of 11
Supporting Information
Yeung et al. 10.1073/pnas.0902930106
SI Text
Testing for Sample Contamination in

47
Analysis of Field Samples.
To
rule out the possibility that the observed high

47
values result from
sample contamination by organic compounds in the stratospheric
samples, one of the samples [20000131(30)1001] was exposed to
water at 25 °C for 24 h to reach CO
2
–water isotope equilibrium. The
exchanged sample had a

47
value of 0.862

0.078‰, identical to
the cylinder CO
2
working standard, whose

47
value reflects CO
2
isotopic equilibrium at room temperature. In contrast, the presence
of contaminants would have resulted in a high

47
value even after
isotopic equilibrium was reached. Hence, no evidence for contam-
ination was observed in the high-latitude samples.
13
C/
12
C KIE Experiments and Modeling.
Our pulsed-photolysis
13
C/
12
C KIE experiments were designed to mimic stratospheric
collision energies. Nascent O(
1
D) atoms from N
2
O photolysis
are highly translationally excited (18.2 kcal mol

1
translational
energy) (1, 2). Therefore, an excess of helium (ultrahigh purity)
was used as a buffer gas to quench the translational energy of
O(
1
D) atoms efficiently while quenching the electronic energy
inefficiently (3–5). Hard-sphere Monte Carlo simulations of the
initial O(
1
D) collision cascade (
n

50,000; [see Nan and
Houston (6)] show that 81% of all O(
1
D)

CO
2
collisions occur
with

2 kcal mol

1
collision energy (E
coll
), and the average
collision energy

E
coll

1.6 kcal mol

1
, similar to the expected

E
coll

in the stratosphere (7, 8). We calculate that under our
typical reaction conditions (
1:100:3,000 N
2
18
O/CO
2
/He, 100-
Torr total pressure), the probability that any given CO
2
molecule
will undergo 2 isotope exchange reactions is

2.5
10

7
. Thus,
the reaction is likely near the single-exchange limit, i.e., no CO
2
molecule undergoes
1 isotope exchange reaction, and effects
due to reaction cycling are negligible.
We expect to observe subtle isotope effects due to differences
in the
18
O(
1
D)–CO
2
collision frequencies between different
isotopologues in the helium-buffer–gas experiments because the
collision-limited rate of reaction for O(
1
D)

CO
2
(9) implies that
the reaction probability for any given O(
1
D)–CO
2
collision is
near unity. For instance, more
18
O(
1
D)

12
C
16
O
2
collisions than
18
O(
1
D)

13
C
16
O
2
collisions occur because the average velocity
for
12
C
16
O
2
is higher than that for
13
C
16
O
2
at a given tempera-
ture. This reduces the rate at which reaction 2 occurs relative to
reaction 1 in the laboratory by
5%. The O(
1
D)–CO
2
collision
frequencies in these experiments scale directly with the collision
pair’s relative velocity, which depends inversely on the pair’s
reduced mass (

) according to Eq.
s1
:
k
2



18
O

44
CO
2

18
O

45
CO
2

k
1


eff
k
1
[s1]
k
represents the rate coefficient for the subscripted reaction
(Reaction
1
or Reaction
2
from the main text), and

represents
the reduced mass of the subscripted reactant pair. Analogous
relationships can be computed for each unique reactant pair.

47
is expected to decrease with increasing


18
O because the
collision frequency for
18
O(
1
D)

44
CO
2
is
5% larger than that
for
18
O(
1
D)

45
CO
2
. Such biases are expected to be small in the
atmosphere because of competing electronic quenching reac-
tions [e.g., O(
1
D)

N
2
3
O(
3
P)

N
2
] (10). These electronic
quenching reactions were minimized in our pulsed photolysis
experiments because helium was used as the buffer gas. Thus, a
13
C/
12
C KIE

k
2
/
k
1

1 in our laboratory experiment will appear
as an anticorrelation in a plot of

47
vs.


18
O. A
13
C/
12
C KIE
1.01 and an O(
1
D) source with

18
O


17
O
100‰ would be
required to generate the polar vortex

47
enrichment the polar
vortex

47
vs.

17
O correlation. On the

47
vs.


18
O plot, the
slope would be

0.0055 by using
k
2
/
k
1

1.01. The negative
experimental slope, however, rules out this KIE
1.
The results of the pulsed laser experiments are given in
Table
S2
. The best-fit slope of the

47
vs.


