Heterogeneous reaction of ozone with hydrocarbon flame soot

The reaction of ozone with toluene and kerosene flame soot was studied over the temperature range 240 to 350 K using a low pressure (a few Torr) flow reactor coupled to a modulated molecular beam mass spectrometer.

A flat-flame burner was used for the preparation and deposition of soot samples from premixed flames of liquid fuels under well controlled and adjustable combustion conditions.

Soot was found to be deactivated in reaction with ozone, the uptake coefficient (γ) being dependent on the time of exposure.

The values of (1.8 ± 0.7) × 10−4 and (3.8 ± 1.5) × 10−4 independent of temperature in the range 240–350 K were determined for the initial uptake coefficient of ozone on toluene and kerosene soot, respectively.

The process of soot ageing (deactivation) was parameterized, the uptake coefficient being expressed as a function of time and gas phase ozone concentration: γ = γ0/(1 + γ0k[O3]t), with temperature independent values of k = (1.1 ± 0.4) × 10−10 and (6.2 ± 2.5) × 10−11 cm3molecule−1s−1, for toluene and kerosene soot, respectively.

From the soot surface saturation experiments the following maximum number of ozone molecules taken up were determined: ≃7 × 1014 for toluene and ≃9 × 1014 molecule cm−2 for kerosene soot.

Experiments on soot ageing confirmed that soot deactivation occurs under real ambient conditions.

The present results support current considerations that heterogeneous loss of ozone on soot has negligible impact on ozone concentration throughout the atmosphere.


Carbonaceous aerosols, which derive from biogenic sources (organic carbon) and from the incomplete combustion of fossil fuels and biomass (black carbon as well as organic carbon), are now recognized for their potential impact on climate.1

Carbonaceous particles directly influence the radiative budget of the atmosphere (an extinction and absorption of solar and terrestrial radiation) and also indirectly as they may serve as condensation nuclei for the formation of clouds.

Besides these “radiative” aspects of the carbonaceous aerosol role in the atmosphere, soot particles providing high surface area for heterogeneous reactions may also influence the chemical composition of the atmosphere by reacting with important atmospheric trace gases.2–4

The present study addresses this last issue and concerns an experimental investigation of the reaction of ozone with soot.

Ozone reaction with soot has been studied in several laboratories, as reported in recent data reviews5,6 (see Discussion section for details), yet the reported values of the uptake coefficient (γ) are highly variable (between 10−3 and 10−8).

This is due to the dependence of the measured values of γ on the surface area of the soot samples accessible to O3, soot type and soot deactivation process leading to a dependence of the uptake coefficient on the time scale of experiments and ozone concentrations used.

In the present study we used a low pressure flow reactor combined with mass spectrometric detection of gas phase species to measure the uptake coefficient of ozone on toluene and kerosene flame soot samples prepared under well controlled and adjustable combustion conditions.

The process of soot ageing (deactivation) was parameterized, the uptake coefficient being expressed as a function of time and gas phase ozone concentration in the temperature range T = (240–350) K.

Experimental section

The mode of flame soot generation and combustion conditions (particularly the fuel/air ratio) are known to influence the chemical reactivity of soot samples.7,8

In the present study we have adapted the flat-flame burner existing in the laboratory (LCSR)9 for the preparation and deposition of soot samples from premixed flames of liquid fuels in a reproducible way and under well controlled combustion conditions.

This soot generation system is shown in Fig. 1.

The liquid fuel pumped with a Gilson pump manufactured for liquid chromatography was first atomized by forcing it through a small orifice by a high pressure (a few atm) nitrogen jet and then vaporized in a heated chamber.

Then oxygen and nitrogen flows were added to this fuel vapor-nitrogen mixture.

All these gases were mixed in a tubular volume filled with glass beads.

The liquid flow rate was regulated and controlled with Gilson pump and the flow rates of nitrogen and oxygen were measured with mass flowmeters (MKS Instruments).

Resulting mixture with known fuel/oxygen ratio was transported to the flat-flame burner.

The matrix of the burner was made of a brass disc drilled with 0.5 mm diameter holes on a 4 cm diameter circular area.

It was placed into a Pyrex tube to prevent the flame perturbation by external air movements.

Additional nitrogen was added into the tube surrounding the burner in order to avoid external oxygen diffusion and preserve real premixed flames with well defined richness.

In order to prevent any fuel condensation all the parts of the system and connection lines were heated at adequate temperature.

The temperature was controlled with thermocouples located at different points of the system.

Toluene and a mixture of hydrocarbons (decane∶propylbenzene∶propylhexane = 74∶15∶11) were used as fuels in the present study.

