Investigating Urban Atmospheric Chemistry Using a Time of Flight Chemical Ionisation Mass Spectrometer a Thesis Submitted To

Investigating urban atmospheric chemistry using a time of flight chemical ionisation mass spectrometer

A thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Science and Engineering

2018

Michael Priestley

School of Earth and Environmental Sciences

Blank page

Contents Abstract ...... 1 Declaration ...... 2 Copyright Statement...... 3 Acknowledgements ...... 4 Alternative format thesis overview ...... 5 1. Introduction ...... 7 1.1. Climate and air quality ...... 7 1.2. The urban atmosphere ...... 10 1.2.1. Volatile organic compounds (VOCs) ...... 12 1.2.2. Oxidants ...... 15 1.2.3. Oxidation ...... 20 1.3. Air quality in the UK ...... 24 1.4. Measuring the chemical composition of the urban atmosphere ...... 28 1.4.1. Mass spectrometry ...... 29 1.4.2. Data acquisition and analysis ...... 33 2. Aims and Objectives ...... 46 3. Paper 1. Observations of isocyanate, amide, nitrate and nitro compounds from an anthropogenic biomass burning event using a ToF‐CIMS ...... 48 4. Paper 2. Observations of organic and inorganic chlorinated compounds and their contribution to chlorine radical concentrations in an urban environment in Northern Europe during the wintertime ...... 50 5. Paper 3. Detection of highly oxidised molecules from the reaction of benzene + OH under different NOx conditions ...... 53 6. Conclusions ...... 54 6.1. Future work ...... 58 Bibliography ...... 63 Appendix A. Co-authorship in peer reviewed publications ...... 74

Word count: 51,590

Blank page

Abstract Atmospheric reactive trace gases perturb the climate system and air quality through many direct and indirect effects. The poor air quality they propagate is amplified in large urban centres where emissions and processing are highly variable, and large populations are exposed. Many chemical processes are poorly understood due to the highly complex interactions and variable composition of the urban atmosphere. These drive the development of new atmospheric measurement instruments that can reliably measure reactive trace gases of very low concentration with high temporal resolution. The iodide time of flight chemical ionisation mass spectrometer (ToF-CIMS) is one such instrument that is able to probe reactive atmospheric systems due to its high linearity of response, reproducibility, and its selectivity and sensitivity towards inorganic species, including chlorinated and brominated species, and multifunctional oxygenated organic species, all of which are of relevance for the study of urban air.

An iodide ToF-CIMS was deployed at the University of Manchester for a two week period in October and November 2014 to assess its ability to detect trace gases relevant to climate and air quality. A biomass burning event (Guy Fawkes Night) was sampled, from which markers of combustion (HCN and HNCO) and other newly detected nitrogen containing species (amides) were quantified and their emission ratios (NEMR) to CO calculated. The HCN NEMR of 1.11 ± 0.62 ppt ppb-1, whilst low, is of a comparable order to other biomass burning studies. Chlorinated organics were also identified throughout the sample period. Their contribution to the steady state Cl radical budget was quantified and compared with the contribution from inorganic Cl species also measured. The detection of day time Cl2 suggests a photochemical mechanism is the cause of production and is a significant source of Cl throughout the day (74%), more so than

ClNO2 (23%) when the shortwave (sw) radiation flux is large. The newly detected ClOVOCs are a negligible source of Cl under both low and high sw flux conditions (3%).

The iodide ToF-CIMS was also deployed at the Jülich plant chamber where many products of benzene oxidation by OH under different NOx conditions were identified. These measurements were contrasted with a nitrate ToF-CIMS that exhibits a different compound detection selectivity. Detection overlap between instruments was observed, however different O:C ratios for species with the same carbon number were found. Products identified by iodide ToF-CIMS in the chamber were identified in the Manchester urban ambient dataset, if they contained 6 carbon atoms and had high N:C (0.3 - 0.5) and O:C ratios (1.5 – 2.0). This suggests the chamber may not be representative of ambient conditions.

Declaration

No portion of the work referred to in the thesis has been submitted in support of an application for another degree of qualification of this or any other university of other institute of learning.

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/display.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/ regulations/) and in The University’s Policy on Presentation of Theses.

Acknowledgements

I would like to express my immense gratitude to Prof. Carl Percival for his supervision and dedication to me and my work despite moving on to pastures new. I would also like to thank Prof. Hugh Coe for his supervision and guidance during the latter part of my time at Manchester. I would like to thank the National Environment Research Council (NERC) for the Doctoral Training studentship I received to fund my work.

I also wish to thank past and present members of the Percival Group, Asan Bacak, Mike le Breton, Tom Bannan, Kimberly Leather, Stephen Worrall and Archit Mehra for their unyielding guidance and friendship. Thank you Ernesto Reyes-Villegas, Yu-Chieh ‘Danny’ Ting, Gillian Young, Nick Marsden, Hazel Jones and Waldemar Schledewitz. I am grateful to have shared on office with others who made the daily grind so much more enjoyable.

Finally, thank you to my parents, for their constant support and encouragement, and to Hannah Josey for your ever present optimism and selfless dedication.

Alternative format thesis overview Measurements of trace gases relevant to air quality using a time of flight chemical ionisation mass spectrometer (ToF-CIMS).