18
O relationship in these
experimental data is

0.0036

0.0008 (2

). By using a
13
C/
12
C
KIE

1.000, our kinetic model of the photochemical experiment
predicts a linear relationship between

47
and


18
O with slope
of

0.003 over the relevant range in


18
O. A
13
C/
12
C KIE

0.999 yields a

47
vs.


18
O slope of

0.0036 (shown in the main
text, Fig. 3). Thus, we assign the value
k
2
/
k
1

0.999 and estimate
the uncertainty in the KIE by varying the quantity
k
2
/
k
1
until the
modeled slope varies by

0.0008. The upper and lower limits of
k
2
/
k
1
(1.000 and 0.998, respectively) were then taken as the 2

uncertainty bounds. The estimate of the integrated effective O(
1
D)
isotope composition was also calculated from this model, although
it was modified to remove the collision frequency scaling.
Although this experiment was unable to isolate the
13
C/
12
C KIEs
in reactions
3a
and
3b
, the laboratory data are consistent with the
13
C/
12
C KIE in reaction
3a
dominating the observed

47
vs.


18
O
relationship. Reactions
3a
and
3b
are both important in the
stratosphere, with the branching ratio between them likely 9:1 at the
average stratospheric collision energy (3a/3b; see ref. 11) and thus
also at the

E
coll

in our experiments. Using this branching ratio, we
calculate that the experiment is approximately 3 times more sen-
sitive to a
13
C/
12
C KIE in reaction
3a
than in
3b
.A
13
C/
12
C KIE
0.993 in reaction
3b
, then, is also consistent with the laboratory

47
vs.


18
O slope. If the observed slope were arising exclusively from
a
13
C/
12
C KIE in reaction
3b
, it would become increasingly negative
with increasing initial

E
coll

because reaction
3b
is favored at higher
E
coll
(11, 12). Our high-collision energy (no buffer gas, initial

E
coll

13.6 kcal mol

1
) experiments also showed depletions in

47
as the
extent of photochemical isotope exchange increased, with no significant
(i.e., factor-of-3) change in

47
vs.


18
O. Other combinations of
13
C/
12
C KIEs in reactions
3a
and
3b
could also yield the laboratory

47
vs.


18
O slope, but the absence of significant temperature and colli-
sion-energy dependences in the laboratory data support our attribution
of the measured isotope effects to reaction
3a
.
A significant

47
change due to reactions other than
1
and
2
in the O(
1
D)–CO
2
photochemical system is unlikely. The KIEs
for the other reactions (e.g., the
17
O
12
C
18
O-producing channel,
which is absent in the pulsed photolysis experiments) would need
to be at least 1.01 to be of comparable importance. In addition,
the

47

18
O–

17
O covariation observed in the continuous-
irradiation experiments (which include the
17
O
12
C
18
O-producing
channel; see below) agreed well with our laboratory model,
which assumes statistical partitioning of isotope exchange prod-
ucts in the O(
1
D)

CO
2
reaction. Although not all of the relative
rates for the O(
1
D)

CO
2
isotope exchange reaction were mea-
sured directly, the agreement between the pulsed-photolysis and
continuous-irradiation experiments suggests that KIEs in the
other O(
1
D)

CO
2
reactions (e.g.,
18
O(
1
D)

16
O
12
C
17
O) are
described well by using statistical product partitioning in this
case. Direct measurements of the other KIEs would be required
to determine whether the statistical partitioning of isotope
exchange products applies rigorously for all O(
1
D)–CO
2
reactant
pairs. As such, we report a
13
C/
12
C KIE of 0.999

0.001 (2

),
which is within the uncertainty of a KIE of 1.000 expected for
stochastic isotope exchange.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
1of11
Continuous Irradiation Experiments and Modeling.
Continuous irra-
diation experiments on the O
2
/O
3
/CO
2
photochemical system
were performed at the University of California, Berkeley, to
examine whether the photochemistry occurring in this system
depletes

47
. A 2-L glass reaction bulb was filled with mixtures
of O
2
and CO
2
, both of known isotopic composition, between
100- and 150-Torr total pressure at 295 K, and the mixture was
irradiated between 0 and 48 h with a pen-ray mercury lamp
through a MgF
2
window. This scheme generates O(
1
D) through
narrow-band photolysis of O
2
at 185 nm (6
10
13
to 1.5
10
14
photons cm

2
s

1
) and subsequent photolysis of O
3
at 254 nm
(2
10
15
photons cm

2
s

1
). Residual O
2
was separated cryo-
genically from the mixture, and O
3
was catalytically decomposed on
hot nickel foil at 60 °C for 15 min before the CO
2
was purified for
isotopic analysis.
A commercially available software program, KINTECUS (13)
(Windows version 3.95, 2008), was used to simulate the kinetics
of the photochemical oxygen isotope exchange between O
2
and
CO
2
in the continuous-irradiation experiments. The species
modeled were O
2
,O
2
(
1