This last mixture (which will be referred as kerosene in the paper) was chosen as a proxy of kerosene as it represents well the combustion of kerosene10 and also facilitates soot preparation due to a small number of hydrocarbon constituents (with lower boiling points) compared with kerosene.

Soot samples from stabilized premixed flames of these liquid fuels were deposited on Pyrex rods (0.9 cm od), which were rotated and moved through the flame as shown in Fig. 1.

The soot was sampled at different locations in the flame (from 1 to 4 cm above the burner surface).

The thickness of soot coverage was determined directly by means of an universal optical microscope (Reichert meF2).

The microscope was sequentially focused on the support Pyrex surface and the surface of the soot.

The soot coverage thickness was determined as the difference between two readings.

Repeating this operation in several points along the soot samples, it was found that the procedure of soot deposition used in this study provided a homogeneous soot coverage (within 20%).

The thickness of soot samples used was varied in the range 20–300 μm.

As one could expect, the mass of the soot deposited on the outer surface of the support Pyrex tube depended linearly on the deposition time.

Examples of such dependencies for the flames with different richness are shown in Fig. 2, where flame richness represents the fuel/oxygen ratio multiplied by stoichiometric coefficient of oxygen.

The flame richness is a key parameter determining the rate of soot formation.

The results from Fig. 2 are presented in Fig. 3 as a dependence of the rate of soot deposition (in mg per 1 cm length of the support tube and per minute) on the toluene flame richness.

The line on Fig. 3 corresponds to a power functione.g.:11 soot deposition rate ≃ (ϕ − ϕc)nwhere ϕc is the critical richness (ϕc ≈ 1.6 for n = 2).

At ϕ < ϕc soot is not formed.

It is clear that from the few data points shown in Fig. 3 it is not possible to determine ϕc and n independently (ϕc is in the range 1.45–1.7 for the values of n between 3 and 1).

However, the important point in the context of the present study was to show existence of the critical richness and strong dependence of the soot formation rate on the flame richness.

The soot samples used in the present study were obtained from the flames with richness between 1.7 and 2.2.

The lower limit of the flame richness was imposed by very low rate of soot formation and the upper one by the difficulties in stabilization of the rich flames over the burner.

Specific (BET) surface area is an important parameter for the quantitative characterization of soot reactivity toward different chemical species.

In the present study BET surface measurements of the soot samples used in the uptake experiments were performed with ASAP-2000 apparatus and nitrogen as adsorbate gas (soot sample was removed from the support tube prior to these measurements).

The values of 175 ± 25 and 120 ± 20 m2 g−1 were found for soot formed in toluene and kerosene flames, respectively.

In the range of the quoted uncertainty and under experimental conditions used no influence on the BET surface area of the flame richness, location in the flame and treatment with ozone was observed.

Interaction of ozone with soot was studied using the flow tube technique and a modulated molecular beam mass spectrometer as the detection method.

A Pyrex tube with the deposited soot sample was introduced into the flow reactor along its axis.

The main reactor (Fig. 4) consisted of a Pyrex tube (45 cm length and 2.4 cm id) with a jacket for the thermostated liquid circulation (water or ethanol).

The movable injector consisted of three coaxial Pyrex tubes as shown in Fig. 4.

The tube with the soot coverage could be moved relative to the outer tube of the injector.

This allowed the variation of the soot sample length exposed to ozone, which was introduced through the side arm of the main reactor.

Thus the reaction time (t) was defined as the soot sample length (l) exposed to O3 divided by the average flow velocity in the reactor (v): t = l/v.

The inner tube of the movable injector was used for circulation of the thermostated liquid inside the tube with soot sample.

This allowed to maintain the same temperature in the main reactor and on the soot surface in the measurements of the temperature dependence of the uptake coefficient.

Fresh soot samples were used in all kinetic experiments.

Generally, freshly prepared soot sample was introduced into the flow reactor and pumped during 15–20 min before being exposed to ozone.

It was verified that pumping for a longer time had no influence on the kinetics of ozone loss (see Fig. 5).

Ozone was produced by an ozonizer (Trailigaz), collected in a liquid nitrogen trap, pumped in order to reduce the O2 concentration and, finally, diluted in Helium and stored in a glass flask.

The absolute concentration of O3 was derived using the reaction of ozone with NO with simultaneous detection of ozone consumed and NO2 formed (Δ[O3] = Δ[NO2]): NO + O3 → NO2 + O2k1 = 2.0 × 10−12exp(−1400/T) cm3molecule−1s−15.The mass spectrometer was calibrated for NO2 by letting a known concentration of NO2 flow into the reactor.


Fig. 5 represents a typical behavior of the ozone concentration when soot sample is moved into the reaction zone.

Fast initial consumption of O3 followed by rapid surface deactivation leading to a decrease of ozone loss rate was observed.