What species can the ToF-CIMS Which species can the ToF-CIMS detect under ambient conditions? observe that are relevant to urban oxidation?

Observations of isocyanate, Observations of organic and amide, nitrate and nitro inorganic chlorinated compounds compounds from an and their contribution to chlorine anthropogenic biomass burning radical concentrations in an urban event using a ToF-CIMS environment in Northern Europe during the wintertime Michael Priestley, Michael Le Breton, Thomas J. Bannan, Kimberly E. Leather, Michael Priestley, Michael le Breton, Asan Bacak, Ernesto Reyes-Villegas, Thomas J. Bannan, Stephen D. Worrall, Frank De Vocht, Beth M. A. Shallcross, Asan Bacak, Andrew R. D. Smedley, Toby Brazier, M. Anwar Khan, James Ernesto Reyes-Villegas, Archit Mehra, Allan, Dudley E. Shallcross, Hugh Coe, James Allan, Ann R. Webb, Dudley E. Carl J. Percival. Shallcross, Hugh Coe, Carl J. Percival.

What urban oxidation products is the iodide ToF-CIMS capable of detecting under different NOx conditions?

Detection of highly oxidised molecules from the reaction of benzene + OH under different NOx conditions

Michael Priestley, Michael Le Breton, Thomas J. Bannan, Stephen D. Worrall, Sungah Kang, Iida Pullinen, Thomas Mentel, Asan Bacak, Dudley E. Shallcross, Gordon McFiggans, Hugh Coe, Carl J. Percival

Blank page

1. Introduction Reactive trace gases comprise <1% of the Earth’s total atmospheric volume but the complexity of that composition is so great that it is not possible to quantify with certainty the absolute number of atmospheric chemical constituents and likely reaches into the 10,000s (Williams 2004). Many atmospheric chemical processes are as yet undetermined or poorly understood and in order to recreate them within earth system models requires much parameterisation and simplification e.g. approximating many compounds as one species. Measurements of atmospheric trace gases are the primary method by which our empirical understanding of traces gases and their interactions can be expanded.

1.1. Climate and air quality Traces gases play a vital role in the Earth’s climate system (Fig. 1). They are responsible for phenomena that can directly and indirectly perturb the Earth’s radiation budget. They can directly interact with radiation, for example the current estimated radiative forcing by -2 O3 relative to pre-industrial levels (1750) is estimated to be 0.4 (± 0.2) Wm (Myhre et al. 2013). Trace gases can interact with the biosphere creating climatic feedbacks that perturb the Earth’s radiation budget. For example O3 is phytotoxic and can reduce the uptake of CO2 by damaged plants (Sitch et al. 2007). Furthermore, O3 damage can cause plants to release more isoprene (Wang et al. 2016), which is a precursor of O3, and so propagates a positive feedback mechanism.

Oxidation of reactive trace gases forms new species with different physico-chemical properties. For example, if by oxidising a material the functionality and polarity of the material is increased, it is expected the new material will have a lower vapour pressure than the starting material (Kroll & Seinfeld 2008) and so more readily condense onto aerosol particles, causing them to grow, or even initiate new particle growth itself (Bianchi et al. 2016). This mechanism forms secondary organic aerosol (SOA) which scatters (Scott et al. 2014) and absorbs light and can alter surface albedo (Lin et al. 2014). If these particles reach a critical size (typically ~ 100s nm (Dusek et al. 2006)), they can activate as cloud condensation nuclei (CCN). This initiates cloud formation that have their own optical properties and thus perturbations on the radiation budget and climate system (France et al. 2013).

Fig. 1. Summary of the interaction of reactive trace gases and their effects on the earth system

Many trace gases and aerosols are toxic to humans and plant life. SOA can comprise a large proportion (~ 50 %) of particulate matter with an aerodynamic diameter of <2.5 µm

(PM2.5) (Kim et al. 2015). These particles are not effectively filtered by the nose and so they are capable of penetrating deep into the lungs where they are absorbed into the blood stream and induce inflammation in the lungs (Terzano et al. 2010). Beyond the physical effects of PM2.5 on the respiratory system, many gas phase precursors and dissolved gas phase constituents of PM2.5 are toxic to humans e.g. poly-aromatic hydrocarbons (PAHs) can directly interact with DNA increasing the genotoxicity of the inhaled particles (Godschalk et al. 2000). This provides a motivation to understand the origins of atmospheric toxic semi-volatile material and the mechanisms that lead to its formation.

In some instances the concentration of atmospheric trace gases causes acute toxicity. This is most commonly associated with natural phenomena such as volcanic emissions (Ilyinskaya et al. 2017) and exposure to forest fire smoke (Henderson & Johnston 2012) or industrial accidents (e.g. D’Silva et al. 1986), for which work place exposure limits and control parameters are required to protect human safety (Health and Safety Executive 2011). More worryingly, the chronic impacts of air pollutants are known to increase morbidity and mortality with recent estimates suggesting 7 million deaths worldwide are

8 attributable to poor air quality (WHO 2014), thus making the reduction of air pollution one of the greatest environmental challenges for policy makers around the globe.