), O
2
(
1
), O(
3
P), O(
1
D), O
3
, and CO
2
and included both singly and doubly substituted
17
O- and
18
O-
isotopologues as well as
13
C-isotopologues of CO
2
, yielding
400
isotope-specific reactions. Rate coefficients for each reaction
corresponding to all common isotopologues were from Sander et
al. (9)
,
whereas photolysis rates (
J
-values) were calculated from
the reported irradiance of the Hg-pen ray lamp at 254 nm (14),
the geometry of the bulb, and the measured cross-sections of O
2
at 185 nm and O
3
at 254 nm (9). The only oxygen KIEs included
in the model are in the ozone recombination reaction (15) and
in O

O
2
isotope exchange (16, 17). Statistical scrambling of
oxygen was assumed to occur in the O(
1
D)

CO
2
isotope ex-
change reaction (18, 19), and scenarios with hypothetical
13
C/
12
C
KIEs of 1.000, 0.999, and 0.998 were considered. The model was
initialized with the experimental partial pressures, temperature,
and starting isotopic compositions of O
2
and CO
2
.
A comparison of the laboratory and modeling results from the
continuous irradiation experiments is shown in
Fig. S2
.We
observed a depletion in

47
with increasing

17
O (a proxy for the
extent of photochemical isotope exchange), in agreement with
the results from our pulsed photolysis experiments and within
the uncertainty bounds we reported for the
13
C/
12
C KIE (0.999

0.001). There is a noticeable disagreement, however, at the
highest extent of reaction (48-h irradiation time). Some of this
error could be attributed to our assumptions about the initial
distribution of CO
2
isotopologues in the model: The initial

47
0 was modeled as an excess of
16
O
13
C
18
O exclusively, although
contributions from
17
O
12
C
18
O and
13
C
17
O
2
may be nontrivial. In
addition, trace water in the reaction bulb (in the gas phase or
adsorbed onto the glass) could catalyze CO
2
–water isotope
exchange reactions and drive the distribution of CO
2
isotopo-
logues toward equilibrium, i.e.,

47

0.86‰ at 298 K. Both of
these uncertainties would be exacerbated as the irradiation time
increases, leading to a marked deviation from the modeled result
at longer irradiation times. Last, the effects of the O
3
decom-
position step on

47
have not been fully elucidated; previous
reports have documented some oxygen isotope exchange when
O
3
is decomposed on hot nickel (10, 20), perhaps with the nickel
oxide layer or surface-adsorbed water, so the postirradiation O
3
decomposition step during CO
2
purification may drive

47
toward its isotopic equilibrium value (

47
0.8‰ at 333 K; see
ref. 21) at the nickel’s temperature. Still, the continuous-
irradiation experiments are consistent with the results of the
pulsed-photolysis experiments. They show that the O(
1
D)

CO
2
isotope exchange reaction drives the CO
2
isotopologue distri-
bution toward a stochastic distribution.
Mixing Effects on

47
.
Unlike

13
C,

18
O, and

17
O,

47
can depend
nonlinearly on mixing fraction, even for 2 CO
2
reservoirs with
similar

47
(22, 23), so mixing relationships are often counter-
intuitive. An illustrative example is shown in
Fig. S3
examining
the abundance of minor molecular hydrogen isotopologues in a
mixing process between 2 hydrogen populations. The 2 mixing
end-members (given as the bottom-left and upper-right corners)
have different initial

D values, although both have their H and
D atoms distributed randomly among all isotopologues. The
concentration of D
2
(R
D2
) varies linearly with mixing fraction
and R
D
, but the stochastic distribution for D
2
molecules scales as
(R
D
)
2
. However, unlike in bulk isotope mixtures (

D in this
example), where the bulk isotope abundance is normalized by a
constant value (e.g., R
VSMOW
), multiply-substituted isotopo-
logue abundances are normalized by the stochastic distribution
at that bulk isotope composition (R
D2, stochastic
), whose value
depends on the mixture’s R
D
, and therefore varies with mixing
fraction. Consequently, this mixture of 2 isotopically dissimilar
hydrogen reservoirs would have a ‘‘

D2
’’
0 despite each
reservoir being initially at a stochastic distribution (where ‘‘

D2
’’