In additional experiments, the soot sample was withdrawn from the reaction zone at different times after the initial exposure to ozone.

Under these conditions, when ozone is no longer in contact with soot surface, ozone concentration relaxed rapidly toward its initial value.

No additional ozone, which could be desorbed from the soot surface was observed.

These measurements indicate that ozone loss on soot is an irreversible process.

The uptake coefficient, γ, was determined as the probability of the irreversible ozone loss per collision of O3 with soot surface: where k′ (in s−1) is the rate constant of ozone loss, ω the average molecular speed, V the volume of the reaction zone and S is the surface area of the soot sample.

To calculate the uptake coefficient two parameters have to be determined experimentally: the rate constant of the heterogeneous ozone loss (k′) and the soot surface area (S) accessible for interaction with ozone.

Determination of k

Kinetics of ozone loss

In order to verify if the first-order loss approximation is applicable for the determination of k′ from ozone loss kinetics, a series of experiments was carried out where similar fresh soot samples of different lengths (2, 4, 6 and 8 cm) were successively exposed to the same initial concentration of ozone.

Variation of the soot sample length is equivalent, in fact, to variation of the reaction time (see Experimental Section).

The results are displayed in Fig. 6a as the kinetics of ozone decay for different times of exposure.

The straight lines are the exponential fits to the experimental points.

The conclusion from these experiments is that for given exposure time and initial ozone concentration the kinetics of ozone loss on soot can be treated with the first-order kinetics formalism and the rate constant can be determined as: where t is the reaction time defined by sample length and flow velocity: t = l/v.

It should be noted at this point that k′ determined in this way is dependent on initial ozone concentration because soot is deactivated during the reaction and the rate of this deactivation is expected to be dependent on the ozone concentration.

The results from Fig. 6a are presented in Fig. 6b as the dependence of k′ (calculated from eqn. (II)) on the time of soot exposure to ozone.

The uncertainties on determination of k′ shown in Fig. 6b are between 13 and 25% (lower for higher k′ values) and include the uncertainty on soot mass measurements.

Good agreement between the values of k′ obtained with soot samples of different length justifies the procedure used for the calculations and shows the good reproducibility of the measurements.

In addition, the graph well illustrates the behavior described above: the ozone loss rate is high at the beginning of the reaction and rapidly decreases during the exposure.

Pressure dependence

The measurements of the rates of heterogeneous loss of gas phase species in flow reactor from the kinetics of the axial decrease of their concentration is complicated by simultaneous occurrence of two processes: diffusion and uptake of the molecules on the reactor wall.

The treatment of the kinetics of heterogeneous decay is simple only in the case of so called “kinetic regime”, when the uptake of molecules on the wall is much slower than their diffusione.g..12

When an efficient heterogeneous loss leads to the important local depletion of the gas phase molecules close to the surface, their diffusion from volume toward the surface becomes a rate limiting step in the heterogeneous reaction and should be taken into account in the processing of the experimental data.

For the determination of the first order rate constant for the kinetic regime we have used an approach based on the kinetic resistance additivity rule:13 where k′obs is first order rate constant measured from heterogeneous decay kinetics as shown above, k′kin and k′dif are the kinetic and diffusion limits of the rate constant, respectively.

In the present work the heterogeneous interactions of the gas phase species with soot samples were studied in a coaxial reactor with an active central rod (covered with soot) and passive surface of the main reactor (covered with halocarbon wax).

The radial diffusion problem for this type of reactor configuration was solved by Gershenzon and co-workers.14–16

For k′dif under fast flow conditions (v → ∞) they give an expression similar to that for a cylindrical reactor: where Kd(q) is dimensionless rate constant of radial diffusion, which is a function of ratio q = r/R, where r is the external radius of the coated internal tube and R is the internal radius of the main reactor.

For our reactor, r = 0.45 and R = 1.2 cm, q = 0.375 and Kd(q) ≈ 4..416

Thus: and where D0 is the diffusion coefficient of O3 in He at 1 Torr pressure (Torr cm2 s−1) and P is the total pressure in the reactor (Torr of He).

A series of experiments has been carried out, where the rate of ozone loss on soot was measured as a function of pressure in the reactor in the range 0.5–5.0 Torr.

It was observed that the rate of O3 loss is strongly dependent on the total pressure in the reactor, with a decrease as pressure increased.

An example of the dependence of 1/kobs on pressure measured at T = 240 K is shown in Fig. 7.

The experimental points measured at different times of soot exposure to ozone can be well fitted with linear regression according to eqn. (IV).

The intercept provides the value of 1/k′kin for a given time of exposure.