Chronic exposure to poor air quality has been linked to a range of physiological conditions such as increased incidences of vascular dementia and Alzheimer’s disease (Jung et al. 2015; Oudin et al. 2016), reduced lung functionality in children (Gehring et al. 2013) and oxidative stress at the cellular level (Rao et al. 2017). In the UK, the enhancement in NO2 concentrations, primarily due to vehicular emission, is estimated to contribute to the early deaths of 40,000 people per year through stress placed on the respiratory and cardiovascular systems (Ornes 2016). Poor air quality of some form is experienced across nearly all geographic and demographic divisions around the globe with recent estimates placing 95% of the world’s population affected (Health Effects Institute 2018).

1.2. The urban atmosphere The troposphere is the lowest portion of the atmosphere ranging from 0 km to ~8 km (~ 200hPa) at the poles and ~14 km at the equator. The tropopause, defined by a minimum in the temperature gradient of the earth’s atmosphere separates the troposphere from the stratosphere above (~10 - 50 km or ~200 - 1 hPa). Gaseous exchange across the tropopause is a slow process relative to the lifetimes of many trace gases so is only relevant for long lived species e.g. chlorofluorocarbons (CFC).

The planetary boundary layer (PBL) describes the lowest portion of the troposphere, typically between 0.2 – 2.0 km that is adjacent to the Earth’s surface (Fig. 2). The PBL is is the consequence of the friction between the bulk atmosphere and the Earth’s surface and is characterised by buoyant and shear turbulence. It is within this layer that air surface gaseous exchange occurs either by emission from the surface or deposition to it. The height of this convective mixed layer changes as a function of time of day. Heating of the surface by solar radiation increases the activity of small scale dynamics and dispersion processes like turbulence by generating static instabilities in the form of large convective eddies driven by buoyancy changes in air masses (Falasca et al. 2013). At night the boundary layer stabilises and decreases in height, leaving behind a residual layer above. This residual layer, which contains the contents of the previous day’s boundary layer, may be entrained into the next day’s mixing layer.

Fig. 2. Diurnal profile of planetary boundary layer adapted from (NikNaks 2012). The collapse of the boundary layer at sunset leaves a residual layer aloft that may preserve the previous day’s pollutants and can be entrained into the next day’s boundary layer as it grows in height from sunrise. The surface layer describes the air mass where wind speed is logarithmic normal to the surface.

The urban boundary layer (UBL) is where the urban surface influences the air above and so is distinct from a rural planetary boundary layer (PBL) (Oke 1976) (Fig. 3). The urban canopy layer (UCL) is a layer that exists at roof level whose airflow processes are governed at the micro-scale as opposed to the local and regional meteorology that governs UBL processes. The regional wind advects the air mass associated with the urban boundary layer forming a plume of urban outflow which carries pollutants into the rural surroundings (Hidalgo et al. 2008)

Large conurbations in all climates exhibit higher air temperatures than surrounding rural areas (described as the Urban Heat Island, UHI) (Hinkel & Nelson 2007). This is caused by the addition of anthropogenic heat sources as well as the contribution of radiative heating to anthropogenic surfaces. This further drives convective mixing and increases the height of the urban boundary layer relative to rural areas. At night, the interaction of the colder air in the rural locality and the warm urban centre may develop an urban heat island circulation (UHIC) analogous to land-sea breezes (Haeger-Eugensson & Holmer 1999). This further complicates the transport of gaseous pollutants and air pollution precursors affecting the urban environment and surrounding rural areas during the night and the next day when photochemistry is initiated again.

Fig. 3. Schematic of the urban boundary layer from Hidalgo et al. (2008). The regional wind above the boundary layer defines the downwind urban plume.

The urban environment contains many different emission sources and so many different types of pollutant e.g. trace metals, inorganic aerosol e.g. sulphates and nitrates, inorganic gases or volatile organic compounds (VOCs) for which a variety of different sources exist. In some instances it is possible to identify the exact processes contributing to the pollution, e.g. specific trace metals from specific industrial processes (e.g. Font et al. 2015) or the contribution of different sources to one type of pollution e.g. residential, traffic and commercial to primary organic aerosol (e.g. Reyes-Villegas et al. 2016). Whereas with other pollutants, particularly VOCs, it is less possible, as multiple sources and secondary processing may produce the same compound (e.g. Bannan et al. 2017).

The complexity resulting from many different VOCs, their emission sources, their chemical and physical processing and ultimately the effects this has on human health and climate is a large motivation for their study.

1.2.1. Volatile organic compounds (VOCs) VOCs are gaseous carbon containing compounds whose structures govern properties such as volatility, reactivity, solubility and therefore atmospheric reaction pathway (Altenstedt and Pleijel, 2000).

Globally the most abundant VOCs are biogenic (BVOCs), specifically terpenes or terpenoids such as isoprene, monoterpenes and sesquiterpenes (Fiore et al. 2012; Park et al. 2013) as well as small oxygenated organic compounds e.g. alcohols, aldehydes and organic acids (Villanueva-Fierro et al. 2004). The global emission rate of BVOCs is estimated to be 1150 Tg C yr−1 (Guenther et al. 1995) of which 523-800 Tg C yr−1 are attributed to isoprene and 30-177 Tg C yr−1 to monoterpenes (Guenther et al. 2012). Emission rates of these compounds vary as a function of plant type and external stresses placed on the plant, e.g. temperature and O3 exposure (Guenther et al. 1995).