0). An analogous plot for CO
2
would have 12 dimensions, but
the qualitative aspects are similar: R
47
mixes linearly, but the
stochastic distribution of mass-47 isotopologues, R
stochastic
47
, de-
pends nonlinearly on the bulk isotope ratios R
13
,R
18
, and R
17
.
Hence, a mixture of 2 dissimilar reservoirs can increase

47
.
In this section, we will show that expected changes in

18
O and

17
O due to intrastratospheric mixing are too small to yield highly
enriched polar vortex

47
. We constructed a mixing model to
simulate the effects of isotopically distinct CO
2
sources on strato-
spheric air. Effects of CO
2
sources on

47
can be calculated simply
by solving the 3-end-member mixing problem between (
i
) tropo-
spheric CO
2
,(
ii
) stratospheric CO
2
undergoing photochemical
isotope exchange, and (
iii
) stratospheric CO
2
derived from a third
source and by accounting for the nonlinearity in mixing. The tropo-
spheric end-member was given an isotopic composition equal to mea-
sured clean troposphere values [based on measurements of air sampled
from Barrow, Alaska, and Cape Grim, Tasmania; (24)], namely

13
C
trop

8‰,

18
O
trop

41‰,

17
O
trop

21‰, and

47trop

0.92‰.
Unfortunately, no measurements of

47
in the tropical troposphere
exist, so

47trop
of the air entering the stratosphere from the tropics is
uncertain. The midlatitude

47
vs.

17
O relationship implies that

47
may be as high as 1.09‰, but changes of

47trop
on this order do not
affect our qualitative understanding of mixing on stratospheric

47
. The
stratospheric end-member was

13
C
strat

8‰,

17
O

80.6‰, and

18
O

98.0‰; this was our integrated effective isotopic composition
for O(
1
D) calculated from the midlatitude

47
variations.
Intrastratospheric Mixing, i.e., Mixing in the Absence of a Third CO
2
Source.
We find that 2-component mixing between tropospheric
and stratospheric air masses does not exhibit sufficient nonlin-
earity to generate the observed

47
enrichments in high-latitude
samples. Mixing nonlinearities in

47
increase as the bulk isotope
compositions of the 2 end-members become more dissimilar, but
stratospheric O(
1
D)–CO
2
isotope exchange alone cannot pro-
duce a CO
2
mixture of sufficiently high

47
. For instance, mixing
effects generate variations in

47

0.1‰ in the troposphere
(22), where bulk isotope compositions of major constituents of
the CO
2
budget vary typically

20‰ (24, 25). A mixture of
tropospheric CO
2
and midlatitude stratospheric CO
2
would
generate a similar enrichment because their bulk isotope com-
positions differ by a similar amount. Even in the extreme case of

18
O
strat

146‰,

17
O
strat

142‰, and

47strat

0‰, the
oxygen isotope composition of CO
2
when it is at photochemical
isotope equilibrium with excess O
2
/O
3
[an unlikely case, but an
ultimately useful one in illustrating this point (20)], 2-component
mixing would yield a maximum

47
change more than an order
of magnitude smaller than, and in the opposite direction of, that
which is observed in polar vortex samples over the same range
in

17
O.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
2of11
Mixing of Gravitationally Separated Air into the Stratosphere.
Gravita-
tional separation of chemical species in the lower stratosphere has
recently been observed by Ishadoya et al. (26), so it may be important
in determining the isotopic composition of stratospheric CO
2
. The
expression in Craig et al. (27) for the enrichment of a species
N
i
versus
a second species
N
in per mil is based on the kinetic theory of gases:


N
i
N



M
gZ
RT

10
3
[s2]
where

M
is the difference in mass,
g
is the acceleration due to
gravity (9.81 m s

2
),
Z
is the column length,
R
is the gas constant,
and
T
is the temperature. By using a reasonable column length of 50 km
at a temperature of 220 K, gravitational separation of CO
2
would
increase

13
C,

18
O, and

17
O by 266‰, 6,223‰, and 1,113‰,
respectively, but would lead to a decrease in

47
by 74‰ at the bottom
of the column. In addition, polar vortex

13
C is not markedly more
enriched from midlatitude

13
C, implying that gravitational separation
is generally negligible for the samples in this study.
Mixing of Tropospheric, Stratospheric, and Mesospheric Air.
We also
explored effects of mixing mesospheric air highly enriched in
18
O
and
17
O with tropospheric and stratospheric air in the polar vortex.
Each polar vortex datum (i.e., with unique

