Heterogeneous reactivity of gaseous nitric acid on Al2O3, CaCO3, and atmospheric dust samples: A Knudsen cell study

Article abstract of DOI:10.1021/jp001254+

The heterogeneous reaction between HNO₃ and various authentic and synthetic mineral dust/mineral oxide surfaces has been investigated using a low-pressure Knudsen reactor operating at 298 K. The surfaces used were Saharan dust from Cape Verde, Arizona dust, CaCO₃, and Al₂O₃. In all cases, a large irreversible uptake was observed. An uptake coefficient of γ = (11 ± 3) × 10^-2 was determined for Saharan dust, and γ= (6 ± 1.5) × 10^-2 was obtained for Arizona dust. The uptake coefficients for HNO₃ on heated CaCO₃ and on unheated CaCO₃ are given by γ = (10 ± 2.5) × 10^-2 and (18 ± 4.5) × 10^-2, respectively, and are in good agreement with previous results. CO₂ and H₂O were formed as gas-phase products. Measurements of the uptake coefficient of HNO₃ on grain-size selected samples of Al₂O₃, γ = (13 ± 3.3) × 10^-2, and systematic variation of sample mass enabled us to show that the geometrical surface area of the dust sample is appropriate for calculation of uptake coefficients in these experiments. The high reactivity of HNO₃ toward dust samples highlights the potentially important role of mineral dust in redistributing nitrate from the gaseous to the particulate phase and modifying tropospheric photochemical oxidation cycles.

Full text of DOI:10.1021/jp001254+

3096
J. Phys. Chem. A 2001, 105, 3096-3106
Heterogeneous Reactivity of Gaseous Nitric Acid on Al2O3, CaCO3, and Atmospheric Dust
Samples: A Knudsen Cell Study
F. Hanisch and J. N. Crowley*
Max-Planck-Institut fu¨r Chemie, DiVision of Atmospheric Chemistry, Postfach 3060, 55020 Mainz, Germany
ReceiVed: March 31, 2000; In Final Form: October 24, 2000
The heterogeneous reaction between HNO3 and various authentic and synthetic mineral dust/mineral oxide
surfaces has been investigated using a low-pressure Knudsen reactor operating at 298 K. The surfaces used
were Saharan dust from Cape Verde, Arizona dust, CaCO3, and Al2O3. In all cases, a large irreversible uptake
was observed. An uptake coefficient of γ ) (11 ( 3) × 10-2 was determined for Saharan dust, and γ ) (6
( 1.5) × 10-2 was obtained for Arizona dust. The uptake coefficients for HNO3 on heated CaCO3 and on
unheated CaCO3 are given by γ ) (10 ( 2.5) × 10-2 and (18 ( 4.5) × 10-2, respectively, and are in good
agreement with previous results. CO2 and H2O were formed as gas-phase products. Measurements of the
uptake coefficient of HNO3 on grain-size selected samples of Al2O3, γ ) (13 ( 3.3) × 10-2, and systematic
variation of sample mass enabled us to show that the geometrical surface area of the dust sample is appropriate
for calculation of uptake coefficients in these experiments. The high reactivity of HNO3 toward dust samples
highlights the potentially important role of mineral dust in redistributing nitrate from the gaseous to the
particulate phase and modifying tropospheric photochemical oxidation cycles.
1. Introduction
estimates that in vast areas of the northern and southern
hemisphere at least 40% of the total nitrate could be found on
dust.10
About 33% of the earth’s land surface is arid and a potential
source region for atmospheric mineral aerosol.1 Mineral aerosol
is a general expression for fine particles of crustal origin that
are generated by wind erosion, and which consist mostly of
silica and silicate minerals. Mineral aerosol is uplifted into the
atmosphere by strong surface winds that travel behind cold
frontal systems.2 Current estimates of annual dust emissions are
in the range of 1000-3000 Tg/year.3 The major source areas
of the world are situated in a broad band that extends from the
west coast of North Africa across the Arabian Peninsula to
central Asia (see, e.g., ref 4). Particles smaller than 10 µm have
atmospheric lifetimes of weeks,4 and mineral aerosol may be
transported over thousands of kilometers enabling Saharan dust
to be transported to, e.g., Florida and Barbados5 and central
Asian dust to the northern Pacific islands.6 Globally, the most
important minerals of the clay fraction (<2 µm) transported in
dust storms are illite, kaolinite, chlorite, and montmorillonite/
smectite,7,4 whereas coarser particles mainly consist of quartz,
feldspars, and carbonates.8
In addition, model calculations11-13 of the [HNO3] to [NOx]
ratio in the troposphere are consistently greater (by a factor of
5-10) than the experimentally determined value of between 1
and 9 (see, e.g., refs 11 and 14). The divergence between
calculation and measurement could be reduced by including in
the model an additional process that converted HNO3 to NOx.
On the other hand, the fact that several models were able to
reproduce well the measured winter [HNO3] to [NOx] ratios
but not the summer ones15 may imply the presence of a missing
sink for HNO3 that shows a strong seasonal variability. The
strong seasonal variation in mineral dust concentrations, with
maximum emissions in the spring and summer months and lower
emissions in the winter and autumn months, means that mineral
dust could make a potentially significant impact on the
distribution of HNO3 in the atmosphere and its seasonal
variability.
There is also strong experimental evidence that indicates an
important role for mineral dust in modifying atmospheric trace
gas distributions. Mineral dust from central Asia has frequently
been found to be associated with nitrate and sulfate, the
concentrations of which increase with transport time over the
Japanese islands.16,17 Similar results have been reported for
several northern Pacific islands,18,19 for the western North
Atlantic,5 for the northwestern Indian Ocean,20 for the USA,21
and for the East Mediterranean Sea.22
Despite the results of field observations and modeling studies,
there have been very few laboratory experiments that investigate
the heterogeneous reactivity of HNO3 on mineral aerosol. In
the present work, we used a newly commissioned Knudsen
reactor to investigate uptake and reaction of HNO3 with selected
authentic dust samples (Saharan dust from Cape Verde and
Arizona dust) and with powder samples of CaCO3 and Al2O3.
Experiments on CaCO3 were performed because calcite is a
frequent constituent of desert dust (up to 30% of the total aerosol
In the past, mineral aerosol has been taken into consideration
by atmospheric scientists mainly in connection with the radiation
budget of the earth. Its properties with respect to absorption
and scattering of solar and terrestrial radiation and its suitability
as cloud condensation nuclei have been subject of investigation.
However, recent modeling studies9,10 have predicted that mineral
aerosol could also have a significant influence on atmospheric
chemistry by promoting heterogeneous reactions. At a loading
of 100 µg/m3, mineral aerosol has been predicted to reduce
[NOy] (NOy: total reactive nitrogen ) NOx + HNO3 + HONO
+ NO3 + 2N2O5 + HNO4 + PAN + RONO2 + ROONO2 +
RNO2, where NOx ) NO + NO2) by 56% during daytime
periods and by 98% during nighttime periods.9 A 3-D model
* To whom correspondence should be addressed.
10.1021/jp001254+ CCC: $20.00 © 2001 American Chemical Society
Published on Web 03/01/2001

Heterogeneous Reactivity of Gaseous Nitric Acid
J. Phys. Chem. A, Vol. 105, No. 13, 2001 3097
TABLE 1: Knudsen Reactor Parameters
Knudsen reactor parameter
volume, V
value
150 cm3
298 K
temperature, T
total pressure, p
HNO3 number density
HNO3 flow
0.4 mTorr
6.5 × 1010-1.3 × 1012/cm3
2.5 × 1013-1.8 × 1015/s
3.5 mm
escape orifice #4 diameter
escape orifice #5 diameter
escape orifice #6 diameter
escape orifice #4 escape rate, kesc,4
escape orifice #5 escape rate, kesc,5
escape orifice #6 escape rate, kesc,6
sample surface area, As
sample collision frequency, ω
5.9 mm
10.0 mm
0.309 (T/M)1/2 s-1
1.098 (T/M)1/2 s-1
4.235 (T/M)1/2 s-1
1.5 and 4.8 cm2
81.31 and 255.24 s-1
molecules through the orifice under molecular flow conditions
is the escape rate constant kesc. Theoretically, the escape rate
constant is related to the orifice size by
Figure 1. Schematic diagram of the experimental setup including
Knudsen reactor, quadrupole mass spectrometer (QMS), and mixing
line with flow controllers (FC). The pressure in the Knudsen reactor
was generally 0.4 mTorr; the pressure in the mixing line held constant
at 0.5 Torr. The pressure in the MS chamber was ≈10-6 Torr.
cjAesc
4V
kesc
)
(i)
mass for northern Saharan dust23) and which, because of its
alkalinity, is thought to make a considerable contribution to the
uptake of acidic trace gases on mineral aerosol. Experiments
on grain-size selected samples of Al2O3 were performed in order
to characterize the experimental setup and data analysis. Saharan
dust from Cape Verde is highly suited for uptake measurements
because it is characteristic of the mineral aerosol that is
transported from the Sahara over the Atlantic Ocean.24
In this study, we present uptake coefficients for HNO3 on
the above-mentioned substrates at room temperature, combined
with an evaluation of the influence of water on the uptake
process.
where cj is the average molecular velocity of the gas molecules,
Aesc is the surface area of the escape orifice, and V is the
Knudsen reactor volume. This expression corresponds to the
“collision” frequency of a gas molecule with the area of the
escape orifice. In practice, the experimental escape rate is smaller
than the theoretical one because of flow resistance in the
connection between Knudsen reactor and MS.