Anthropogenic VOCs are more typically saturated or aromatic compounds (Atkinson 2000) that are attributed to activities such as; industrial processes, e.g. power generation and manufacturing; transport, e.g. vehicular emissions and fuel related evaporative emissions; and waste management, including solid and water waste treatment and burning (Huang et al. 2017). For anthropogenic saturated, unsaturated and aromatic compounds, global emission is estimated to be 105 Tg C yr−1 or around 10% of the BVOC emission (Guenther et al. 2012). However, anthropogenic emissions in Europe are believed to be of a comparable order to emitted biogenic non-methane VOCs (Simpson et al. 1999) as a result of extensive deforestation and the demanding consumption of fossil fuels.

As many VOCs are highly reactive, they do not have sufficient time to mix well in the atmosphere therefore their concentrations vary both spatially and temporally depending on the source (Borbon et al. 2013, Monks et al. 2015) (except ethane which has a hemispheric background) (Fig. 4). Whilst global emissions of anthropogenic VOCs may be lower than biogenic emissions, the effect of emissions from urban conurbations are disproportionately large compared with the spatial footprint and populations that are exposed to their emission. For example, whilst Asian megacities comprise approximately 2% of the continental land mass, their combined anthropogenic emissions comprise ~10- 15% of that area affecting ~30% of the population (Guttikunda et al. 2005).

Fig. 4. Emissions map of non-methane VOC. The influence of anthropogenic activity and the high spatial variability of VOC emissions are apparent (National Atmospheric Emissions Inventory 2015).

In Europe and the UK, there are a number of contributing sectors to urban air pollution including traffic, residential, industrial and commercial. Emissions are typically either; fugitive, e.g. from cooking in the commercial and residential sectors (Reyes-Villegas et al. 2018) and solvent evaporation from industrial processes; or from incomplete combustion sources, such as domestic wood burning, engine combustion from traffic or power station fuel consumption (Leggett 1996).

Biomass burning is one major emission source of trace gases to the atmosphere. As a global emission source, forest fires in the tropics, specifically equatorial regions contribute the most. Human influenced biomass burning is common around the globe as a method of agricultural burning and land clearing (Lemieux et al. 2004), space heating and cooking (Desai et al. 2004). This can lead to a swift degradation of air quality in enveloped urban centres where large populations reside and brings several health and economic impacts (Kunii et al. 2002; Aouizerats et al. 2015), negatively impacting the health of billions (Soldatova et al. 2011). In Europe, domestic biofuel burning for space heating using wood burners is growing in popularity and recent UK Government findings suggest that domestic wood fuel use in open fires and closed wood burners has until

2015 been underrepresented by a factor of 3 suggesting it is a significant source of emission to the atmosphere (Department for Energy and Climate Change 2016).

Open burning and incineration of waste is another form of anthropogenic biomass burning and is estimated to contribute 7% of VOC emissions globally (Wiedinmyer et al. 2014) although estimates on absolute emissions are still relatively uncertain. On a local scale, small scale anthropogenic fires that may originate from biomass or waste burning contributes to poor air quality (Christian et al. 2009) and is a relevant issue regarding UK air quality (Lohmann et al. 2000).

Where transformations of the primary emitted compounds occur through a variety of oxidative mechanisms (discussed later) secondary products are formed. As VOCs oxidise, the increase in oxygenated functionality forms more polar molecules and so facilitates a greater degree of intermolecular bonding, reducing the vapour pressure of the material and so its volatility (Kroll & Seinfeld 2008). Broadly, a material with a vapour pressure of greater than 1 Pa (under typical atmospheric conditions) will exist in the gas phase whereas those with a vapour pressure less than 10-4 Pa will be largely in the condensed phase (Valorso et al. 2011). The material that partitions to the aerosol phase, either on pre-existing surfaces or self-nucleating, results in the growth of SOA.

Many facets of SOA formation are poorly understood and so are the subjects of much study by flow tube experiments and in environmental chambers (Rissanen et al. 2014, Kiendler-Scharr et al. 2009). This work typically aims to quantify the yield of SOA from different precursors, under different ambient conditions e.g. NOx, RH, aerosol seed (Stirnweis et al. 2017) and to study the mechanistic processes of SOA formation e.g. auto-oxidation (Ehn et al. 2014), condensed phase reactions (Slade et al. 2017), reactive uptake (Surratt et al. 2010), hygroscopicity (Suda et al. 2014) and product distributions (Aljawhary et al. 2013).

The interaction between biogenic VOCs and anthropogenic pollution can further complicate the attribution of secondary pollutant formation to any one phenomenon. For example SOA formation over China is expected to be higher during the summer due to a larger isoprene source during that season, compared with more dominant anthropogenic VOC emission during the winter (Hu et al. 2017). The additional complication of mixed VOC precursors is another aspect of SOA formation that is of current interest to environmental chambers studies (Ahlberg et al. 2017).

Within the urban environment, there are different loss mechanisms for air pollutants. For example, the lifetime of vehicular emissions at kerbside is controlled by dispersive loss mechanisms and so is a dynamical process (Sanchez et al. 2016). In turbulent

14 conditions such as a street canyon, air parcels with elevated concentrations swiftly mix but may remain elevated to a point where urban background levels are maintained at a measureable level. For reactive trace gases, chemical reactivity of the pollutant is an important factor determining the pollutant lifetime. Oxidation is the major governing process for chemical transformations of trace gases in the urban atmosphere that relies on catalytic mechanisms to recycle oxidants enabling the loss of VOCs from the atmosphere.