18
O,

17
O, and

47
values) was fit individually because of the presence of mesospheric
filaments in the polar vortex (28), and consequent heterogeneity in
the mixing fraction and end-member isotopic composition between
samples, at the relevant altitudes. Modeled mixing fractions (
f
trop
,
f
strat
, and
f
meso
) and isotopic compositions for each of the high-
latitude CO
2
samples are given in
Table S4
. Because of the number
of adjustable parameters and the uncertainties in those parameters,
these results are likely only meaningful to within an order of
magnitude. Still, our modeled

18
O
strat
and

18
O
meso
values are an
order of magnitude larger than those suggested by Liang et al.
(2007). No laboratory measurements of the fractionations in O
2
due
to Lyman-

photolysis exist; because the Lyman-

lines fall on the
edge of the
E
3
u

4
X
3
g

(1, 0) absorption band of O
2
, a small
absolute cross-section error for 1 O
2
isotopologue can translate into
a significant isotopic fractionation error. Relaxing the mesospheric

18
O vs.

17
O relationship predicted by Liang et al. (2007), i.e.,

17
O

0.3

18
O, increases the modeled

18
O
strat
values and
lowers the predicted

18
O
meso
values, but the mesospheric CO
2
mixing end-member is still enriched overall to
100,000‰ in

18
O
and

17
O. A simple calculation (assuming 360 ppm CO
2
and an
initial

18
O

23.5‰ in O
2
) predicts that the mesospheric O
2
end-member should have a

18
O value on the order of

100‰.
Potential Effects of CO
2
Photolysis on Mesospheric CO and CO
2
.
Bhattacharya, et al. (29) performed laboratory experiments in
which they photolyzed CO
2
with either a Kr lamp (120–160 nm)
or a Hg lamp (185 nm) and measured the
17
O and
18
O compo-
sitions of the product CO and O
2
. The 120- to 160-nm results
suggest that mass-dependent depletions of order 50‰ and
100‰ in
17
O and
18
O of CO, respectively, are possible, which
would enrich the remaining CO
2
reservoir in
17
O and
18
O. The
185-nm results were quite different: Photolysis of CO
2
near
natural isotopic abundance showed mass-independent enrich-
ments in
17
O up to 150‰ with little change in
18
O in the CO and
O
2
products, whereas photolysis of
13
C-labeled CO
2
showed
mass-dependent enrichments in
17
O and
18
Ointhe
13
CO and O
2
products of 50 to 100‰, respectively. Taken alone, these latter
isotope dependences could result in mesospheric CO that is
significantly enriched in
17
O,
18
O, and
13
C, and perhaps even
13
C
18
O. The remaining mesospheric CO
2
would be correspond-
ingly depleted in heavy isotopologues.
If similar isotopic fractionations occur in CO and CO
2
when CO
2
is photolyzed by UV radiation in the mesosphere, then the popu-
lation of mesospheric CO downwelling into the stratospheric polar
vortex could have a large impact on the

47
values of CO
2
in the
polar vortex. Slow oxidation of CO in the polar vortex (which can
be elevated a thousandfold relative to background stratospheric CO
levels) over several months could result in a subpopulation of CO
2
with elevated concentrations of
16
O
12
C
17
O,
16
O
12
C
18
O, and
16
O
13
C
18
O (or other mass-47 isotopologues of CO
2
), given the sign
and magnitude of known KIEs for the CO

OH reaction (30, 31).
The slow oxidation of CO may only produce temporary increases
of heavy-isotopologue abundances because large fractionations due
to KIEs will decrease as CO oxidizes quantitatively to CO
2
. Still,
CO
2
photolysis in the mesosphere, followed by oxidation of the
product CO in the mesosphere and/or stratosphere, may allow the
heavy oxygen and carbon isotopes in CO
2
to ‘‘reorganize’’ in a
manner that could elevate

47
values of CO
2
in the polar vortex. The
mixing of the remaining (unphotolyzed) mesospheric CO
2
with
stratospheric CO
2
in the polar vortex, however, may have the
opposite effect on

47
values of CO
2
.
The full impact of this mechanism is not well-constrained,
however, because (
i
) the isotopic composition of mesospheric
CO
2
has not been measured, (
ii
) not all of the KIEs for the
CO

OH reaction are known at the relevant temperatures and
pressures, and (
iii
) the physical origin of the photolytic fraction-
ations associated with CO
2
photolysis is not understood. The first
2 points were discussed in the main text, and so will not be
discussed here. For the physical origin of the photolytic frac-
tionations associated with CO
2
photolysis, Bhattacharya et al.
(30) suggested that dissociation rates for certain CO
2
isotopo-
logues could be enhanced if their excited vibrational states
(which get populated when the molecule absorbs UV light)
overlapped sufficiently with the vibrational states of another
CO
2
electronic state. This would facilitate conversion of the
electronically excited species CO
2
(
1
B
2
), which is promoted from
ground-state CO
2
when it absorbs UV light and does not have
enough energy to dissociate into O