In the present setup, kesc is determined in a pulsed valve
experiment in which a short pulse of an unreactive gas is
introduced into the Knudsen reactor. The decay of its concentra-
tion in the Knudsen reactor is monitored with the MS and found
to be exponential:
2. Experimental Section
C(t) ) C0 exp(-kesct)
(ii)
This is the first description of the newly commissioned
Knudsen reactor setup, which is therefore presented in detail.
A schematic diagram is given in Figure 1.
However, pulsed valve experiments with HNO3 do not provide
reliable escape rate constants because HNO3 is a molecule with
a high affinity for surfaces and loss to nonpassivated reactor
walls competes with effusion through the orifice. To determine
the escape rate constants for the six orifice sizes, we performed
pulsed valve experiments with a number of other gases: N2
(28 atomic mass units (amu)), Ar (40 amu), NO2 (46 amu), SO2
(64 amu), and CFCl3 (136 amu). From the results obtained we
derived a general molecular mass dependent expression for the
escape rate constants. The average molecular velocity cj (in units
of cm/s) depends on the molar mass M of the gas molecule, as
shown by eq iii:
2.1. Knudsen Reactor/Mass Spectrometer (MS). The
Knudsen reactor consists of two parts. The upper part, supporting
the isolation plunger, is made from Pyrex glass and is fixed to
the gas mixing line; the lower part, which contains the sample,
is made of stainless steel and can be heated to 450 K. Both the
plunger and the lower part are coated with Teflon (DuPont FEP
121-A). The main compartment is 32 mm high and has a
diameter of 70 mm; the sample compartment is 15 mm high
and has a diameter of 30 mm. The total volume, including the
connections to the MS and to the pressure gauge, is just 150
cm3. The volume of the sample compartment was kept small
(10.6 cm3) in order to reduce artifacts caused by changes in the
Knudsen reactor volume after the plunger is raised. The pressure
inside the Knudsen reactor was measured with a pressure gauge
(MKS Baratron 627, 0.01-100 mTorr). The Knudsen reactor
was coupled to the MS via a variable diameter escape orifice
(hole diameters in mm: 10.0, 5.9, 3.5, 2.0, 1.2, and 0.7). In the
present experiments, we used the 10.0, 5.9, and 3.5 mm orifices.
Before entering the MS, the molecular beam was modulated
by a punched wheel chopper at 244 Hz, permitting phase-
sensitive detection. The molecular beam was ionized with a
cross-beam rhenium ion source. Ions were mass selected with
a quadrupole mass analyzer and detected with a secondary
electron multiplier. Data were accumulated by an ion-counting
preamplifier and were finally stored in a PC.
4
x
cj ) 1.46 × 10 T/M
(iii)
where T is the absolute temperature (298 K) and M is the molar
mass (in g). The appropriate temperature and mass dependent
expressions for calculating kesc are given in Table 1. They hold
for HNO3 (63 amu) as soon as the walls of the Knudsen reactor
are passivated with HNO3.
The flows of the reactant gases through the mixing line were
controlled by actively regulating the pressure at 0.5 Torr. Gases
entered the Knudsen reactor either by a second needle valve or
via a pulsed valve. All needle valves were constructed of Teflon
and were heated to 313 K to eliminate the influence of
temperature fluctuations. The needle valves were used for
continuous-flow steady-state experiments, the pulsed valve was
used for experiments in which a dosed pulse of gas was
introduced into the Knudsen reactor (pulse duration ∼20 ms).
The MS signal is proportional to the molecular flow through
the Knudsen reactor. The parameter describing the loss of

3098 J. Phys. Chem. A, Vol. 105, No. 13, 2001
Hanisch and Crowley
2.2. Reactant Gas Preparation. HNO3 was prepared by
dissolving 80 g of KNO3 (Aldrich) in 150 mL of concentrated
H2SO4 (95-97%; Aldrich) at 273 K. The HNO3 vapor was
transferred to a trap at liquid nitrogen temperature and finally
stored as anhydrous HNO3 liquid in a blackened bulb at 233
K. Gas-phase samples of HNO3 with N2 (mixing ratios of 0.1,
0.04, and 0.005) were prepared using standard manometric
techniques and stored in 2 L blackened glass vessels. Optical
absorption measurements performed on these samples revealed
the presence of NO2 impurity at <2% level. DNO3 was prepared
using D2SO4 (98% in D2O; Aldrich) instead of H2SO4.
2.3. Dust Sample Characterization and Preparation. Al2O3
(R-Al2O3) and CaCO3 were purchased from Sigma-Aldrich. For
Al2O3, we chose four disjunctive size classes (d < 10 µm, 10
µm < d < 44 µm, 44 µm < d < 74 µm, and 74 µm < d < 149
µm). CaCO3, ATD (“Arizona test dust”; Powder Technology,
Inc., MN, USA), and SDCV (Saharan dust from Cape Verde
islands, provided by L. Gomes) particles were smaller than 10
µm. In addition, we carried out some experiments on polished
and unpolished CaCO3 (104) and R-Al2O3 (112h0) single crystals
(purchased from MaTecK, Ju¨lich, Germany).
The mineral dust samples were prepared by mixing 30, 60,
or 100 mg of dust into a paste with water or ethanol and
dispersing on a round optical glass flat (surface area As ) 1.5
or 4.8 cm2), which was then placed in the sample compartment.
Experiments using H2O for dispersing the mineral dust gave
the same results as the ones using ethanol. Ethanol was used
for the majority of the experiments because it dries much faster.
The bulk density of the sample was determined from its weight
and height, which was measured using calibrated feeler gauges.
The CaCO3 and Al2O3 single crystals had a size of 10 × 10
× 0.5 and 10 × 10 × 1 mm3, respectively. Dust samples and
single crystals were usually heated to 363 K for 5 h under
vacuum (p ≈ 2 × 10-6 Torr) and held under vacuum overnight
in the Knudsen reactor prior to the experiment. BET surface
areas of the mineral dust were determined with a Quantachrome
Autosorb 6B. Scanning electron microscopy of unsputtered
surfaces was carried out with a LEO Gemini 1530 operated at
1.5 kV.
2.4. Measurement Procedure and Data Analysis. Uptake
coefficients, γ, were determined in steady-state, continuous-
flow experiments at room temperature (298 K). To make sure
that molecule-surface interactions dominate over molecule-
molecule interactions, Knudsen experiments must be carried out
in the Knudsen pressure regime or in the molecular flow regime.
The Knudsen regime requires the mean free path to be similar
to the dimensions of the reactor; at lower pressures, molecular
flow dominates. In the present experiments, a total Knudsen
reactor pressure of p ) 0.4 mTorr was used, which is in the
molecular flow regime. The HNO3 concentration was deter-
mined from the HNO3 mixing ratio, which was checked with
UV spectroscopy, and the Knudsen reactor pressure. The HNO3
concentration was varied between 1.3 × 1012, 5.2 × 1011, and
6.5 × 1010 cm-3. The corresponding HNO3 flows through the
Knudsen reactor into the MS were between 1.8 × 1015 and 2.5
× 1013 s-1 depending on the size of the escape orifice.