The slope of the straight line, α = R2/4.4D0, defines the correction factor which has to be applied to a measured value of kobs in order to obtain the value of kkin: For 1 Torr total pressure (generally used in our experiments): One can note that the slopes of the three straight lines corresponding to different exposure times are identical in the range of experimental uncertainty, which indicates that the correction factor α is independent of the magnitude of k′obs.

Similar series of experiments have also been carried out at T = 298 and 350 K. The values for the correction factor α determined at T = 240, 298 and 350 K are 2.2 × 10−3, 1.7 × 10−3 and 1.4 × 10−3 s, respectively.

As one could expect the correction factor is lower at higher temperature, where the gas phase diffusion is faster.

Eqn. (IV) including the empirically found correction factors α was used for the calculation of k′kin from the experimental values of k′obs for the O3–soot interaction.

Maximum corrections on k′obs were around 40%.

For the intermediate temperatures T = 265 and 320 K extrapolated values of α were used.

The knowledge of the parameter α = R2/4.4D0 allows the determination of D0, the diffusion coefficient of O3 in He at 1 Torr total pressure.

Thus, for T = 298 K the value of D0 = (205 ± 15) cm2 s−1 (with 2σ statistical uncertainty) can be calculated.

This value is a factor of two lower than recently reported: D0(O3–He) = (410 ± 25) Torr cm2 s−1.17

The reason for the lower value of D0 of the present study is not clear.

It can be an artifact of our experiments or questionable applicability of the Kd(q)-values15,16 to our reactor configuration.

It is interesting to note that a similar systematic (factor of two) discrepancy was observed between literature values for HO2 and OH diffusion coefficients in helium and those obtained in our unpublished study applying the kinetic resistance additivity rule to our coaxial reactor.

Additional experiments with variation of the geometry of the reactor (q = r/R parameter) are under way to confirm the applicability of eqn. (III) (more specially of the Kd(q) coefficients16), at least, to the reactor configuration used in the present study.

Dependence on flame richness

Fig. 8 shows the results of the experiments where the rate of ozone loss on soot was measured on soot samples of similar mass and length, however generated under different combustion conditions (flame richness).

One can note that no dependence of the reaction rate on the flame richness was observed in the range of the quoted experimental uncertainty (14–25%).

Although a rather narrow range of flame richness was used in these experiments, however, as was shown in Experimental section, even a slight variation of this parameter had an important influence on the flame structure and especially on the rate of soot formation.

Dependence on soot sampling position

The rate of ozone/soot interaction was studied as a function of soot sampling position in the flame.

Experiments were carried out with similar (mass and length) soot samples collected at 1, 2.5 and 4 cm above the burner surface.

As shown in Fig. 9 no influence of the soot ageing in the flame on the reactivity of soot toward ozone was observed under experimental conditions used.

All the experiments described below were carried out with soot sampled at 2.5 cm height above the burner surface.

Surface area

Specific experiments have been carried out in order to determine the soot surface area accessible to ozone.

For that, the rate of ozone loss on soot k′ was studied as a function of the mass of soot deposed on the unity length of the support tube.

In fact, this dependence is equivalent to the dependence of k′ on the thickness of soot coating.

Results are shown in Fig. 10.

For all the plots corresponding to a different time of soot exposure to ozone two regimes are observed:18 the first one, where k′ is linearly dependent on the mass of soot sample and the second one (saturation region), where k′ is independent of the sample mass.

Linear dependence of the reaction rate on mass, hence thickness, of soot coating indicates that the entire surface area of soot samples is involved in the interaction with ozone and, consequently, the BET surface area should be used for calculations of the uptake coefficient.

The soot mass range, corresponding to the linear regime, is different for different exposure times.

This could be expected, since the soot mass probed by ozone is defined by two concurrent processes: diffusion of ozone into soot bulk and ozone loss rate which decreases upon deactivation of soot sample.

The uptake measurements in the present study were carried out with soot sample masses lower than 0.4 mg cm−1, where linear dependence of the reaction rate on soot mass was observed for exposure times higher than 30 s.

Uptake coefficient

The kinetics of ozone consumption on soot surface was investigated as a function of the initial concentration of ozone.

The results obtained with different initial concentrations of ozone are shown on Fig. 11.

We observe similar behavior with an almost total consumption of ozone in the first seconds of the reaction, followed by a progressive return to the initial concentration of ozone.

This behavior indicates that a deactivation process occurs, the rate of which increases with increasing initial ozone concentration.

Thus the reaction rate (and consequently the uptake coefficient) depends on two parameters: exposure time and ozone concentration.

Considering that soot reactivity is defined by the number of active sites on its surface, which are consumed in the reaction with ozone and in order to describe the process of soot deactivation by only one parameter, we attempted to express the uptake coefficient as a function of the number of ozone molecules taken up per unit surface area of soot.