1.2.2. Oxidants The importance of an oxidant to the total oxidising capacity of the atmosphere is dependent on reactivity and concentration. The four major atmospheric oxidants are OH,

O3, NO3 and Cl.

median rate [daytime] a [night time] a Oxidant coefficient c molecule cm-3 molecule cm-3 molecule-1 s-1 cm3 alkanes alkenes OH 3.94 × 106 1.72 × 104 5 × 10-12 6 × 10-11 7 9 -16 -11 NO3 7.38 × 10 2.46 × 10 5 × 10 2 × 10 12 12 -14 O3 2.71 × 10 1.97 × 10 - 2 × 10 Cl 1.00 × 104 b → 0 b 5 × 10-10 5 × 10-10 Table 1. Average polluted daytime and night time concentrations and alkane alkene reactivity’s of the most important atmospheric oxidants in a polluted atmosphere. a Unless specified otherwise, concentrations are taken from Calvert et al. (2000). b Estimated by Bannan et al. (2015). c taken from McGillen et al. (2006a, 2006b)

Table 1 summarises their average concentrations and reactivity’s. Tropospheric OH concentrations are greatest during the day and alkane reactivity is high. This is also true of Cl whose reactivity can be 200 times greater than that of OH, however its typical tropospheric concentrations are lower by the same order. During the day, alkene oxidation by O3 is also competitive. At night OH and Cl concentrations decrease and alkane oxidation by NO3 is dominant. NO3 oxidation of alkenes is also competitive with oxidation by O3. The higher reactivity of alkenes with a multitude of oxidants contributes to their shorter atmospheric lifetimes.

In the troposphere O3 is formed by the interaction of NOx and VOC in a catalytic mechanism known as the HOx cycle, discussed below.

O3 is unreactive towards alkanes but is reactive towards alkenes producing a range of oxygenated compounds. Alkene ozonolysis proceeds via the 1,3 cyclo-addition of ozone

15 across the alkene double bond to form a primary ozonide. The ozonide undergoes a concerted cycloreversion (Criegee 1975) to form a carbonyl compound and a Criegee intermediate (CI). The primary ozonide is usually more than 200 kJ mol-1 below the enthalpy of the reactants, ozone and the alkene. Therefore, on decomposition of the chemically activated ozonide, there is sufficient energy for the newly formed CI to undergo unimolecular decomposition. The CIs readily dissociate into two radicals (HCO and OH) providing another OH source. CIs may also re-arrange via dioxirane into the bis-oxy carbonyl oxide and rearrange to form a carboxylic acid (Orzechowska and Paulson, 2005). Carboxylic acids are generated from CIs by reaction with water vapour forming hydroxy-hydroperoxide (HMHP) which degrades to form formic acid and water.

Excited state HMHP can also stabilise, lose OH and react with O2 to form the carboxylic acid and HO2, again contributing to the HOx cycle.

OH formation is typically initiated by the photolysis of O3 into molecular oxygen and an excited state atomic oxygen O(1D) (eq 1). O(1D) reacts with water vapour to form two hydroxyl radicals (eq 2). Typically, O(1D) preferentially deactivates to O(3P) (eq 3). In the marine boundary layer, where relative humidity (RH) is high, approximately 10% of O(1D) forms OH (Monks, 2005). OH concentrations are highest throughout the tropics where high RH and incident ultraviolet (UV) radiation dominate, as well as during the summer at higher and lower latitudes, where urban areas generate high levels of O3.

�� → �� + �( �) (1)

� �( �) + �� → � �� (2)

�( ��) + � → �( ��) + � (3)

In polluted air masses associated with urban environments, HONO photolysis can be a significant source of OH (e.g. Lee et al. 2016) as can the photolysis of aldehydes (RCHO) to form an alky radical (or hydrogen radical if R = H) and an organic moiety

HCO (Lin et al. 2012) (see HOx cycle below) (eq 4). Both H and HCO enter the oxidation chain by reaction with O2 to form HO2 (eq 5). The CO formed from HCO oxidation can then be oxidised to CO2 as is observed during VOC oxidation (Jenkin and Clemitshaw, 2000). Alkene ozonolysis is another source of OH radicals (as discussed previously).

�� → � + �� (4)

�� + �� → �� + �� (5)

Although OH global tropospheric average concentrations are approximately 0.06 parts per trillion (ppt) (106 molecules cm-3) (Prinn, 2003), its high reactivity causes OH to be the most significant atmospheric oxidising agent, reacting with many inorganic species and all organic compounds except chlorofluorocarbons (CFCs) and halons without H atoms (Atkinson, 2000). Table 2 summarises the global emission of the most important trace gases and their removal by OH.

Trace gas Global emission Removal by Removal by rate (Tg/yr) OH (%) OH (Tg/yr) CO 2800 85 2380 Isoprene 570 90 513 Methane 530 90 477

SO2 300 30 90

NO2 150 50 75 Terpenes 140 50 70 DMS 30 90 27 Ethane 20 90 18 Table 2. Trace gas emissions and OH contribution to their removal (Prinn, 2003)

The high reactivity of OH causes its lifetime to be short (� ~1 s) indicating its contribution to oxidation is local and not long ranged. The short lifetime of OH suggests that its concentration throughout the troposphere is highly variable and sensitive to its sources, sinks and environmental conditions. As most OH generating mechanisms are photochemical, OH is considered the most important oxidant during daylight hours. However, as previously mentioned, alkene ozonolysis can lead to OH formation which does not require incident light.

