Field campaigns have been carried out with the FAGE (fluorescence assay by gas expansion) technique in remote biogenic environments in the last decade to quantify the in situ concentrations of OH, the main oxidant in the atmosphere. These data have revealed concentrations of OH radicals up to a factor of 10 higher than predicted by models, whereby the disagreement increases with decreasing NO concentration. This was interpreted as a major lack in our understanding of the chemistry of biogenic VOCs (volatile organic compounds), particularly isoprene, which are dominant in remote pristine conditions. But interferences in these measurements of unknown origin have also been discovered for some FAGE instruments: using a pre-injector, all ambient OH is removed by fast reaction before entering the FAGE cell, and any remaining OH signal can be attributed to an interference. This technique is now systematically used for FAGE measurements, allowing the reliable quantification of ambient OH concentrations along with the signal due to interference OH. However, the disagreement between modelled and measured high OH concentrations of earlier field campaigns as well as the origin of the now-quantifiable background OH is still not understood. We present in this paper the compelling idea that this interference, and thus the disagreement between model and measurement in earlier field campaigns, might be at least partially due to the unexpected decomposition of a new class of molecule, ROOOH, within the FAGE instruments. This idea is based on experiments, obtained with the FAGE set-up of the University of Lille, and supported by a modelling study. Even though the occurrence of this interference will be highly dependent on the design and measurement conditions of different FAGE instruments, including ROOOH in atmospheric chemistry models might reflect a missing piece of the puzzle in our understanding of OH in clean atmospheres.
OH radicals are the most important oxidant in the atmosphere, and the
detailed understanding of their formation and reactivity is key for the
understanding of the overall chemistry. Upon reaction with volatile organic
compounds (VOCs, such as methane and isoprene), OH oxidation leads to the
production of organic peroxy radicals (
An alternative explanation for the unexpectedly high OH concentrations
measured in biogenic, low-NO environments is that the measurements suffer
from an unidentified interference. Indeed, all of these measurements have
been carried out using a technique named FAGE (fluorescence assay by gas
expansion). Briefly, ambient air is rapidly expanded into a low-pressure
volume, where OH radicals are excited by 308 nm light and the resulting
fluorescence is detected (Heard and Pilling, 2003). Calibration of the
fluorescence signal allows the determination of absolute concentrations
(Dusanter et al., 2008). Interferences can arise from different sources such
as photolysis of suitable precursors by the fluorescence excitation laser
(e.g.
The third source of interference, the generation of OH radicals during the
expansion into the FAGE cell, is more difficult to identify because only one
photon is needed and hence the interfering species would appear as ambient
OH. Following the large disagreements between measurements and models, the
group of W. Brune at Penn State University (State College) redesigned a concept to quantify such possible
interferences (Mao et al., 2012), which had first been tested by Dubey et
al. (1996): a pre-injector device is installed just above the inlet into the
FAGE cell, which injects regularly into the airflow a high concentration of a
species rapidly reacting with OH radicals. This way all ambient OH radicals
are scavenged before entering the FAGE cell, and any remaining signal can be
identified as interference. The difference between the signal with and
without the scavenger allows the quantification of the real ambient OH. The
use of this technique was reported for the first time in 2012, showing results
for a field campaign in a forest in California (Mao et al., 2012). It led to
the identification of a large fluorescence signal following scavenging of all
ambient OH radicals, corresponding to up to 50 % of the total OH
concentration. The OH concentrations obtained with the scavenger agreed well
with models, while the OH concentrations obtained without the scavenger
exceeded modelled concentrations by up to a factor of 3. Other groups also
developed a pre-injector system in the following years (Griffith et al.,
2016; Novelli et al., 2014a; Tan et al., 2017). Using this system, Novelli et
al. (2014a) observed strong interferences in their FAGE system during
three field campaigns in remote biogenic environments in Germany, Finland and
Spain, while Griffith et al. (2016) were able to account for the observations
through known interferences by
Novelli et al. (2014a) proposed that ozonolysis of
alkenes, leading to the formation of Criegee intermediates and the subsequent
decomposition of these Criegee intermediates within the FAGE cell, was
responsible for the interference (Novelli et al., 2017). However, using
different FAGE systems, Rickly and Stevens (2018) and Fuchs et al. (2016)
could not confirm this source: even though they detected internally formed OH
when mixing high concentrations of
Following several years of interference studies in various environments,
recent work from W. Brune's group (Feiner et al., 2016) concluded that the
interference observed in their FAGE system
was due to a rather long-lived species because the interference persists
into the evening; had been observed in different environments
dominated by MBO (2-methyl-3-buten-2-ol), terpenes or isoprene; hence, it must originate from a class
of species rather than from only one species such as isoprene; strongly increased with increasing O(
In this work we present experimental and modelling evidence that this
sought-after species could be the product of the reaction between
In the first part, the experimental evidence for the interference generated in the UL-FAGE (FAGE instrument of the University of Lille) by the presence of ROOOH molecules will be presented. It should be noted that the intensity of interferences or even the presence at all can depend on the design of the FAGE instrument (inlet design, pressure drop, residence time, etc.), and the results presented here are only valid for the FAGE instrument of the University of Lille. Other FAGE instruments need to be tested individually for the possible presence of an interference in OH measurement due to the presence of ROOOH. In the second part, model calculations are used in order to estimate the steady-state concentration of ROOOH molecules that can possibly build up in different environments.