The uptake probability γ is defined as the ratio of the number
of collisions with the surface that lead to removal from the gas
phase to the total number of collisions with the surface. This is
mathematically equivalent to
Figure 2. The uptake of HNO3 onto ≈60 mg of Al2O3. Upper panel:
small grain size (d < 10 µm). Lower panel: large grain size (44 µm <
d < 74 µm). The dashed lines indicate the size and shape of HNO3
uptake in the absence of a dust sample, which has already been
subtracted for the experimental trace. Details of the experimental
conditions are found in Table 2, experiments A23 and A25, respectively.
reactive surface area of the sample As (see Table 1) and is
calculated from
cjAs
4V
ω )
(v)
The unimolecular rate constant kuni is defined for a pseudo-
first-order removal of the reactant gas from the gas phase to
the reactive surface, which depends on the number of available
surface sites for the heterogeneous interaction. kuni is obtained
by measuring the relative height of the signal before and during
exposure to the dust surface:
S0
kuni
)
- 1 kesc
(vi)
(
)
S(t)
where S0 is the MS signal of the reactant gas prior to the lifting
of the plunger and S(t) is the time dependent signal over the
whole duration of the experiment. Note that S(t) is corrected
for the small uptake of HNO3 observed when no dust was
present in the sample compartment of the Knudsen reactor (see
Figure 2).
The uptake coefficient is thus given by
kesc S0
γ )
- 1
(vii)
(
)
kuni
ω
S(t)
γ )
(iv)
ω
For the determination of γ, HNO3 was monitored at its strongest
fragment at m/e 46 (NO2+). The integration time for the signal
where ω is the collision frequency of a molecule with the

Heterogeneous Reactivity of Gaseous Nitric Acid
J. Phys. Chem. A, Vol. 105, No. 13, 2001 3099
TABLE 2: Summary of Uptake Experiments with HNO3
onto Al2O3
was 0.8 s with a duty cycle of 1.7 s. In some experiments, HNO3
was also monitored at m/e 30 (NO+) and m/e 63 (HNO3+) in
order to make sure that the m/e 46 signal was representative of
the HNO3 concentration and that there were no contributions
from NO2. The ratios m/e 63 to m/e 46 and m/e 63 to m/e 30
remained constant during the exposure, showing that HNO3 does
not decompose to NO2 on the reactor walls or react with the
dust sample to form significant amounts of gaseous NO, NO2,
or HONO. In the experiments on CaCO3, H2O was monitored
on m/e 17 (OH+) or m/e 18 (H2O+) and CO2 on m/e 44 (CO2+).
2.5. Error Analysis. Three quantities in eq vii contributed
to the experimental error of γ: kesc, ω, and S. kesc was determined
with an accuracy of 4.3% for the 10 mm escape orifice and
4.5% for the 5.9 mm escape orifice (4.4% on average). The
error of ω was mainly determined by the uncertainty in the
Knudsen reactor volume which was estimated to be 6.7%. The
sample surface area As, corresponding to the surface area of
the glass flat covered with dust or to the surface area of the
single crystal, did not significantly contribute to the error of ω
because it could be measured very precisely. Finally, the
uncertainty of the MS signal (S) after opening the plunger was
determined by the random noise and came to maximal 2.5%
for the low flow experiments. The three errors add up to a total
error of 8.4% for the resulting γ. However, in some cases, the
reproducibility of the measured uptake coefficient was consider-
ably worse than this, indicating that systematic errors, probably
related to surface preparation and presentation, are the dominant
source of error. We therefore quote an error of ( 25%, based
on the maximum observed fluctuations from the mean value of
γ that were observed in the course of these studies. Later we
consider potential errors associated with the usage of the
geometric surface area to calculate the uptake coefficient rather
than a surface area that takes into account the internal surface
of the sample.
geometric sample Ø escape
surface area mass
orifice
(mm)
[HNO3]
(cm-3
γinit
expt graina
(cm2)
(mg)
)
× 10-2
A1
1
2
3
4
1
1
2
3
4
1
2
3
4
1
1
2
3
4
1
4
4
4
1
2
3
4
4.8
4.8
4.8
4.8
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
100
100
110
120
100
100
100
100
100
60
60
60
60
100
60
60
60
60
60
100
60
30
60
60
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
5.9
10
10
10
10
10
10
10
10
10
1.3 × 1012
1.3 × 1012
1.3 × 1012
1.3 × 1012
1.3 × 1012
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
6.5 × 1010
1.3 × 1012
1.3 × 1012
1.3 × 1012
1.3 × 1012
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
6.5 × 1010
6.5 × 1010
6.5 × 1010
6.5 × 1010
4.4
2.9
4.0
4.1
5.5
4.6
2.9
5.3
4.4
5.3
3.1
4.1
4.1
9.7
6.5
5.9
7.1
A2
A3
A4
A5
A6
A7
A8
A9
A10
A11
A12
A13
A14
A15
A16
A17
A18
A19
A20
A21
A22
A23
A24
A25
A26
5.7
11.9
10.2
11.8
10.4
12.6
8.2
10
10
10
60
60
13.2
13.4
a Grain 1, (d < 10 µm), BET surface area ) 1.5 m2/g; Grain 2, (10
µm < d < 44 µm), BET surface area ) 0.5 m2/g; Grain 3, (44 µm <
d < 74 µm), BET surface area ) 0.3 m2/g; Grain 4, (74 µm < d <
149 µm), BET surface area ) 0.1 m2/g.
with d < 10 µm (upper panel) takes place in two phases: In
the first four minutes of exposure (6 min < t < 10 min), the
HNO3 signal quickly recovers to ≈75% of its original value,
whereas at t > ∼10 min, the recovery is much slower. In the
case of HNO3 onto larger grain-sized Al2O3 (44 µm < d < 74
µm, lower panel), the recovery of the MS signal proceeds
continuously and has almost reached the steady-state level after
37 min of exposure. These differences can be understood in
terms of external surface saturation and diffusion into the bulk.
We examined this aspect of the uptake process in more detail
by taking electron micrographs of both Al2O3 (d < 10 µm) and
Al2O3 (44 µm < d < 74 µm) samples that had been prepared
in an identical manner to those used in the Knudsen reactor
experiments. These micrographs are displayed in Figure 3a (d
< 10 µm) and Figure 3b (44 µm < d < 74 µm). They show
the different surface morphologies of the two grain sizes. In
addition, we carried out BET surface area measurements of each
dust sample. The measured Al2O3 BET surface areas are: (d <
10 µm) ) 1.5 m2/g, (10 µm < d < 44 µm) ) 0.5 m2/g, (44 µm
< d < 74 µm) ) 0.3 m2/g, and (74 µm < d < 149 µm) ) 0.1
m2/g. Clearly, and as expected, the total surface area is
significantly larger for the smaller particles (by a factor of ≈15),
explaining the larger capacity for this sample to remove HNO3
from the gas phase, as evidenced by the slower signal recovery.
The difference in the time dependence of the HNO3 uptake onto
different grain sizes of Al2O3, as exemplified in Figure 2, is
related to the difference in the total available surface area and
to differences in the rate of diffusion into the internal voids of
the dust sample. This is expected to take place on a longer time
scale than uptake on external surface sites and is faster for the
larger grain size.
3. Results and Discussion
The results are presented in separate sections according to
the substrate used. For each substrate, sample mass, geometric
surface area, and HNO3 flow were varied. For Al2O3, the grain
size was also systematically varied. In addition, experiments
were conducted using Al2O3 and CaCO3 single crystals. In all
cases, the initial rate of uptake of HNO3 onto the dust surface
was determined, which is taken as a measure of the reactivity
of fresh mineral dust surfaces. The role of adsorbed water on
the uptake process and on the formation of reaction products
was also examined.
3.1. Uptake of HNO3 onto Al2O3. In total, 26 experiments
were carried out using Al2O3 samples with different grain
sizes: d < 10 µm, 10 µm < d < 44 µm, 44 µm < d < 74 µm,
and 74 µm < d < 149 µm. The results are presented in Table
2, along with the experimental conditions in each case.
Typical raw data from an experiment to investigate the uptake
of HNO3 onto ≈ 60 mg of Al2O3 are shown in Figure 2. After
a steady flow of HNO3 has been established, the isolation
plunger is lifted (at t ) 6 min) and the surface is exposed to
the HNO3 flow. Because of uptake of HNO3 onto the Al2O3,
the number of molecules exiting through the escape orifice into
the MS decreases immediately. As the exposure time increases,
the HNO3 signal recovers, indicating a decrease in the number
of available sites for reaction, resulting in an apparent reduction
in the uptake coefficient. At t ) 40 and 37 min respectively,
the plunger is lowered and the HNO3 signal returns to its initial
steady-state value.