NO3

The Nitrate radical (NO3) is the most important night time oxidising agent. Whilst the reaction of NO3 with alkanes is orders of magnitude slower than the corresponding reaction with OH, the reaction of NO3 with alkenes at night can dominate the loss of VOCs. Stutz et al. (2010) showed than reaction with alkenes is responsible for more than

70% of the nocturnal loss of NO3. NO2 generated by NO oxidation can be further oxidised by O3 to form NO3 (eq 6). However as NO3 is photochemically labile, it readily degrades back to NOx and oxygen during daylight hours (eq 7, 8). The exact pathway is wavelength dependent.

�� + �� → �� + �� (6)

�� → �� + � (7)

�� → �� + �� (8)

When no incident light can photolyse NO3, its concentration builds. Without incident light

OH formation from the production of O3 cannot occur and NO2 photolysis stops, preventing further O3 formation. Net NO3 production removes NO2 and O3 from the atmosphere which further reduces O3 concentrations and OH production. NO3 reversibly reacts with NO2 to form nitrogen pentoxide N2O5 (eq 9), which is readily taken up by aqueous aerosol where it can react with H2O(aq) forming HNO3(aq), highlighting another heterogeneous NOx removal pathway. Another loss mechanism of NO3 is the reaction with NO to form NO2 (eq10).

�� + �� + � ⇌ �� + � (9)

�� + �� → �� (10)

- N2O5 is also capable of reacting with aqueous chloride Cl (aq) present in the aerosol to form nitryl chloride (ClNO2) (eq 11) which is an important source of chlorine radicals (Cl), another relevant atmospheric oxidant (eq 12).

− − �� + �� → �� + �� (11)

�� → �� + �� (12)

Cl The reaction of Cl with VOCs is analogous to those for OH (eq 13, 14, 15) but Cl oxidation is typically most relevant during the morning before OH concentrations have reached their maximum due to its photochemical release from Cl containing precursors.

�� + � → �(−�) + �� (13)

� + �� + � → �� + � (14)

�� + � → �� (15)

Cl radicals can abstract hydrogen to form hydrogen chloride and an alkyl radical (eq 13) that is oxidised by O2 (eq 14), or add to the reactant forming a radical that picks up O2 to form a chloro peroxy radical (eq 15).

The prevalence of chlorine radicals further disturbs the HOx cycle as Cl catalytically destroys O3 and subsequently forms chlorine monoxide radicals (ClO) that can react with

HO2 and NO2 to form HOCl and ClONO2 respectively, both of which can be taken up onto aerosol surfaces (Fig. 5) and act as a sink or reservoir of Cl and or NO2.

Fig. 5. Recycling of chlorine atoms. Pink indicates a source of Cl. Red indicates activation of the cycle due to

the presence of NOx. Inside the blue dashed circle represents the aqueous phase, outside represents the gas phase. Adapted from (Seinfeld & Pandis 1998).

Cl ions dissolved in aerosol particles originate from a variety of different sources, the most prevalent being NaCl typically from marine environments, although HCl and NH4Cl are other common Cl sources frequently measured in the atmosphere. The chloride is activated by one of several channels to produce gaseous chlorine containing species.

Typically this may be via the formation of ClNO2, Cl2 or HOCl, all of which exist in equilibrium between the gaseous and aerosol phase. ClNO3 and HCl are other gas phase sources of Cl. Formation of nitrated chloride compounds relies on a source of NO3 and is typically associated with anthropogenic behaviour, thus ClNO2 and ClONO2 are typically considered anthropogenic in origin, with ClONO2 considered important only for the stratosphere and upper troposphere. In some regions such as in China, Cl2 correlates well with SO2 and so is expected to originate from industrial power generation facilities that burn coal (Liu et al. 2017).

Measurements of such chlorinated species are relatively sparse. ClNO2 has been measured in the US (e.g. Kercher et al. 2009), mainland Europe (e.g. Phillips et al. 2012)

19 and China (e.g. Wang et al. 2017) with emphasis on the polluted marine boundary layer (e.g. Osthoff et al. 2008) as Cl- from sea salt aerosol was thought to be the most important source of Cl, although sampling inland locations implicates anthropogenic - sources of Cl as the major source (e.g. Mielke et al. 2011). Few studies of ClNO2 have been made in the UK even though it has a significant role in the oxidant chemistry of London (Bannan et al. 2015).

Biomass burning is another source of Cl- to the atmosphere including many ClVOCs (Lobert et al. 1999). Anthropogenic sources of Cl include waste burning (Lemieux et al. 2004), fugitive emission from waste sources (Christian et al. 2009) and water treatment (Ghernaout & Ghernaout 2010) and may be especially significant for indoor environments where bleach cleaning is common (Wong et al. 2017). Even fewer measurements of other potential precursors of Cl radical have been made and so present a current gap in current understanding of the impact Cl has on atmospheric oxidation.