With the goal of forming sizeable amounts of trioxide (ROOOH), experiments
have been carried out in a pump-probe UL-FAGE, described already in detail in earlier publications (Fuchs et al.,
2017; Hansen et al., 2015; Parker et al., 2011). Briefly, a gas mixture
containing the VOC (isoprene,
Schematic view of the experimental set-up.
Experiments start with a fresh mixture (i.e. with the photolysis laser
manually covered), and 40 decays are then recorded every 0.5 s for 20 s. After
40 photolysis pulses the laser is covered again for 2 min to allow the
mixture to completely refresh and, in order to improve the
OH concentration time profiles following the photolysis of
600 ppb
The initial isoprene concentration (3
Model used to estimate the accumulation of ROOOH in the photolysis cell before entering the FAGE cell; all rate constants have been taken from the most recent IUPAC evaluations (Atkinson et al., 2004, 2006).
A yield of 1 is estimated for the formation of ROOOH in the reaction of
This model was run 40 times for 0.5 s, with the final concentrations of the
different species obtained at each run being used as initial concentrations
in the following run, always adding 1.4
Evolution of different species in the photolysis cell as a function
of the number of photolysis pulses. Full black line describes evolution of
The goal of this model is not to precisely describe the ongoing chemistry but rather to get a good idea of how much ROOOH is possibly accumulated. The
model uses different simplifications:
OH radicals only react with species present in the model, i.e. no wall
loss or reaction with impurities is taken into account; the possible
photolysis of ROOOH at 266 nm or a heterogeneous loss on the reactor walls
is not taken into account; no reaction of OH with the products of
the photolysis beam has been
considered homogeneous; the inhomogeneity of the beam profile of our
photolysis laser has not been considered; the decrease in available
All these simplifications lead to an uncertainty in the final ROOOH
concentration, possibly of up to a factor of 10. Most of the simplifications
will lead to an overestimation of the final ROOOH concentration (either ROOOH
is consumed or less is formed), except for the inhomogeneous photolysis beam
where the direction of uncertainty is not easy to determine (higher formation
of ROOOH in the hotspots of the laser beam and lower formation in the rest of the
volume). The model predicts the formation of around
[ROOOH]
The model predicts the consumption of most isoprene over the 40 photolysis
pulses, which should lead to a decrease in the decay rate, given the much
faster rate constant of OH with isoprene compared to the reaction products. A
single-exponential decay was then fitted to the experimental OH profiles from
Fig. 2, and the resulting pseudo-first order decay rates are shown as blue
dots in Fig. 4. It can be seen that the decay rate decreases over the 40
shots by around 20 s
Results of fitting a mono-exponential decay to the raw signal of the
experiments shown in Fig. 1. Blue dots: OH decay rates from the
mono-exponential fit between 0.02 s and the end of the data set (left
Photolysis of
In order to better understand the origin of the increase in the LIF signal, additional experiments have been carried out.
Photolytically generated interferences need two photons for generating one
fluorescence photon and can thus be identified by either varying the
fluorescence excitation laser energy (the signal intensity would increase
with the square of the excitation laser energy) or by changing the repetition
rate of the excitation laser (photolytically generated interferences appear
because the air mass within the excitation volume is not completely renewed
between two excitation laser pulses (200
The blue dots on the lower graphs show the decrease in the decay rate with an
increasing number of photolysis pulses, of the same order of magnitude for
all three experiments, as expected (photolysis energies as well as isoprene
and
Summary of results from Fig. 5.
From the observation that the increase in residual LIF signal with an increasing number of photolysis pulses is independent of both (a) the fluorescence excitation laser energy and (b) the repetition rate of the excitation laser, we conclude that the observed interference in the UL-FAGE is not due to a photolytic process.