CO under the experimen-
tal conditions, into the electronically excited species CO
2
(
3
B
2
),
which does have enough energy to dissociate. If this mechanism
is a general feature of the dynamics of CO
2
photodissociation, it
could lead to additional mass-independent fractionations at
wavelengths other than at 185 nm. Unfortunately, this hypothesis
has not been tested at other wavelengths yet, and without a firm
theoretical understanding, or at least an empirical wavelength
dependence, of the isotopic fractionations associated with CO
2
photolysis, convolving the 120- to 160-nm results with the
185-nm results from Bhattacharya, et al.’s experiments, and then
convolving those with the actinic flux to make a prediction for

47
of mesospheric CO and CO
2
is not possible.
1. Felder P, Haas B-M, Huber JR (1991) The photoreaction N
2
O
3
O(
1
D)

N
2
(
1
) studied
by photofragment translational spectroscopy.
Chem Phys Lett
186:177–182.
2. Springsteen LL, Satyapal S, Matsumi Y, Dobeck LM, Houston PL (1993) Anisotropy and
energy disposal in the 193-nm N
2
O photodissociation measured by VUV laser-induced
fluorescence of O(
1
D).
J Phys Chem
97:7239 –7241.
3. Matsumi Y, Shamsuddin SM, Sato Y, Kawasaki M (1994) Velocity relaxation of hot O(
1
D)
atoms by collisions with rare gases, N
2
, and O
2
.
J Chem Phys
101:9610 –9618.
4. Heidner RF, III, Husain D (1974) A study of the collisional quenching of O(
1
D
2
)bythe
noble gases employing time-resolved attenuation of atomic resonance radiation in the
vacuum ultraviolet.
Int J Chem Kinet
6:77– 87.
5. Shi J, Barker JR (1990) Kinetic studies of the deactivation of O
2
(
1
g

) and O(
1
D).
Int
J Chem Kinet
22:1283–1301.
6. Nan G, Houston PL (1992) Velocity relaxation of S(
1
D) by rare gases measured by
Doppler spectroscopy.
J Chem Phys
97:7865–7872.
7. Takahashi K, Taniguchi N, Sato Y, Matsumi Y (2002) Nonthermal steady state
translational energy distributions of O(
1
D) in the stratosphere.
J Geophys Res
107:4290 – 4296.
8. Kharchenko V, Dalgamo (2004) A thermalization of fast O(
1
D) atoms in the strato-
sphere and mesosphere.
J Geophys Res
109:D18311.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
3of11
9. Sander SP, et al. (2006)
Chemical Kinetics and Photochemical Data for Use in Atmo-
spheric Studies: Evaluation Number 15, JPL Publication 06
-
2
(Jet Propulsion Labora-
tory, Pasadena, CA).
10. Johnston JC, Ro
̈ ckmann T, Brenninkmeijer CAM (2000) CO
2

O(
1
D) isotopic exchange:
Laboratory and modeling studies.
J Geophys Res
105:15213–15229.
11. Mebel AM, Hayashi M, Kislov VV, Lin SH (2004) Theoretical study of oxygen isotope
exchange and quenching in the O(
1
D)

CO
2
reaction.
J Phys Chem A
108:7983–
7994.
12. Perri MJ, Wyngarden ALV, Lin JJ, Lee YT, Boering KA (2004) Energy dependence of
oxygen isotope exchange and quenching in the O(
1
D)