An example is presented in Fig. 12, where the straight line represents an exponential fit to the experimental points.

One can note that all the data corresponding to the exposure time from 0.5 to 15 min and to the initial ozone concentration varied by an order of magnitude can be represented by a simple expression: γ = γ0exp(−βΔ[O3])where γ0 is the initial uptake coefficient, β a constant characterizing the deactivation processes and Δ[O3] the number of ozone molecules taken up by 1 cm2 surface area.

This approach allows the determination of the initial uptake coefficient by extrapolation of the experimental data to the beginning of the reaction, when Δ[O3] = 0.

The values of γ0, derived in this way for kerosene and toluene soot at different temperatures in the range 240–350 K are reported in Tables 1 and 2, respectively.

Although eqn. (VI) reproduces well the values of γ at initial rapid stage of the reaction, its application for the calculation of the time dependence of the uptake coefficient at a given concentration of ozone at longer reaction times, for instance for atmospheric application, seems to be rather difficult and most probably incorrect.

That is why we have tried to find a parametric representation of γ as a function of time and ozone concentration.

First, it was observed that reciprocal of the uptake coefficient for given initial concentration of ozone can be well represented by a linear function of the exposure time: where C is a coefficient characterizing the soot deactivation rate (decrease of γ) and depending on the concentration of ozone.

Examples of such plots observed at different initial ozone concentrations are shown in Fig. 13.

The slopes of the straight lines in Fig. 13 provide the values of C, which are presented in Fig. 14 as a function of the concentration of O3.

It should be noted at this point that in the data analysis the ozone concentration was considered to be constant along the reaction zone (limited by soot sample length) while it was not the case.

In fact, the ozone concentration decreased along the soot sample from its maximum value ([O3]0) at the upstream end of the soot sample to the minimum one at the downstream end of the reaction zone.

In the present experiments this ozone consumption ranged from ≈50% to a few percent depending on the exposure time and the mean ozone concentration along the reaction zone was used in the calculations (an uncertainty of near 20% on this ozone concentration scaling was estimated and taken into account in the final results).

Another point is that the uptake coefficient being dependent on [O3], the ozone concentration profile along the soot sample could lead to the “gradient of soot reactivity” along the soot sample length.

This issue was neglected and not considered in the calculations.

Fig. 14 shows that the coefficient C can be well approximated by linear function of the ozone concentration: C = k[O3].

The value of the constant k, which has a dimension of a second order rate constant (cm3 molecule−1 s−1), can be easily determined from the slope of the straight line in Fig. 14.

Thus, the uptake coefficient can be, finally, represented by the following expression: which can be written as: The values of k for ozone reaction with kerosene and toluene soot determined at different temperatures are given in Tables 1 and 2, respectively.

One can note that reactivity of soot samples under investigation as well as the rate of soot deactivation were found independent of temperature in the range T = 240–350 K. The following values of γ0 and k can be recommended from this study for this temperature range: γ0 = (1.8 ± 0.7) × 10−4 and k = (1.1 ± 0.4) × 10−10 cm3 molecule−1 s−1 (toluene)γ0 = (3.8 ± 1.5) × 10−4 and k = (6.2 ± 2.5) × 10−11 cm3 molecule−1 s−1 (kerosene).Quoted uncertainties are near 40% for both γ0 and k and represent a combination of statistical and estimated systematical errors, including those on k′, mass of soot samples, specific surface area and absolute ozone concentrations measurements.

Two limiting cases for γ in eqn. (VII) can be noted: γ0k[O3]t ≪ 1 resulting in γ = γ0 and γ0k[O3]t ≫ 1 resulting in γ = 1/k[O3]t.

In the first case (low ozone concentrations) the uptake coefficient is independent of the volume O3 concentration and exposure time.

In the second case, it is inversely proportional to both [O3] and time of exposure.

The parameters γ0 and k in eqn. (VII) which describes decrease of soot reactivity with time were obtained from the experimental data corresponding to the first 10–20 min of soot exposure to ozone.

In order to verify that eqn. (VII) still holds when the ratio [O3]∶t is varied while the product [O3]t is kept in the sensitivity range of our experiments, soot samples were exposed to relatively low ozone concentration, [O3]0 = 5.5 × 1011 molecule cm−3, for up to 140 min.

The results obtained for temporal behavior of the uptake coefficient are shown in Fig. 15.

The experimental points were fitted with eqn. (VII), where γ0 and k were variable parameters.

The curve in Fig. 15 corresponds to the best fit obtained with γ0 = 4.2 × 10−4 and k = 4.7 × 10−11 cm3 molecule−1 s−1.

These values are in good agreement with those reported above considering the experimental uncertainties.