Comparison of the uptake profiles in Figure 2 reveals
differences in the uptake process. The HNO3 uptake onto Al2O3
From our experiments, it is not possible to draw quantitative
conclusions concerning the capacity of mineral dust to react

3100 J. Phys. Chem. A, Vol. 105, No. 13, 2001
Hanisch and Crowley
Figure 4. Uptake of HNO3 onto Al2O3: dependence of the uptake
coefficients on the flow of HNO3. The open symbols represent data
obtained with the 5.9 mm escape orifice; the filled symbols represent
data obtained with the 10 mm escape orifice. Insert: initial uptake
coefficient (γobs) vs sample mass (expts A20-A22). The initial HNO3
concentration was held constant at 5.2 × 1011 cm-3; the Al2O3 sample
was the 74-149 µm grain size. The solid curves were calculated
according to the pore diffusion model as described in the text with dh
) 112 µm, ABET ) 0.1 m2/g, Ft ) 3.97 g/cm3, and Fb ) 1.8 g/cm3.
Figure 3. Electron micrographs of the Al2O3 samples prepared as usual
for the Knudsen reactor experiments. (A) grain size d < 10 µm and
(B) grain size 44 µm < d < 74 µm. In both figures, the full width of
the micrograph is 270 µm.
Curve A, τ ) 2, γpd ) 5.5 × 10-3; Curve B, τ ) 20, γpd ) 5.5 × 10-3
;
Curve C, τ ) 2, γpd ) 2.4 × 10-3; and Curve D, τ ) 20, γpd ) 2.4 ×
10-3
.
with HNO3 under atmospheric conditions. The use of bulk
samples in the experiments means that the dust presentation
and interaction with the gas phase is different from that which
takes place in the atmosphere. However, one parameter of the
heterogeneous uptake can accurately be determined: the initial
uptake probability, i.e., the uptake probability immediately after
the lifting of the plunger. Therefore, the further discussion of
our results focuses on this parameter:
increase in kuni with decreasing grain size because of the
increasing internal surface area. kuni, however, was completely
unaffected by the variation of grain size. This important aspect
of the uptake is discussed in detail in section 3.6. The use of
different geometric surface areas (i.e., 4.8 cm2 rather than the
standard 1.5 cm2, experiments A1-A4) yielded values for γinit
which are in the same range as the values of the experiments
A6-A13. Experiments A1-A4 were performed at a higher
HNO3 flow than experiments A6-A13, but experiment A5
shows that in this regime within the limits of uncertainty the
flow has no influence on the initial uptake (compare to A8).
A double series of experiments (A6-A13) was performed
with 60 and 100 mg and a third series (A20-A22) with 30, 60,
and 100 mg of Al2O3 spread on 1.5 cm2 glass flats. γinit did not
show any dependency on the sample mass, i.e., γinit did not
depend on the thickness of the bulk sample. Only the uptake
capacity of the Al2O3 bulk increases with increasing sample
mass.
In Figure 4, we plot the initial uptake coefficient vs the flow
of HNO3. The HNO3 flow was varied by using different mixing
ratios in N2 in the storage bulb and different escape orifices.
Values of γ obtained in experiments with the 5.9 mm escape
orifice are lower than the ones obtained with the 10 mm escape
orifice, and both sets show an increase in γ with decreasing
HNO3 flow. This can be explained by Knudsen reactor wall
effects. After a steady concentration of HNO3 has been
established, HNO3 adsorbed on the walls is in equilibrium with
the gaseous HNO3 in the Knudsen reactor. As soon as the
plunger is lifted this equilibrium is perturbed because of fast
uptake of HNO3 onto the dust surface, and the reactor walls
act as an additional HNO3 source. That HNO3 is lost, indeed,
from the reactor walls can clearly be seen after lowering the
plunger, after which the HNO3 signal takes up to two minutes
kesc S0
γinit
)
- 1
(viii)
(
)
ω
S(topen
)
where topen is the time at which the sample was first exposed
by lifting the isolation plunger. Because the major goal of the
experiments using Al2O3 was to establish the experimental
procedure and to identify the correct surface area for calculating
the uptake coefficient, a series of diagnostic experiments were
carried out in which important experimental parameters such
as the sample mass and the geometric surface area were varied.
In six sets of experiments (Expts A1-A4, A6-A9, A10-
A13, A15-A18, A19 and A21, and A23-A26; Table 2) we
examined in detail the influence of the substrate’s grain size
on the initial HNO3 uptake, with the mass of substrate held
constant. γinit was found to vary within the limits of uncertainty
if the results for one middle range grain size, 10 µm < d < 44
µm, are excluded. We assume that this batch was somehow
contaminated.
The invariability of γinit with grain size has an important
consequence. First, the BET surface, which varies over 1 order
of magnitude (see footnotes to Table 2) is not the appropriate
reactive surface area for the calculation of γ. If diffusion of
HNO3 into the internal pores of the dust sample contributed
significantly to the initial uptake, we should have seen an

Heterogeneous Reactivity of Gaseous Nitric Acid
J. Phys. Chem. A, Vol. 105, No. 13, 2001 3101
TABLE 3: Summary of Uptake Experiments with HNO3
onto CaCO3
geometric sample
surface mass
expt area (cm2) (mg)
Ø escape
orifice
(mm)
sample
condition
[HNO3]
(cm-3
γinit
)
× 10-2
B1
4.8
4.8
4.8
1.5
1.5
1.5
1.5
1.5
1.5
1.5
3.1
3.1
3.1
3.1
4.8
1.5
1.5
1.5
1.5
1.5
1.5
1.5
100 unreacted, dry
100 reacted, dry
100 reacted, dry
100 unreacted, dry
100 reacted, dry
100 unreacted, dry
100 unreacted, damp
60 unreacted, damp
60 reacted, damp
60 unreacted, damp
100 unreacted, dry
100 unreacted, dry
100 unreacted, dry
100 reacted, dry
100 unreacted, dry
60 unreacted, damp
60 reacted, damp
60 reacted, damp
60 unreacted, dry
60 unreacted, dry
60 unreacted, dry
60 unreacted, dry
5.9
5.9
5.9
5.9
5.9
5.9
5.9
10
10
10
10
10
10
10
10
10
10
10
10
10
1.3 × 1012
1.3 × 1012
1.3 × 1012
1.3 × 1012
1.3 × 1012
5.2 × 1011
6.5 × 1010
1.3 × 1012
1.3 × 1012
5.0
0.9
1.6
5.1
2.1
5.2
9.9
7.3
7.3
B2
B3
B4
B5
B6
B7
B8
B9
B10
B11
B12
B13
B14
B15
B16
B17
B18
B19
B20
B21
B22
5.2 × 1011 13.7
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
5.2 × 1011
4.3
6.9
6.3
2.8
7.0
Figure 5. Uptake of HNO3 onto CaCO3: raw data (expt B16). Upon
lifting the plunger at t ≈ 6 min, some H2O is released from the dust
sample. After ≈17 min exposure, an extra flow (≈2.3 × 1016 s-1) of
H2O was added to the reactor (m/e 17). The HNO3 signal at m/e 46
has been shifted up by 1000 counts for clarity.
1.0 × 1011 18.3
6.5 × 1010 16.2
6.5 × 1010 15.1
6.5 × 1010
9.3
6.5 × 1010 11.6
6.5 × 1010 10.2
10
10
6.5 × 1010
7.8
to increase to its original steady-state value, showing that the
walls have been depassivated and again act as a sink for HNO3.
Degassing of HNO3 from the walls during the uptake process
distorts the results because the MS signal is enhanced with
respect to the ideal case without degassing. This effect means
that measured uptake coefficients are lower than the actual ones
and become more important with larger kuni because the
difference between S0 and S(topen) is larger. It also explains why
uptake coefficients measured with the 5.9 mm escape orifice
are smaller than the ones measured with the 10 mm escape
orifice, because the drop in signal for the small escape orifice
is larger because of a longer residence time in the reactor. For
this reason, uptake coefficients measured with the 10 mm escape
orifice at small HNO3 flows are closest to the real value because
kuni is minimized. The preferred uptake coefficient for HNO3
onto Al2O3 is thus an average of the results from experiments
A23, A25, and A26 and is (13 ( 3.3) × 10-2
.