CO
2
reaction: A crossed
molecular beam study.
J Phys Chem A
108:7995– 8001.
13. Ianni JC (2003) A comparison of the Bader–Deuflhard and the Cash–Karp Runge–Kutta
integrators for the Gri-Mech 3.0 model based on the chemical kinetics code Kintecus.
Computational Fluid and Solid Mechanics
, ed Bathe KJ (Elsevier, Oxford).
14. Reader J, Sansonetty CJ, Bridges JM (1996) Irradiances of spectral lines in mercury pencil
lamps.
Appl Opt
35:78 – 83.
15. Janssen C, Guenther J, Mauersberger K, Krankowsky D (2001) Kinetic origin of the
ozone isotope effect: A critical analysis of enrichments and rate coefficients.
PhysChem
Chem Phys
3:4718 – 4721.
16. Wiegell MR, Larsen NW, Pedersen T, Egsdard H (1997) The temperature dependence of
the exchange reaction between oxygen atoms and dioxygen molecules studied by
means of isotopes and spectroscopy.
Int J Chem Kinet
29:745–753.
17. Kaye JA, Strobel DF (1983) Enhancement of heavy ozone in the Earth’s atmosphere?
J
Geophys Res
88:8447– 8452.
18. Baulch DL, Breckenridge WH (1966) Isotopic exchange of O(
1
D) with carbon dioxide.
Trans Faraday Soc
62:2768 –2773.
19. Yung YL, Lee AYT, Irion FW, DeMore WB, Wen J (1997) Carbon dioxide in the
atmosphere: Isotopic exchange with ozone and its use as a tracer in the middle
atmosphere.
J Geophys Res
102:10857–10866.
20. Shaheen R, Janssen C, Ro
̈ ckmann T (2007) Investigations of the photochemical isotope
equilibrium between O
2
,CO
2
, and O
3
.
Atmos Chem Phys
7:495–509.
21. Wang Z, Schauble EA, Eiler JM (2004) Equilibrium thermodynamics of multiply substi-
tuted isotopologues of molecular gases.
Geochim Cosmochim Acta
68:4779 – 4797.
22. Eiler JM, Schauble E (2004)
18
O
13
C
16
O in the Earth’s atmosphere.
Geochim Cosmochim
Acta
68:4767– 4777.
23. Affek HP, Eiler JM (2006) Abundance of mass-47 CO
2
in urban air, car exhaust and
human breath.
Geochim Cosmochim Acta
70:1–12.
24. Affek HP, Xu X, Eiler JM (2007) Seasonal and diurnal variations of
13
C
18
O
16
O in air:
Initial observations from Pasadena, CA.
Geochim Cosmochim Acta
71:5033–5043.
25. Ciais P, Meijer HAJ (1998) The
18
O/
16
O isotope ratio of atmospheric CO
2
and its role in
global carbon cycle research.
Stable Isotopes:Iintegration of Biological, Ecological and
Geochemical Processes
, ed Griffiths H (BIOS Scientific, Oxford), pp 409 – 431.
26. Ishidoya S, Sugawara S, Morimoto S, Aoki S, Nakazawa T (2008) Gravitational separa-
tion of major atmospheric components of nitrogen and oxygen in the stratosphere.
Geophys Res Lett
35:L03811.
27. Craig H, Horibe Y, Sowers T (1988) Gravitational separation of gases and isotopes in
polar ice caps.
Science
242:1675–1678.
28. Plumb RA, et al. (2003) Global tracer modeling during SOLVE: High-latitude descent
and mixing.
J Geophys Res Atmos
108:8309.
29. Bhattacharya SK, Savarino J, Thiemens MH (2000) A new class of oxygen isotopic
fractionation in photodissociation of carbon dioxide: Potential implications for atmo-
spheres of Mars and Earth.
Geophys Res Lett
27:1459 –1462.
30. Feilberg KL, Johnson MS, Nielsen CJ (2005) Relative rates of reaction of
13
C
16
O,
12
C
18
O,
12
C
17
O and
13
C
18
O with OH and OD radicals.
Phys Chem Chem Phys
7:2318 –2323.
31. Ro
̈ ckmann T, et al. (1998) Mass-independent oxygen isotope fractionation in atmo-
spheric CO as a result of the reaction CO

OH.
Science
281:544 –546.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
4of11
Fig. S1.

17
O values vs. long-lived stratospheric tracer mixing ratios. In both images, measurements reported in this work (ER-2 and Balloon data) are compared
with previous high-latitude (
48°N) (
Left
) and polar vortex (
Right
) measurements. The measured

17
O correlations with N
2
O and CFC-12 do not reveal obvious
distinctions among the mid- and high-latitude CO
2
samples;

47
measurements, in contrast, reveal strong mesospheric or other polar vortex influence. See
Field
and Laboratory Results
for details.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
5of11
Fig. S2.
Comparison of results from continuous irradiation experiments with calculated results obtained from a kinetics model of the photochemical
experiments. Error bars represent 2

standard errors of each analysis.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
6of11
Fig. S3.
How mixing 2 arbitrary reservoirs of molecular hydrogen can produce a nonstochastic isotopologue distribution. Dashed lines represent mixing vect
ors
for 2 end-members initially at the stochastic distribution. The resulting mixture has an excess of D
2
relative to its stochastic distribution. The isotopic ratios used
in this plot are for illustration purposes only; qualitatively similar effects will apply for all mixing relationships.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
7of11
Table S1. Long-lived tracer data for stratospheric samples
Air type
Sample name

