseone found by Hanisch and Crowley [12] and Michel andcoworkers [22], who reported uptake coefficient of theorder of 10−6 using the pore diffusion theory [20].

BIBLIOGRAPHY

1. Bonasoni, P.; Cristofanelli, P.; Calzolari, F.; Bonafe, U.;

Evangelisti, F.; Stohl, A.; Sajani, S. Z.; van Dingenen,

R.; Colombo, T.; Balkanski, Y. Atmos Chem Phys 2004,

4, 1201–1215.

2. Dentener, F. J.; Carmichael, G. R.; Zhang, Y.; Lelieveld,

J.; Crutzen, P. J. J. Geophys Res-Atmos 1996, 101,

22869–22889.

3. Horvath, M.; Bilitzky, L.; Huttner, J. Ozone; Elsevier:

New York, 1985; pp. 26–29.

4. Jacob, D. J. Atmos Environ 2000, 34, 2131–2159.

5. Kamm, S.; Mohler, O.; Naumann, K. H.; Saathoff, H.;

Schurath, U. Atmos Environ 1999, 33, 4651–4661.

6. Suzuki, S.; Hori, Y.; Koga, O. Bull Chem Soc Japan

1979, 52, 3103–3104.

7. Jonas, P.; Charlson, R.; Rodhe, H. Aerosols in Climate

Change; Cambridge University Press, 1995.

8. Schulz, M.; Balkanski, Y. J.; Guelle, W.; Dulac, F.

J Geophys Res-Atmos 1998, 103, 10579–10592.

9. Dickerson, R. R.; Kondragunta, S.; Stenchikov, G.;

Civerolo, K. L.; Doddridge, B. G.; Holben, B. N. Science

1997, 278, 827–830.

10. Dentener, F. J.; Crutzen. P. J. J Geophys Res-Atmos

1993, 98, 7149–7163.

11. Zhang, Y.; Sunwoo, Y.; Kotamarthi, V.; Carmichael, R.

J Appl Meteorol 1994, 33, 813–824.


12. Hanisch, F.; Crowley, J. N. Atmos Chem Phys 2003, 3,

119–130.

13. Bauer, S. E.; Balkanski, Y.; Schulz, M.; Hauglustaine,

D. A.; Dentener, F. J Geophys Res-Atmos 2004, 109, art.

no.-D02304, doi:10.1029/2003JD003868.

14. Bian, H. S.; Zender, C. S. J Geophys Res-Atmos 2003,

108, 4672.

15. Caloz, F.; Fenter, F. F.; Tabor, K. D.; Rossi, M. J. Rev

Sci Instrum 1997, 68, 3172–3179.

16. Atkinson, R.; Baulch, D. L.; Cox, R. A.; Hampson,

R. F.; Kerr, J. A.; Rossi, M. J.; Troe, J. J Phys Chem

Ref Data 1997, 26, 521–1011.

17. Li, W.; Gibbs. G. V.; Oyama, S. T. J Am Chem Soc 1998,

120, 9041–9046.

18. Li, W.; Oyama, S. T. J Am Chem Soc 1998, 120,

9047–9052.

19. Karagulian, F.; Rossi, M. J. Phys Chem Chem Phys 2005,

7, 3150.

20. Keyser, L. F.; Moore, S. B.; Leu, M. T. J Phys Chem

1991, 95, 5496–5502.

21. Sullivan, R. C.; Thornberry, T.; Abbatt, J. P. D. Atmos

Chem Phys 2004, 4, 1301–1310.

22. Michel, A. E.; Usher, C. R.; Grassian, V. H. Atmos En-

viron 2003, 37, 3201–3211.

23. Alebic-Juretic, A.; Cvitas, T.; Klasinc, L.

Ber Bunsen Ges Phys Chem 1992, 96, 493–

495.

24. Hanisch, F.; Crowley, J. N. Atmos Chem Phys 2003, 3,

119–130.

25. Michel, A. E.; Usher, C. R.; Grassian, V. H. Geophys

Res Lett 2002, 29, art. no.-1665.

26. Garland, J. A. US ERDA Symposium Series, Richland,

WA, USA, 1974.

27. de Reus, M.; Dentener, F.; Thomas, A.; Borrmann, S.;

Strom, J.; Lelieveld, J. J Geophys Res-Atmos 2000, 105,

15263–15275.

the heterogeneous decomposition of ozone on atmospheric mineral dust surrogates at ambient...

Documents