Additional experiments have been carried out using identical OH
concentrations but much higher isoprene concentrations than in the above
experiments. Under these conditions, there is still formation of high
concentrations of
The results of these experiments are shown in Fig. 6. For the conditions in
the left graphs
Experiments with high isoprene concentrations:
From these observations, it can be concluded that the increase in LIF
intensity at long reaction times observed in the experiments presented in
Fig. 4 is consistent with being generated by the product of the reaction
between
To further support the hypothesis that the observed increase in residual LIF
signal is due to an interference generated by the product of the reaction of
Photolysis of
For the lowest concentration (left graphs in Fig. 7), a high formation of
ROOOH can be expected: under these conditions OH radicals react slowly with
butane, and the reaction with the nascent
Photolysis of
Experiments with
In both experiments the LIF intensity at long times does not change (
The increase in residual LIF signal in Fig. 4 over the 40 photolysis pulses
is around 0.005 arb. units. This can be compared with the raw OH decays shown
in Fig. 2: the initial LIF signal of
It can hence be concluded that in the UL-FAGE, an interference signal
corresponding to [OH]
In order to estimate if ROOOH concentrations in this range can possibly be accumulated in remote biogenic environments, calculations using global and box models have been performed.
The global distribution of ROOOH species produced by the
Modelled mean diurnal peak ROOOH volume mixing ratio (in ppt) during
northern
Crucially, the model simulated the abundances of a number of peroxy radicals
resulting from the oxidation of emitted VOCs:
Steady-state ROOOH abundances were calculated “offline” using the modelled
abundances of hourly [OH] and [
Variation in ROOOH as a function of NO (
To confirm these global model results, a steady-state box model, constrained
to observations (including OH, NO, isoprene and
In this work we have shown that the product of the reaction of
Underestimation by models of OH concentrations measured in remote, biogenic
environments: the global model predicts ROOOH peak concentrations in remote
environments that are possibly high enough to explain, at least partially,
the observed disagreement between model and measurements (Whalley et al.,
2011; Lelieveld et al., 2008; Hofzumahaus et al., 2009; Tan et al., 2017). Variability of interferences observed in field campaigns: the box model
calculations have shown that the concentration of ROOOH species varies with
NO, VOC concentration and J(O Interference observed from Interferences observed in the SAPHIR chamber: Fuchs et al. (2012)
have carried out experiments under low-NO conditions by comparing OH
concentrations measured by FAGE and DOAS (Fuchs et al., 2012). Most of the
time the agreement between both techniques was excellent, but on a few days
towards the end of the campaign higher OH concentrations were measured by
FAGE compared to DOAS. The NO concentrations on these days were lower, making
the formation of ROOOH more likely than on days with excellent agreement
between FAGE and DOAS (Table 2 in Fuchs et al., 2012).
The results presented in this work thus propose a plausible solution to
answer many open questions; it is, however, not very likely that they can explain
an increase in the interference at night, such as observed by Novelli et
al. (2014a). Of course, the uncertainties are currently high on both the
observed FAGE interference per ROOOH molecule as well as the maximum ROOOH
concentration that can accumulate in real environments. The first point could
be improved through well-designed chamber studies under very low-NO
concentrations: such experiments have already been carried out (Nguyen et
al., 2014), and a detailed analysis of the data might support the conclusions
from this work. The second point is more difficult to ameliorate because the
steady-state ROOOH concentration directly scales with its removal rate, and
currently nothing is known about the fate of ROOOH. Perhaps the table can be
turned by using the evolution of the observed interferences to learn about
the fate of ROOOH?
Nonetheless, even with current uncertainties the implications of our understanding of daylight atmospheric oxidation chemistry are significant. We provide a plausible mechanism for how and why high OH levels in some environments are bolstered by a false signal, in a sense validating our current generation of models and reducing the need for speculative chemistry to explain the difference in simulated and observed OH of earlier field campaigns in pristine environments. With further observations and model development, the outcome will be to improve our ability to predict the OH budget in pristine environments and the impacts of changes on the global chemistry–climate system.
Data are available upon request.
The supplement related to this article is available online at:
SB and CS developed the FAGE instrument; CF and CS designed the experiments and MA carried them out; VF, SAN and ATA developed the model and performed the simulations. CF prepared the paper with contributions from all co-authors.
The authors declare that they have no conflict of interest.
This project was supported by the French ANR agency under contract no. ANR-11-LabX-0005-01 CaPPA (Chemical and Physical Properties of the Atmosphere), the Région Hauts-de-France, the Ministère de l'Enseignement Supérieur et de la Recherche (CPER Climibio) and the European Fund for Regional Economic Development. Alexander T. Archibald and Scott Archer-Nicholls thank NERC-NCAS and the Walters Kundert Trust under whose auspices this work was enabled. Valerio Ferracci thanks the European Research Council for funding through the Atmospheric Chemistry-Climate Interactions (ACCI) project, project number 267760. UM-UKCA runs in this work used the ARCHER UK National Supercomputing Service. The authors thank Paul Wennberg for very helpful discussions. Edited by: Frank Keutsch Reviewed by: two anonymous referees