,K

47
,‰

17
O, ‰
N
2
O, ppbv
CH
4
, ppbv
CFC-11, pptv
CFC-12, pptv
ER-2 Samples
V
20000312(25)1120
443.61
1.614
5.76
111.5
948
19
147
V
20000131(30)1001
448.08
1.555
4.38
142.1
1,066
31
198
V
20000203(10)1173
409.37
1.436
3.69
192.8
1,250
72
293
VE
20000127(5)1060
445.10
1.233
2.45
227.6
1,385
112
362
M
20000106(30)1169
467.43
1.071
2.13
256.4
1,495
157
421
M
20000111(25)2021
358.26
1.075
0.18
307.9
1,726
246
526
Balloon samples
M
1-A010-R(2035)
931.14
0.976
6.46
54.0
835
0.13
46.0
M
3-A01-R(1141)
896.15
0.923
6.29
56.7
843
0.23
51.2
M
7-A026-R(1057)
778.08
0.912
6.12
92.1
957
0.63
100.2
M
5-A013-R(2079)
861.84
0.913
5.91
77.8
913
0.15
78.2
M
8-A017-E(1113)
757.89
1.004
5.00
107.9
1,014
1.10
123.6
M
10-A022-E(1186)
709.67
0.916
4.26
155.9
1,176
5.90
208.6
Nonisotopic tracer measurement methodology can be found in Flocke, et al. (1) and Froidveaux, et al. (2).
1. Flocke F, et al. (1999) An examination of chemistry and transport processes in the tropical lower stratosphere using observations of long-lived an
d short-lived compounds obtained
during STRAT and POLARIS.
J Geophys Res Atmos
104:26625–26642.
2. Froidevaux L, et al. (2006) Early validation analyses of atmospheric profiles from EOS MLS on the Aura satellite.
IEEE Trans Geosci Remote Sens
44:1106 –1121.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
8of11
Table S2. Results of pulsed photochemical experiments
Temperature, K

13
C, ‰

18
O
initial
,‰

18
O
final
,‰


18
O*, ‰

47, final
,‰

(

47, final
,‰)
Photolysis experiments
300

10.76
29.84
66.41
36.57

0.009
0.012

10.56
29.95
54.03
24.08
0.036
0.025

23.79
15.50
58.44
42.94
0.104
0.017

24.24
15.48
55.89
40.41

0.018
0.009

10.69
30.08
48.79
18.71
0.038
0.013

10.55
29.63
41.17
11.54
0.057
0.012
229

24.13
15.28
58.33
43.05

0.024
0.032

25.15
14.46
59.86
45.40
0.036
0.023
Blank experiments
300

10.67
29.81
29.72

0.09
0.150
0.011

19.22
20.84
20.88
0.04
0.127
0.012
229

25.09
14.22
14.29
0.07
0.165
0.028
Blank experiments were run without photolysis step.
*


18
O


18
O
final

18
O
initial
.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
9of11
Table S3. Results of continuous irradiation experiments
Sample name*
Irradiation time, h

18
O, ‰

17
O, ‰

47
,‰
Starting CO
2
0.0
6.4
0.0
1.10
ANOM #1
48.0
67.4
24.8
0.72
ANOM #1
48.0
67.4
24.8
0.85
ANOM #2
4.5
13.1
2.3
0.96
ANOM #2
4.5
13.2
2.3
1.04
ANOM #4
10.0
19.0
4.8
0.75
*Samples with the same names are replicate analyses for

47
, taken as separate aliquots of the same experiment.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
10 of 11
Table S4. Tropospheric–stratospheric–mesospheric mixing model results
Mixing model results
Sample
f
trop
*
f
strat
f
meso

18
O
strat
,‰

18
O
meso
,‰

18
O, ‰

17
O, ‰

47
,‰
20000312(25)1120
0.925
0.075
2.6
10

6
82.5
1.76
10
5
44.55
5.76
1.614
20000131(30)1001
0.943
0.057
2.1
10

6
80.0
1.86
10
5
43.62
4.38
1.555
20000203(10)1173
0.952
0.048
1.4
10

6
77.7
2.07
10
5
43.06
3.69
1.436
20000127(5)1060
0.967
0.033
1.7
10

6
87.9
1.47
10
5
42.79
2.45
1.233
*Mixing fraction of tropospheric air, calculated as 1

f
strat

f
meso
.
Yeung et al.
www.pnas.org/cgi/content/short/0902930106
11 of 11