3.2. Uptake of HNO3 onto CaCO3. A total of 22 experiments
with CaCO3 powder samples were carried out, and the results
are presented in Table 3 and in Figures 5 and 6. For the CaCO3
experiments, we observed a strong influence of surface water,
both in the uptake coefficients and in the formation of gas-
phase products. Here we adopt the arbitrary definition of “damp”
as applying to dust that has not been heated before the uptake
of HNO3 was measured and “dry” for dust samples that had
been heated in the standard manner to 363 K and kept in the
evacuated Knudsen reactor overnight.
3.2.1. Uptake Coefficient. The raw data from an experiment
to investigate the uptake of HNO3 onto CaCO3 is shown in
Figure 5. The presence of water on the CaCO3 sample is
indicated by the peak at m/e 17 immediately after lifting the
plunger at t ) 6 min. This short-lived signal originates from
surface-adsorbed water that had already been released from the
sample to the gas phase and trapped in the isolated volume under
the plunger. This water signal is missing in experiments where
the sample was heated. Initial uptake coefficients on fresh
CaCO3 samples were in the range of 0.05-0.18 and, within
experimental uncertainty, were independent of the geometric
surface area (expt B1 vs B4).
Figure 6. Uptake of HNO3 onto CaCO3: dependence of the uptake
coefficients on the flow of HNO3. The open symbols represent data
obtained with the 5.9 mm escape orifice; the filled symbols represent
data obtained with the 10 mm escape orifice.
be taken as a test of the reproducibility of the uptake measure-
ment: the values show reasonable agreement, though not within
the expected experimental error. Comparison of these γinit values
to the γinit values from experiments B10 and B16, respectively,
shows that under certain conditions the uptake coefficient can
be two or three times higher. The difference in the results can
be explained by the presence of water, because both the initial
uptake coefficient and the rate of surface saturation are highly
dependent on the water content of the CaCO3 sample, with a
higher uptake on damp CaCO3 (e.g., B16) and a smaller uptake
on dry CaCO3 (e.g., B19-B22).
For experiments on dry as well as on damp CaCO3 samples,
a decrease in the uptake coefficient with increasing flow and
decreasing escape orifice size was observed. In Figure 6, the
stars and the cross are results from experiments on damp CaCO3
and the squares are from experiments on dry CaCO3. It is
interesting to note that the impact of high HNO3 flows is bigger
for experiments on damp substrates because they have higher
uptake probabilities. In a single experiment, we investigated the
Uptake coefficients as a function of flow are plotted in Figure
6. Two series of experiments (B11-B13 and B19-B22) can

3102 J. Phys. Chem. A, Vol. 105, No. 13, 2001
Hanisch and Crowley
effect of adding water during the uptake process in order to see
if a chemically aged surface (after exposure to HNO3) can
thereby be reactivated. The addition of a flow of H2O (2.3 ×
1016 s-1) at t ≈ 22.5 min has no influence on the rate of decline
of the HNO3 uptake coefficient (Figure 5), suggesting that this
is not the case, at least on the time scales and the H2O
concentrations available in this experimental setup. The final
result for dry CaCO3 is γ ) (9.7 ( 2.4) × 10-2, which is the
value we determined at low HNO3 concentrations.
Several experiments were also performed on prereacted
CaCO3. After saturation of the sample with HNO3 (i.e., the
sample was exposed to HNO3 until the uptake coefficient had
decreased to almost zero), the HNO3 flow was shut off and the
sample was left in the Knudsen reactor, which was evacuated
overnight. A total of 24 h after the first exposure, the experiment
was repeated. We found that the initial reactivity is partly or
almost fully restored, depending on sample heating. Experiments
B9 and B17-B18 were performed on damp CaCO3 that already
had reacted with HNO3 the day before. Uptake coefficients
almost reached the value of the unreacted sample. However,
the time required for the uptake to reach saturation is strongly
reduced, indicating a decrease in bulk reactivity. Apparently,
external surface sites that had been saturated are again available
for reaction after 24 h. At the same time, the number of reactive
sites inside the bulk of the sample has decreased. Experiments
B2-B3, B5, and B14 were performed on dry CaCO3. Uptake
coefficients of HNO3 on prereacted dry CaCO3 were ca. one-
third of the original value. An exception is experiment B2 which
was carried out only 1 h after experiment B1 and where the
initial uptake coefficient is even lower. In addition, all experi-
ments on dry, prereacted CaCO3 showed that less HNO3 was
needed to deactivate the sample than in experiments on damp,
prereacted CaCO3.
We interpret our results in terms of surface reactivation by
nitrate recrystallization. Several studies have defined the
important role of surface adsorbed water in the reaction of HNO3
with NaCl surfaces, where a water mediated recrystallization
process is responsible for regeneration of reactive surface
sites.25-28 We transfer this model to CaCO3, although CaCO3
is hydrophobic, because water is probably bound primarily to
surface defects such as edges and steps, and we extend the model
also to the internal surface. Our results with single crystals (see
later) suggest that the relatively smooth crystal surfaces
themselves do not significantly contribute to the uptake process.
3.2.2. Product Formation. HNO3 is expected to react with
CaCO3 according to
Figure 7. Uptake of DNO3 onto CaCO3. The MS signals of DNO3
and CO2 were put on an absolute basis by comparison with calibrated
flows of each gas.
of the m/e 44 curve with the m/e 20 curve in Figure 7 shows
that CO2 production peaks at about 13 min, whereas D2O (HDO)
production peaks at about 16 min. Experiments with HNO3
displayed a similar behavior with regard to the delayed formation
of H2O and CO2. The delayed formation of the products is an
interesting feature of HNO3 uptake onto CaCO3. We note that
the delayed detection of CO2 and H2O cannot be due to
differences in molecular flow characteristics compared to HNO3,
because both H2O and CO2 have smaller molecular masses and
therefore larger values for kesc
.
In a series of experiments, we were able to show that the
induction period is dependent on the HNO3 flow, with a shorter
induction period at higher HNO3 flows. This observation is in
qualitative agreement with results by Fenter et al.,29 who
observed a delayed production of CO2 and H2O in the reaction
of HNO3 with CaCO3 and also of HCl in the reaction of HNO3
with NaCl and KCl.30 The present experiments revealed that
the yields of CO2 and H2O were highly variable and depended
not only on the HNO3 flow but also on the water content of the
CaCO3 sample. The stoichiometry of reaction 1 was generally
not fulfilled, indicating that reaction 1 does not properly describe
the more complex processes occurring on the dust sample.
By calibrating the MS for CO2 and H2O and integrating the
signals over the whole reaction time, we could show that on
average 2.7 HNO3 molecules were adsorbed for every CO2
molecule released for “damp” dust sample. For “dry” samples,
3.4 HNO3 molecules were consumed per CO2 molecule released.
Fenter et al.29 also observed a reduced yield of CO2 in the
reaction of HNO3 on dried CaCO3 pellets, but they also showed
that CO2 could be released from the prereacted surface by adding
a flow of water. They hypothesized that either physisorbed
HNO3 remained inactive in the absence of water or that HNO3
formed an intermediate product with CaCO3 and that CO2
release was inhibited in the absence of surface water.
The ratio of H2O produced to HNO3 taken up was lower than
predicted by reaction 1 by about a factor of 4 or more. Our
observations for H2O are in qualitative agreement with findings
of Fenter et al.,29 who attribute the missing water to its
interaction with the CaCO3 substrate. In separate experiments,
in which the uptake of water on CaCO3 was investigated, we
were able to show that, after stopping the H2O flow, as much
H2O degassed from the sample as had been adsorbed previously,
i.e., the uptake process is reversible. These results suggest that
it is not CaCO3 that retains the water but the Ca(NO3)2 formed
in reaction 1, which is much more hydroscopic than CaCO3.
2HNO3(g) + CaCO3(s) f
Ca(NO3)2(s) + CO2(g) + H2O(g) (1)
The formation of gas-phase products in the reaction of HNO3
with CaCO3 was investigated in experiments performed at high
HNO3 flows (1.8 × 1015 s-1) and with maximal CaCO3 sample
surface area (4.8 cm2). Under these conditions, both CO2 and
H2O could be observed by mass spectrometry.