Thus, expression (VII) is applicable for the description of soot deactivation process under largely varied conditions: exposure time from a few minutes up to a few hours, initial concentration of ozone from 5 × 1011 to ≃1013 molecule cm−3.

In another series of experiments soot reactivity toward O3 was studied as a function of time of the soot exposure under outdoor conditions.

The day-time concentrations of ozone and NO2 in ambient air were measured to be (75 ± 10) and (2.5–8) ppb, respectively.

Soot samples with similar masses (≈0.3 mg cm−1 × 6 cm) were exposed outside for different times (up to 7 h).

The initial uptake coefficient (γ0) was chosen as a parameter to characterize the outdoor soot ageing process.

This parameter was determined by introduction of the aged soot samples into the flow reactor and measuring their reactivity toward ozone.

Fig. 16 shows the measured values of the uptake coefficient as a function of the number of ozone molecules taken up by the soot surface in the reactor for soot samples being exposed outside for different times.

The time, 2 min, for the fresh soot corresponds, in fact, to the time which is necessary for installation of the freshly prepared soot sample in the flow reactor.

The deactivation of soot under outdoor conditions is clearly observed.

The values of γ0 obtained from the data in Fig. 16 are presented in Fig. 17 as a function of the time of soot ageing.

The line in Fig. 17 is a fit to the experimental points according to eqn. (VII).

The best fit is obtained with γ0 = 4 × 10−4 and k = 4.3 × 10−13 cm3 molecule−1 s−1.

The value of γ0 agrees with that measured above for ozone reaction with kerosene, however the value of k is around two orders of magnitude lower than corresponding value measured in experiments on soot deactivation in the reactor.

This lower rate of soot deactivation under outdoor conditions, compared to that in the low pressure reactor, could be expected since the results obtained in a kinetic regime (reactor) could not be directly applied to the experiments on soot ageing outside, where the diffusion of gas phase species toward our “compact” (not dispersed) soot sample is a limiting factor of the heterogeneous reaction (diffusion regime).

It should be noted that these diffusion limitations do not occur in the case of small atmospheric soot particles (with diameter of a few tens nm) dispersed in the atmosphere and eqn. (VII) may be applicable under such atmospheric conditions.

Number of active sites

Another parameter characterizing the soot reactivity toward ozone is the maximum number of ozone molecules which can be lost on the unity surface area of soot sample.

If reaction is considered as non catalytic (one ozone molecule lost per active site) which seems to be the case at least for the first rapid reaction stage this parameter can be considered as a number of active (toward ozone) sites on the soot surface.

To determine this parameter we measured the total number of ozone molecules consumed on the soot surface from the beginning of the reaction up to the complete soot sample deactivation (defined as an absence of observable ozone loss within experimental accuracy).

Exposure time to reach the total soot surface saturation was in the range 30–60 min depending on soot sample mass and ozone concentration used.

The results obtained for toluene soot are shown in Fig. 18, where the number of ozone molecules taken up is presented as a function of mass of soot samples generated in flames of different richness.

The number of active sites per mg of soot derived from the slope of the straight line of Fig. 18 is 1.2 × 1018 sites mg−1, independent of flame richness.

Similarly, for the kerosene soot the value of 1.1 × 1018 sites mg−1 was determined.

Appling the BET surface area (175 and 120 m2g−1, respectively for toluene and kerosene soot), one calculates the respective values of the number of active sites per unity of surface area: 7 × 1014 sites cm−2 and 9 × 1014 sites cm−2.

It can be noted that these numbers measured with the flame soot are very close to those reported by Pöschl et al19. and Kamm et al20. for spark discharge soot aerosol particles: 5.6 × 1014 and 6.5 × 1014 sites cm−2, respectively.

Concerning the results reported in this section it should be noted that if the 1/t-dependence of γ (eqn. (VII)) is assumed to be valid for long exposure times (t → ∞), then there is no surface saturation and the maximum number of ozone molecules taken up by the soot surface can not be determined since the number of consumed ozone molecules does not level off at any value and is going to infinity with time.

Considering this as well as the method applied to determine the total number of consumed ozone molecules (experimental sensitivity limitations in determining the soot surface saturation), the total number of ozone molecules consumed per unit soot surface area determined in the present study has to be considered as a lower limit.


Comparison with literature data

Kinetics of ozone reaction with soot has been the subject of numerous previous studies (Table 3).

Stephens et al.,21 using a Knudsen cell combined with mass spectrometric detection of gas phase species, have measured the uptake coefficients of ozone on commercial ground charcoal (BET surface area 37 m2g−1): (0.2–4.1) × 10−3 and (2.7–11.3) × 10−5 for initial and steady state uptake coefficients, respectively.

The uptake coefficient was found to be dependent on the carbon sample, flow rate and concentration of ozone.