1.2.3. Oxidation

The HOx cycle describes the relationship between OH/HO2 and their analogous counterparts RO/RO2. Fig. 6 summarises the major atmospheric oxidation pathway for

VOCs and demonstrates the system’s sensitivity to NOx concentrations. As described previously, OH is generated either by O3 photolysis, HONO photolysis, ozonolysis of alkenes or reduction of HO2 which originates from an aldehyde source.

The presence of NOx is a decisive factor in controlling the mechanistic route of radical propagation within the HOx cycle. Low and high NOx are qualitative expressions used to distinguish the two major routes.

15 3 -1 -1 In pristine environments devoid of NOx, the slow HO2 + O3 (~ 2x10 cm molecule s ) reaction is favourable and net O3 production is negative. The main mechanism of

OH/HO2 recycling is the oxidation of CO to CO2. OH abstraction or addition to the VOC creates a carbon centred radical which quickly picks up O2 to generate a peroxy radical

(RO2). The radical chain reactions are mainly terminated by the condensation reactions of RO2 and HO2.

As NOx is introduced into the system the HO2 + NO reaction is more dominant and so the photocatalytic cycle of O3 production is activated, generating more OH and so oxidising more VOC.

Under low NOx conditions, RO2 competes with HO2 to oxidise NO thus forming alkoxy radicals (RO). RO regenerates HO2 by its reaction with O2 to form carbonyl species. This

20 forms the basis of the HOx cycle and describes the recycling of OH with the consequence of VOC oxidation.

Under high NOx conditions, the reaction of oxygenated organic species such as RO2 with

NOx becomes favourable. RO2 reacts with NO to form organic nitrates. RO2 can reversibly react with NO2 to form alkyl peroxynitrates or if the R group contains a peroxy acyl group, then a reversible reaction forms peroxy acyl nitrates (PANs). Both alkyl peroxynitrates and PANs are thermally unstable at standard temperature and pressure so readily decompose in the low troposphere, typically with lifetimes of 0.01-1s and ~30 minutes respectively (Atkinson 2000). Although if lofted into the upper atmosphere, the formation of the organic nitrate is favourable as temperature decreases and thus they are more stable and can be transported over large distances. At this new location they can mix downwards and release RO2 and NOx. The formation of nitrites RONO from the RO + NO reaction is feasible although typically only under laboratory conditions (Orlando et al. 2003). Other organic nitrogen containing compounds formed under high NOx conditions are nitro compounds by the reaction of RO with NO2 e.g. the formation of nitrophenol from phenoxy radical (Yuan et al. 2015). The increasing NOx concentration increases the formation of O3 to the point where the NO + O3 reaction becomes competitive. Under these VOC limited conditions, O3 concentrations then reduce. Loss of

OH also becomes possible by reaction with NO2 to form nitric acid (HNO3) which is soluble and irreversibly lost through deposition.

Fig. 6. HOx cycle. Blue: reactions in a NOx free environment. Black: additional reactions in a NOx

environment. Red: additional reactions in a high NOx environment. Adapted from Atkinson (2000).

The termination products of RO/RO2 with NOx typically form material with lower vapour pressures than the starting material (Valorso et al. 2011), suggesting their increased propensity to condense. The increased oxygenated functionality also increases their hygroscopicity. The formation of epoxide functional groups increases reactive uptake as hydrolysis reactions on aqueous aerosol surfaces are favourable. The interactions and phase changes of these semi-volatile compounds are complex and constitute an entirely separate field of study.

Fig. 7. Auto-oxidation mechanism where intra-molecular hydrogen shifts propagate peroxy radical formation within the same compound

Whilst classical oxidation is typically thought to consist of bimolecular reactions, the unimolecular auto-oxidation mechanism is another atmospherically relevant oxidation pathway that was first demonstrated to be relevant for longer lived (τ = 30 - 60 s) peroxy and alkoxy radicals formed from biogenic VOC oxidation i.e. isoprene + OH (Crounse et al. 2011).

Auto-oxidation describes an intramolecular hydrogen shift to a radical centred oxygen atom, typically an alkoxy or peroxy radical centre. The hydrogen may be extracted from a C-H or O-H bond and so move the radical centre to the remaining carbon or oxygen atom from which the hydrogen was abstracted. If carbon centred, addition of O2 forms a new peroxy radical and radical propagation may continue (Fig. 7). Continued oxidation forms material with lower vapour pressures that more readily condense. More recently, this mechanism has been implicated in new particle formation observed over boreal forests (Ehn et al. 2014).

1.3. Air quality in the UK Improving poor air quality has been an aim of UK legislation for many decades. Bans on coal use for industrial applications had been enforced by various acts of parliament since the industrial revolution e.g. the Public Health Act (UK Government, 1875) but the Clean Air Act of 1956 is often cited as the beginning of modern day air quality legislation (United Kingdom Parliament, 1956). The aim of the Act was to reduce emissions of black soot and SO2 that formed the basis of ‘pea souper’ smog events occurring in the cold London winters when more coal was consumed for space heating and industrial activities. By the 1970s the prevalence of the vehicular combustion engine led to it being targeted by various acts of UK parliament and directives set by the European Commission (EC) to limit the emissions of carbon monoxide (CO), hydrocarbons and sulphur (European Commission 1970, United Kingdom Parliament 1974).