To better differentiate between water produced in the reaction
of nitric acid with CaCO3 and water present in the backgound,
we performed an additional series of experiments with DNO3
instead of HNO3. A typical experiment is shown in Figure 7.
DNO3 was monitored on m/e 64 (not shown) and m/e 46, and
D2O, on m/e 20. As expected, the reaction led to the formation
of D2O, though HDO was also detected at a similar concentra-
tion. D2O (and HDO) production was as variable as H2O
production in the reaction of HNO3 with CaCO3. Comparison

Heterogeneous Reactivity of Gaseous Nitric Acid
J. Phys. Chem. A, Vol. 105, No. 13, 2001 3103
TABLE 4: Summary of Uptake Experiments with HNO3
onto ATD
geometric
surface area
(cm2)
sample Ø escape
mass
(mg)
orifice
(mm)
[HNO3]
(cm-3
γinit
expt
)
× 10-2
C1
C2
C3
C4
C5
C6
C7
C8
4.8
1.5
1.5
1.5
1.5
1.5
1.5
1.5
100
100
100
100
60
60
60
5.9
5.9
5.9
5.9
10
10
10
1.3 × 1012
1.3 × 1012
5.6 × 1011
6.5 × 1010
1.3 × 1012
5.6 × 1011
6.5 × 1010
6.5 × 1010
2.1
2.8
3.4
5.0
4.0
6.6
6.8
4.7
60
10
TABLE 5: Summary of Uptake Experiments with HNO3
onto SDCV
geometric
surface area
(cm2)
sample Ø escape
mass
(mg)
orifice
(mm)
[HNO3]
(cm-3
γinit
expt
)
× 10-2
Figure 8. Uptake of HNO3 onto an unpolished CaCO3 single crystal.
A, raw data; B, correction curve for uptake to the sample chamber; C,
corrected data, offset by 2000 counts for clarity.
D1
D2
D3
D4
D5
D6
D7
D8
D9
4.8
1.5
1.5
1.5
1.5
1.5
1.5
1.5
100
100
100
100
100
60
5.9
5.9
5.9
5.9
5.9
10
1.3 × 1012
1.3 × 1012
5.6 × 1011
5.6 × 1011
6.5 × 1010
1.3 × 1012
5.6 × 1011
6.5 × 1010
6.5 × 1010
6.5 × 1010
6.5 × 1010
6.1
7.2
7.1
11.7
9.2
10.6
12.4
12.1
11.8
10.2
10.7
We also attempted to gain insight into the reaction mechanism
by taking electron micrographs of the CaCO3 sample both before
and after exposure to HNO3. No change in surface morphology
was observed, and the definition of the crystal edges of the
CaCO3 sample was not measurably modified by treatment with
HNO3. Similar results were obtained for the Al2O3, ATD, and
SDCV samples, and we conclude that the amounts of HNO3
sufficient for saturation in our experiments were too low to
significantly alter the surface morphology.
60
10
60
60
10
10
1.5 (unheated)
D10 1.5 (unheated)
D11 1.5 (hot)
60
60
10
10
listed in Table 5. Uptake coefficients were found to be
independent of the geometric surface area of the sample and
lie between 6 × 10-2 and 12 × 10-2. A slight dependence of
the uptake coefficient on the HNO3 flow was found, and a value
of γ ) (11 ( 3) × 10-2, the uptake coefficient for low HNO3
concentrations, is the preferred result.
3.3. Single Crystals of Al2O3 and CaCO3. In comparison
with dust sample surfaces, surfaces of single crystals have very
few or no defect sites. We conducted three experiments each
on polished and unpolished Al2O3 and CaCO3 single crystals,
using the 3.5 mm escape orifice and an HNO3 flow of 2.6 ×
1013 s-1. For Al2O3, we determined γ ) (1.6 ( 1.4) × 10-3 on
unpolished surfaces and γ ) (6.7 ( 1.9) × 10-4 on polished
surfaces. For CaCO3, we determined γ ) (1.75 ( 0.39) × 10-3
on unpolished surfaces (Figure 8) and γ ) (9.6 ( 1.2) × 10-4
on polished surfaces, where the quoted errors have been chosen
to reflect the reproducibility based on the total of three
experiments. Uptake coefficients on single crystals are smaller
than uptake coefficients on dust samples by about 2 orders of
magnitude. In addition, uptake coefficients are smaller on
polished surfaces than on unpolished surfaces, and the correction
to S(t) for uptake of HNO3 to the walls of the sample
compartment was of a similar size as the uptake of HNO3 onto
the single-crystal surface (Figure 8). Both results underline the
importance of defect sites for the reactivity of the sample. It is
not clear whether the uptake is reduced because there are less
defect sites for direct HNO3 uptake, or because there is less
water bound to defect sites. Bearing in mind the effects described
in the previous section we believe that the absence of water is
the reason for the reduced uptake.
3.4. Uptake of HNO3 onto Arizona Dust. A total of 8
experiments with ATD were carried out, and the results are listed
in Table 4. Uptake coefficients were in the range between 2 ×
10-2 and 7 × 10-2 and found to be independent of the geometric
surface area of the sample and of the HNO3 flow. A slight
tendency of increasing uptake coefficients with decreasing
HNO3 flows was observed, with slightly higher γ values for
experiments with the bigger escape orifice. The final result is
γ ) (5.7 ( 1.5) × 10-2, which is the mean uptake coefficient
for low HNO3 concentrations.
SDCV contains a considerable amount of water. Heating of
SDCV inside the Knudsen reactor resulted in loss of 1.2 mg
water per g of dust. Interestingly, it did not make a difference
for the uptake coefficient if we heated the sample or not
(experiments D9 and D10). Also even on hot SDCV (353 K;
experiment D11), the uptake coefficient was as high as on dust
at 298 K. This is explained by dehydration curves of clay
minerals which show that substantial loss of water does not
occur at temperatures below 673 K (illite) or 773 K (kaolinite,
chlorite). For example, kaolinite loses 10% of its weight in the
form of water by heating between 773 and 823 K.31 The 1.2‰
weight loss after heating at 353 K as in the present experiments
cannot be considered to represent a substantial loss of water
and explains why uptake of HNO3 is not influenced by the
standard “drying” procedure.
Although no evidence for product formation from reactive
uptake was observed, the irreversibility of the uptake process
could be confirmed in the following experiment. After extended
exposure to HNO3, the dust sample was isolated from the
Knudsen reactor (plunger closed) and the HNO3 flow disabled.
After the signal at m/e 46 had decreased to about 5% of its
steady-state flow value, the plunger was lifted. The absence of
an increase in the HNO3 signal was taken to confirm that
desorption of physisorbed HNO3 from the surface was insig-
nificant.
3.6. Role of Internal Surface Area in Calculation of γinit
.
In section 3.1, we briefly addressed the role of pore diffusion
for calculation of the uptake coefficient. If pore diffusion is
significant on the time scale of the measurement of the initial
uptake coefficient, this parameter, defined in the following as
3.5. Uptake of HNO3 onto Saharan Dust. A total of 11
experiments with SDCV were carried out, and the results are
geom
γ
, should be corrected for the internal surface area. In this
init

3104 J. Phys. Chem. A, Vol. 105, No. 13, 2001
Hanisch and Crowley
TABLE 6: Comparison of Uncorrected and Corrected
Uptake Coefficients on Al2O3
applicable, the data should lay in the linear mass-dependent
region as defined by Underwood et al,36 where the relevant
surface area is the total external and internal surface area. As
average
grain size
(µm)
geom
already pointed out in section 3.1, the constancy of γ
with
init
geom
init
γ
γBET
γpd
varying grain size is a strong indication that the internal surface
of the sample is not of great significance for the HNO3 uptake.
If the pore diffusion correction were applicable to HNO3, the
values of γinit for the experiments with HNO3 on 30, 60, and
100 mg of Al2O3 (74 µm < dh < 149 µm; expts A9, A13, and
A20-A22) should display a linear dependence on sample mass;
the independence of γinit upon changing the sample mass
suggests this is not the case. In the insert to Figure 4, we plot
values of γinit obtained under identical conditions (initial HNO3
concentration, grain size, etc.), except that the sample mass was
varied. The solid lines are the result of a pore diffusion
calculation for uptake onto this substrate. Despite variation of
τ over a broad range (2-20), the model was not able to
reproduce the (lack of) dependence of the initial uptake
coefficient on sample mass.