Using similar experimental apparatus Rogaski et al22. studied the interaction of ozone with a commercial amorphous carbon (Degussa FW-2, BET surface area 460 m2g−1).

The O3 uptake coefficient (determined considering a geometric surface area) was (1 ± 0.7) × 10−3 and slightly decreased with time.

Fendel et al23. used flow tube technique to investigate the reaction of ozone with carbon aerosol (of 10–100 nm radius, BET surface area 395 m2 g−1) produced in a spark discharge generator.

The ozone volume mixing ratio as a function of the aerosol concentration was monitored for (15–30) s reaction time.

The resulting uptake coefficient was found to be strongly dependent on the initial concentration of ozone: γ, calculated using specific surface area, decreased from 3.3 × 10−3 to 2.2 × 10−4, ozone being increased from 160 to 915 ppb.

In the flow tube study of Longfellow et al24. the kinetics of ozone reaction with methane soot was monitored by using chemical ionisation mass spectrometry.

Considering geometric surface area, the authors report the value of 7 × 10−2 for initial uptake of ozone on 3.5 mg of fresh soot, which drops to ≃4 × 10−3 on soot exposed to O3 (≃3 × 1010 molecule cm−3) for 30 min and to 1.6 × 10−4 for 7 h of exposure (soot sample of 2.2 mg).

Following the authors and assuming a 30-fold (for 3.5 mg sample) decrease in γ if available surface area instead of geometric one is applied, one gets: γ0 ≃ 2.3 × 10−3, γ0/γ(30 min) ≃ 18 and γ0/γ(7 h) ≃ 275.

It is interesting to note that the application of eqn. (VII) for γ derived in the present study (with k = 1.5 × 10−10 cm−3 molecule−1 s−1, which is an upper limit of k for toluene soot) to these data gives: γ0/γ(30 min) ≃ 20 and γ0/γ(7 h) ≃ 261, i.e. very close to the experimental values.24

This shows that the empirical eqn. (VII) derived from the experimental data for the first minutes of the ozone + soot reaction and using relatively high initial concentrations of ozone could be applicable at longer reaction times and lower ozone concentrations.

All studies cited above dealt with the first rapid stage of ozone–soot interaction.

On contrary, in a aerosol chamber study of Kamm et al20. the experiments were carried out on a time scale of 72 h, where the second slow stage of the uptake process was investigated, the rapid initial reaction stage being not resolved on the time scale of the experiments.

A reaction probability γ = 1.2 × 10−6 at T = 296 K was observed and was shown to decrease with a characteristic time of about 12 h.

This slow exponential dependence of γ with time differs from the 1/t dependence of γ derived in our study for the fast initial reaction stage.

In another static aerosol reactor study by Disselkamp et al25. the values of ≃10−3 and ≃10−8 were reported, respectively for the initial and steady state uptake coefficients of ozone on carbon black aerosols (Degussa FW1).

Interestingly, if eqn. (VII) with k ≃ 10−10 cm3 molecule−1 s−1 is applied for estimation of γ under the experimental conditions of the study by Disselkamp et al.25

(ozone partial pressure ≃3.5 × 10−5 atm, characteristic exposure time around 50 min) one calculates γ ≃ 4 × 10−8 in fair agreement with their experimental value.

Analysis of the data available in the literature (Table 3) shows that the values of the initial uptake coefficient measured in different studies with different soot samples and under different experimental conditions are all near 10−3 (with uncertainty factor of 4).

However quantitative data on evolution of γ with time due to deactivation of soot are very scattered: so called “steady state” values of γ vary between 10−4 and 10−8.

This could be expected, since the reaction rate depends strongly on two parameters, exposure time and ozone concentration.

Thus the measured values of γ are strongly dependent on the time scale of experiments and concentrations of ozone used.

In this respect, the analytical expression derived in the present study for the uptake coefficient, where this latter is represented as a function of exposure time and concentration of ozone, seems to be a useful tool for determination of the heterogeneous reaction rate under given experimental or ambient conditions.

In the present study the uptake coefficient of ozone on kerosene and toluene flame soot surface was found to be independent of temperature in the range (240–350) K. This is in disagreement with the results of previous studies,20,24,26,27 where relatively strong temperature dependence of the uptake coefficient was observed.

On methane soot previously exposed to ozone Longfellow et al24. measured γ ≃ 4 × 10−3 at T = 298 K and a factor of three lower value, (1.3 ± 0.3) × 10−3, at T = 253 K. Il’in et al26. reported a temperature factor of 780 and 1000 K for ozone reaction on fresh and aged flame soot, respectively.

Finally, in the most recent temperature dependence study by Chunghtai et al27. the reaction of ozone with n-hexane flame soot beyond its initial stage was shown to proceed with an activation energy of (12.9 ± 0.5) kJ mol−1.