The nonapplicability of the pore diffusion model is further
corroborated by the results for γBET and γpd which show a strong
dependency on grain size. In contrast, Caloz et al.37 were able
to align uptake coefficients for ClNO2 on KBr that varied over
2 orders of magnitude by applying the correction for pore
diffusion. In our study, the opposite is the case, and initially
consistent data are converted to uptake coefficients that vary
over a factor of 3.
Strictly speaking, the fact that we did not observe a
dependence of γ on grain size for Al2O3 does not prove the
assumption that the other substrates used in this study behave
in the same manner. However, together with the results for
HNO3 onto CaCO3 (Fenter et al.29) and for HNO3 onto NaCl
and KBr, our assertion that sample surface presentation has no
influence on the initial uptake of HNO3 appears justified.
In equations x and xi, the Thiele modulus φ is a measure of
the surface reaction rate to the porous diffusion rate. If diffusion
does not play a role with respect to surface reaction, φ must
become very large. As a consequence, η approaches zero and
eq x reduces to
2.4
27
59
12.6 × 10-2
(8.2 × 10-2
13.2 × 10-2
13.4 × 10-2
2.2 × 10-4
16 × 10-4
)
11 × 10-4
12 × 10-4
112
34 × 10-4
34 × 10-4
section, we compare the uncorrected (i.e., geometric) uptake
coefficients for Al2O3 (expts A23-A26) to the corrected ones,
taking into account the four particle size classes, and explain
why we think that no correction is needed for uptake experi-
ments with HNO3.
A simple correction method is to use the BET area of a
sample instead of the geometrical surface area:
Ageom
geom
init
γBET
)
γ
(ix)
ABET
where Ageom is the geometrical surface area of the sample (1.5
cm2), γigneiotm ) 13 × 10-2, and ABET is the BET surface area of
the sample, calculated for the sample mass of 60 mg using the
specific BET areas given in section 3.1 and Table 2. Calculated
values for γBET are presented in Table 6.
A more sophisticated correction method which takes into
consideration the relative rate of surface reaction and pore
diffusion effects has been presented by Keyser and co-
workers.32-34 This method was originally developed for het-
erogeneous reactions on vapor deposited ice surfaces but has
also been applied to reactions on granular surfaces. To take into
account the porosity of the Al2O3 grains bigger than 10 µm,
we calculated the correction for pore diffusion according to
Keyser et al.34
1
2
h
1
2
geom
init
γ
) γpdFbABETdh + η
-
(x)
(
(
))
dh
where η is the “effectiveness” factor, η ) (tanh φ)/φ, and φ is
the Thiele modulus whereby
geom
init
γ
) γpdFbABET0.5dh
(xii)
3Fb
h
1
2
pd
x
3τγ
φ )
-
(xi)
With this formula we calculated corrected γ values of 8 × 10-2
for grains with d < 10 µm and 1.4 × 10-2 for the other sizes.
The result for the smallest grains is expected to be the most
accurate because the porous nature of individual larger grains
(which themselves are aggregates of smaller particles, see Figure
3) precludes accurate estimation of dh.
We therefore conclude that neither the simple BET correction
nor the pore diffusion correction are appropriate for the uptake
experiments with HNO3. A correction for surface enlargement
due to the roughness of the surface (equation xii) may be
appropriate and would result in a reduction of the initial uptake
coefficients by a factor of about 2.
3.7. Comparison with Literature. Laboratory investigations
of the interaction of HNO3 with any of the surfaces used in the
present study are limited in number. For comparison with
literature data, we use the uncorrected initial uptake coefficients
based on the geometric surface area, which is generally the
reported quantity.
Our results for the uptake of HNO3 onto CaCO3 are in
excellent agreement with a previous Knudsen reactor investiga-
tion29 in which an uptake coefficient of γ ) (10 ( 2) × 10-2
was obtained. For damp CaCO3, the uptake coefficients from
(
)
2(Ft - Fb)
dh
The input parameters for the calculation of the pore diffusion-
corrected uptake coefficient γpd are the uncorrected uptake
geom
init
coefficient γ , the total height of the Al2O3 sample h (h ) hi
+ he, where hi and he are the internal and external thickness of
the substrate), the average grain size of the Al2O3 particles dh,
the true density of crystalline Al2O3 Ft () 3.97 g/cm3), the bulk
density of the Al2O3 sample Fb, and the tortuosity factor τ. The
sample heights were between 420 µm (for d < 10 µm) and 210
µm (for 74 µm < d < 149 µm), resulting in bulk densities of
between 0.9 and 1.8 g/cm3, respectively. On the basis of
inspection of the SEM micrographs, we determined dh ) 2.4
µm for grain class 1, and for the other classes, we took the
arithmetic mean of the respective upper and lower size limits.
The tortuosity factor was set to τ ) 2, following calculations
for HNO3 on NaCl grains.35 Our calculated values for γpd are
listed in Table 6.
Values of γBET and of γpd are 2 to 3 orders of magnitude
lower than uncorrected γ values and, with the exception of the
smallest grain size, are in good agreement with each other. This
agreement shows that, if the pore diffusion model were

Heterogeneous Reactivity of Gaseous Nitric Acid
J. Phys. Chem. A, Vol. 105, No. 13, 2001 3105
experiments B10 (γ ) 14 × 10-2) and B16 (γ ) 18 × 10-2
)
aerosol mass (or 75% of the calcite content) can be converted
to nitrate. This implies that 5 µg/m3 can remove a total of 0.5
µg HNO3 which is equivalent to ≈200 ppt at the earth’s surface.
This simple calculation is very crude in nature but indicates
that interaction with mineral dust may be an important loss
process for tropospheric HNO3 and shows why modeling studies
of mineral dust interactions with HNO3 have revealed an
important role for this process.
are also in excellent agreement with previous measurements29
on damp CaCO3 pellets, in which an uptake coefficient of γ )
(15 ( 3) × 10-2 was determined. Another Knudsen reactor
study for HNO3 onto CaCO336 found uptake coefficients between
2.3 × 10-3 and 4 × 10-4, depending on the sample mass. These
uptake coefficients are ca. 2 orders of magnitude lower than
the present values and those obtained by Fenter et al.29 The
reason for this discrepancy is unclear.
The reaction of HNO3 with a dry Na2CO3 surface has also
been investigated using a Knudsen reactor29 and an annular
reactor38 to obtain uptake coefficients of 7.6 × 10-2 and 1.5 ×
10-2, respectively. Further evidence for the reaction of HNO3
on mineral particles is provided by the work of Mamane and
Gottlieb,39 who used electron microscopic and bulk analysis
techniques to observe and quantify nitrate formation on single
particles that were exposed to 0.04 ppm HNO3 under conditions
similar to ambient (70% humidity at room temperature). They
found that ≈30 mg nitrate per g of aerosol were formed under
favorable conditions. Although the data on the reactivity of
HNO3 to mineral oxide and mineral dust surfaces is still very
limited, observations of gas uptake and nitrate formation provide
strong evidence for an efficient reactive uptake process.
3.8. Atmospheric Implications. The atmospheric implica-
tions of the present results are assessed by taking the Saharan
dust sample used in the laboratory to be mineralogically
representative of atmospheric dust aerosol. The mineralogical
composition of SDCV that we used has been described in the
literature (e.g., see ref 40) and closely simulates atmospheric
particles of crustal origin.41 The clay fraction (<2 µm) of dust
from Cape Verde shows a kaolinite-illite-chlorite assemblage
which is typical for central Saharan dust, though the predomi-
nance of illite and kaolinite is characteristic of dusts in other
regions of the world.
We note that our experimental set up does not allow one
potentially important parameter, the humidity of the sample, to
be varied over an appropriate range. We have however shown
that the reactivity of the Saharan dust sample is not influenced
by strong heating and that the reactivity of CaCO3 is slightly
increased by the presence of H2O vapor. The fact that large
uptake coefficients were obtained for the “dry” conditions of
the Knudsen reactor experiments, where the relative humidity
is orders of magnitude less than that observed anywhere in the
lower troposphere, suggests that an uptake coefficient of 0.1
(taken from the data on Saharan dust) may be considered a
conservative estimate of the true uptake coefficient. The greatest
effect of humidity is expected to be related to the capacity of
the dust to remove HNO3.
4. Conclusions
We have determined the uptake coefficient for the reaction
of HNO3 with authentic mineral dust samples for the first time.
A value of γ ) 11 × 10-2 was determined for SDCV at 298
K. The result for ATD was γ ) 6 × 10-2. Measurements of
the uptake coefficient of HNO3 on Al2O3 (γ ) 13 × 10-2
)
enabled us to show that the geometrical surface area of the dust
sample is most appropriate for calculation of the uptake
coefficients in this study. The uptake coefficients for HNO3 on
dry CaCO3 and on damp CaCO3 are given by γ ) 10 × 10-2
and 18 × 10-2, respectively, and are in good agreement with
previous results.29 The reaction was found to be sensitive to
the amount of surface-adsorbed water. Experiments with HNO3
on single crystals (γ ≈ 10-3) showed that the lack of surface
defects results in a reduction of the uptake coefficient. Our
results indicate a generally very high reactivity of HNO3 toward
dust samples, highlighting the potentially important role of
mineral dust in modifying the distribution of nitrate between
the gaseous and particulate phases.
Acknowledgment. We gratefully acknowledge L. Gomes
for supplying the Cape Verde loess sample and D. Levi (Du
Pont) for provision of the FEP 121-A Teflon suspension used
to coat the Knudsen reactor. Thanks are due to Cedric du Fresne
and Joachim Huth for conducting the BET and the SEM
measurements, respectively. We acknowledge partial funding
from the BMBF (AFS-07AF210B8) and the German-Israeli
Foundation for Scientific Research and Development (I156-
303.06/97).
References and Notes
(1) Tegen, I.; Fung, I. J. Geophys. Res. 1994, 99, 22897.
(2) Carmichael, G. R.; Zhang, Y.; Chen, L.-L.; Hong, M.-S.; Ueda, H.
Atmos. EnViron. 1996, 30, 2407.
(3) Li, X.; Maring, H.; Savoie, D.; Voss, K.; Prospero, J. M. Nature
1996, 380, 416.
(4) Prospero, J. M. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 3396.
(5) Savoie, D. L.; Prospero, J. M. Geophys. Res. Lett. 1982, 9, 1207.
(6) Duce, R. A.; Unni, C. K.; Ray, B. J.; Prospero, J. M.; Merrill, J. T.
Science 1980, 209, 1522.
The rate of removal of HNO3 by uptake onto mineral dust
can be approximated in a simple model, which assumes spherical
geometry for particles and thus underestimates the surface area
density. We assume that the lifetime τ for removal of HNO3
by dust is given by
(7) Chester, R. Sci. Geol. Mem. 1990, 88, 23.
(8) Gomes, L.; Gillette, D. A. Atmos. EnViron. 1993, 27A, 2539.
(9) Zhang, Y.; Sunwoo, Y.; Kotamarthi, V.; Carmichael, G. R. J. Appl.
Met. 1994, 33, 813.
(10) Dentener, F. J.; Carmichael, G. R.; Zhang, Y.; Lelieveld, J.; Crutzen,
P. J. J. Geophys. Res. 1996, 101, 22869.
(11) Chatfield, R. B. Geophys. Res. Lett. 1994, 21, 2705.
(12) Hauglustaine, D. A.; Ridley, B. A.; Solomon, S.; Hess, P. G.;
Madronich, S. Geophys. Res. Lett. 1996, 23, 2609.
(13) Jacob, D. J.; Heikes, B. G.; Fan, S.-M.; Logan, J. A.; Mauzerall,
D. L.; Bradshaw, J. D.; Singh, H. B.; Gregory, G. L.; Talbot, R. W.; Blake,
D. R.; Sachse, G. W. J. Geophys. Res. 1996, 101, 24235.
(14) Ridley, B.; Walega, J.; Huebler, G.; Montzka, D.; Atlas, E.;
Hauglustaine, D.; Grahek, F.; Lind, J.; Campos, T.; Norton, R.; Greenberg,
J.; Schauffler, S.; Oltmans, S.; Whittlestone, S. J. Geophys. Res. 1998, 103,
8323.
4
γcjA
τ )
(xiii)
where A is the dust surface area density in cm2/cm3. If we
assume a conservatively low (i.e., background) dust loading of
5 µg/m3, we obtain A ≈ 10-7 cm2/cm3. Our measured uptake
coefficient γ ) 0.1 then leads to a HNO3 lifetime with respect
to processing by dust of ≈4 h. This is considerably shorter than
the gas phase photochemical lifetime due to reaction with OH
and photolysis which is ≈300 h in the lower troposphere. In
addition, field observations have shown that up to 10% of the
(15) Thakur, A. N.; Singh, H. B.; Mariani, P.; Chen, Y.; Wang, Y.;
Jacob, D. J.; Brasseur, G.; Mu¨ller, J.-F.; Lawrence, M. Atmos. EnViron.
1999, 33, 1403.
(16) Nishikawa, M.; Kanamori, S. Anal. Sci. 1991, 7, 1127.

3106 J. Phys. Chem. A, Vol. 105, No. 13, 2001
Hanisch and Crowley
(17) Okada, K.; Naruse, H.; Tanaka, T.; Nemoto, O.; Iwasaka, Y.; Wu,
P.-M.; Ono, A.; Duce, R. A.; Uematsu, M.; Merrill, J. T. Atmos. EnViron.
1990, 24A, 1369.
(18) Savoie, D. L.; Prospero, J. M. Nature 1989, 339, 685.
(19) Prospero, J. M.; Savoie, D. L. Nature 1989, 339, 687.
(20) Savoie, D. L.; Prospero, J. M.; Nees, R. T. Geophys. Res. Lett.
1987, 92, 933.
(21) Wolff, G. T. Atmos. EnViron. 1984, 18, 977.
(22) Mamane, Y.; Ganor, E.; Donagi, A. E. Water, Air, Soil Pollut. 1980,
14, 29.
(23) Loye¨-Pilot, M. D.; Martin, J. M.; Morelli, J. Nature 1986, 321,
427.
(30) Fenter, F. F.; Caloz, F.; Rossi, M. J. J. Phys. Chem. 1994, 98, 9801.
(31) Grim, R. E. Clay Mineralogy; McGraw-Hill: New York, 1953.
(32) Keyser, L. F.; Moore, S. B.; Leu, M. T. J. Phys. Chem. 1991, 95,
5496.
(33) Keyser, L. F.; Leu, M.-T. J. Colloid Interface Sci. 1993, 155, 137.
(34) Keyser, L. F.; Leu, M.-T.; Moore, S. B. J. Phys. Chem. 1993, 97,
2800.
(35) Leu, M. T.; Timonen, R. S.; Keyser, L. F.; Yung, Y. L. J. Phys.
Chem. 1995, 99, 13203.
(36) Underwood, G. M.; Li, P.; Usher, C. R.; Grassian, V. H. J. Phys.
Chem. A 2000, 104, 819.
(37) Caloz, F.; Seisel, S.; Fenter, F. F.; Rossi, M. J. J. Phys. Chem.
1998, 38, 7470.
(38) Msibi, I. M.; Shi, J. P.; Harrison, R. M. J. Atmos. Chem. 1993, 17,
339.
(39) Mamane, Y.; Gottlieb, J. J. Aerosol Sci. Technol. 1990, 21, 225.
(40) Coude´-Gaussen, G.; Rognon, P.; Le Coustumer, M. C. R. Acad.
Sci. 1994, 319, 1343.
(41) Desboeufs, K. V.; Losno, R.; Vimeux, F.; Cholbi, S. J. Geophys.
Res. 1999, 104, 21287.
(24) Rognon, P.; Coude´-Gaussen, G.; Revel, M.; Grousset, F. E.;
Pedemay, P. Sedimentology 1996, 43, 359.
(25) Beichert, P.; Finlayson-Pitts, B. J. J. Phys. Chem. 1996, 100, 15218.
(26) Allen, H. C.; Laux, J. M.; Vogt, R.; Finlayson-Pitts, B. J.;
Hemminger, J. C. J. Phys. Chem. 1996, 100, 6371.
(27) Ghosal, S.; Hemminger, J. C. J. Phys. Chem. A 1999, 103, 4777.
(28) Davies, J. A.; Cox, R. A. J. Phys. Chem. 1998, 102, 7631.
(29) Fenter, F. F.; Caloz, F.; Rossi, M. J. Atmos. EnViron. 1995, 29,
3365.