The reason for the disagreement between our results and those from other studies is not clear.

Although all previous temperature dependence studies dealt mainly with the second slow reaction stage (aged soot), while the results of the present study are relevant rather to the first rapid reaction stage on fresh soot sample.

In several studies the gas phase products of the ozone reaction with soot surface were detected.

Stephens et al21. observed an oxygen molecule formation for each ozone molecule lost on the surface.

In addition, they have found CO and CO2 in the reaction products with the yield of 20 to 40%.

The formation of O2 was supported by study of Rogaski et al.,22 where O2 was observed, however its yield was not quantified.

Disselkamp et al25. measured a stoichiometric coefficient of 2 for CO2 production in reaction of ozone with soot, however they have not found CO in the reaction products, in contradiction with the results of Stephens et al21. and Smith et al.28

In the present study the possible reaction products CO, CO2, O2 were difficult to detect by mass spectrometry because of the presence of very high residual signals at m/z = 28 (CO+, N2+) and 32 (O2+) and very low sensibility (very high background noise) of the mass spectrometer at m/z = 44 (CO2+).

Concerning the reaction mechanism, Stephens et al21. and Pöschl et al19. proposed a Langmuir type kinetic model, consisting in reversible adsorption of O3, followed by slow reaction on the surface.

Kamm et al20. interpreted their experimental results by suggesting a slow catalytic ozone destruction process occurring after the fast initial stage of the surface reaction on fresh soot.

The fast initial catalytic decomposition of ozone was postulated by Smith and Chughtai.29

However, Disselkamp et al.,25 on the basis of the measurements of CO2 formation at the initial reaction stage, concluded that ozone loss is not a catalytic process.

In the present study, the ozone loss on soot was observed to be an irreversible process with the reaction rate depending on the number of active sites on the surface.

From another side, empirical eqn. (VII) derived from the experimental data corresponding to the first 10–20 min of soot exposure to ozone, implies (if it is correct at longer reaction times) a slow catalytic ozone destruction after the fast initial uptake phase.

Finally, our experimental data concerning the measurements of ozone loss rate at the first rapid reaction stage do not allow to draw any conclusion about the reaction mechanism at the molecular level, and about possible catalytic processes which may contribute to ozone destruction on much longer time scales.

Atmospheric implication

Previous two-dimensional model calculations4 suggested that a direct ozone loss on soot aerosols has the potential of explaining the lower stratospheric ozone depletion at northern hemisphere middle latitudes.

However the ozone loss on soot was considered as a catalytic process with a reaction probability of 10−3, excluding soot deactivation and subsequent decrease of its reactivity.

Experimental data from the present study allow for a rough estimation of the potential importance of the direct ozone loss on soot under real atmospheric conditions.

We have determined the number of ozone molecules taken up by the soot surface during the initial rapid stage of the ozone/soot interaction where the measured uptake coefficient drops by a factor of ≃100 from its initial value to γ ⩽ 10−6.

These numbers were 8.9 × 1014 and 7.1 × 1014 molecule cm−2 for kerosene and toluene flame soot, respectively.

Combining these values (we use 1015 molecule cm−2 in the calculations below) with the maximum atmospheric soot surface area density of ≃10−5 cm2cm−3 (based on soot concentrations of ≃10 μg m−3 measured in urban area near heavy traffic roads30–32 and surface to mass ratio of 140 m2g−1 which can be derived from the data reported by Blake and Kato33 for fractal geometry of soot particles) one can calculate the maximum amount of ozone which can be taken up by freshly emitted soot in a very polluted area.

The obtained value, ≃1010 molecule cm−3, is negligible compared with the typical ozone concentrations in the troposphere (⩾1012 molecule cm−3).

Considering that in the upper troposphere and lower stratosphere soot mass density is a factor of ≃104 (refs. 33–35) and in the free troposphere a factor of ≃10–103 (refs. 36–38) lower than the value used in the above calculations, the direct ozone loss due to the initial rapid reaction with soot can be considered as negligible throughout the atmosphere.

Concerning the second slower stage of the ozone reaction with deactivated soot aerosol (γ ≃ 10−7–10−8) the lifetime of ozone toward its reaction with soot can be estimated as: where ω is the mean molecular speed of ozone and σ is the area density of atmospheric soot.

Taking the maximum values for γ, ω and σ (10−7, 360 m s−1 at T = 298 K and 10−5 cm2cm−3, respectively) a lower limit of τ ⩾ 3.5 years can be determined for these extreme conditions, indicating a negligible contribution of the ozone/soot reaction to ozone loss processes in the atmosphere.

This is in agreement with the conclusions of previous laboratory studies20,24,25 and stratospheric measurements35,39.