In the 1980s and early 1990s, The UK Parliament wrote into law various directives set by the EC as the Motor Fuel (Lead content of Petrol) Regulation (United Kingdom Parliament 1981), the Air Quality Standards Regulations (United Kingdom Parliament 1989), the Environmental Protection Act (United Kingdom Parliament 1990) and the Road Vehicles Regulations (United Kingdom Parliament 1991). These included mitigating against high concentrations of VOCs and NOx which are O3 precursors and major contributors to photochemical air pollution, first observed in 1940s Los Angeles.

This smog is formed by photo-oxidative processes, as described in the sections on O3 and the HOx cycle, which is a distinctly different form of air pollution to the ‘pea soupers’ of the late 19th and early 20th centuries that were caused by the direct emission of pollutants.

In the UK during the 1980s and early 1990s, the urban and rural automatic ambient air quality monitoring networks were formed in order to collect the required data to measure the state of UK air. These networks were amalgamated and now, the automatic urban rural network (AURN) comprises 127 sites across the UK measuring a range of air pollutants for which current legislation requires concentrations are limited. Other air quality networks maintained by the Department for Environment Food and Rural Affairs (DEFRA) include the (Non-automatic and) Automatic Hydrocarbon Network, Automatic London Network and many other non-automatic networks including but not limited to the Black Carbon Network and Toxic Organic Micro-Pollutants (TOMPs) Network.

In 1996 the EU air quality framework (96/62/EC European Council 1996) and its four daughter directives became the EU’s principal air quality mechanism, which was consolidated into the 2008 Ambient Air Quality Directive (2008/50/EC, European

Parliament & Council of the European Union 2008) and written into UK law in 2010. This, combined with the National Emissions Ceilings Regulations (Department for Environment Food & Rural Affairs 2002) comprises the current UK legal framework on air quality.

The 2008 Ambient Air Quality Directive (2008/50/EC) sets legally binding national air quality objectives (NAQOs) that detail the limits on measured concentrations of air pollutants and their permissible frequency of exceedances. These limits are relevant to either, human health or vegetation and ecosystems. The specific pollutants controlled by the legislation are lead, particulate matter (PM10, PM2.5), NO2, CO, O3, SO2 as well as the VOCs: poly-aromatic hydrocarbons (PAHs), benzene, 1,3-butadiene. As described, historically these pollutants have been identified as the most important components of air pollution and so many infrastructures exist to monitor their concentrations and emissions.

Currently in the UK pollutant concentrations have been steadily reducing since 1970

(Department for Environment Food & Rural Affairs 2014) (Fig. 8, 9). For instance NO2,

PM2.5 and PM10 concentrations have demonstrated a moderate reduction and the longevity of episodic high pollution events is also reducing, although the trend appears to be less pronounced at the end of the sample period and ozone concentrations remain constant at rural background and increase at urban background locations (Department for Environment Food & Rural Affairs 2018).

-3 Fig. 8. Comparison of annual average levels of NO2, O3, PM10 and PM2.5 (μg m ) from 1987-2017 reproduced from Department for Environment Food & Rural Affairs (2018). Decreasing trends are observed

for NO2 and PM measurements but O3 remains static.

Whilst the success stories of UK air pollution reductions reflect progress made towards achieving cleaner air, interventions to decrease polluting vehicle emissions such as CO

25 and VOCs mean NOx reduction has not occurred as predicted (Carslaw et al. 2016). This is significant as transport accounted for 49% of UK NOx emissions in 2016 (Department for Environment Food & Rural Affairs 2016). The use of coal for power generation and increasing popularity of diesel cars during the early 2000’s due to government incentives also prevented greater reduction in NOx emissions and still contribute significantly to the

UK NOx budget.

The result is the UK continues to breach the European Parliament air quality directive limits for NO2 of more than 18 recorded exceedances of NO2 concentrations greater than 200 µg m-3 (European Parliament & Council of the European Union 2008) and so faces on-going legal action from the European Commission (EC). This limit is breached quickly within the year, for instance the legal limit was reached in London before the end of January 2018 further emphasising the current inability to limit highly polluting incidences (New Scientist 2018). This highlights the complexity of the air quality problem. As with many components of the earth system, an isolated approach to perturbing a single component often results in undesirable consequences for other components elsewhere in the system, suggesting that from this point, meaningful air quality reduction requires a holistic approach. For example, NO2 reduction is the current focus of much air quality legislation, however the impact this reduction will have on O3, a trace gas whose production increases at lower levels of NOx and has its own NAQO, has yet to be confronted and might even be visible in Fig. 8.

Fig. 9. UK VOC emissions (kilotonnes) from 1970 to 2016 split by the six largest sectors. Emissions have reduced by a third from 1991 to 2016 with the biggest reductions made in the transport and extraction and distribution of fossil fuels sectors (National Atmospheric Emissions Inventory 2016).

Since 1970, emissions of NMVOCs peaked in 1991 at 2,850 kt yr-1, dominated by the transport and extraction and distribution of fossil fuels sectors (Fig. 9). By 2016 emissions from these sectors have reduced drastically such that emission from transport is now the smallest of the six major contributors.

In the UK, exceedances for NO2 and PM occur much more frequently in urban and industrial centres (Department for Environment Food & Rural Affairs 2018) and so in order to gain control of air quality, this is where focus must be applied in the understanding of processes that cause the exceedances and where the introduction of new technologies, policies and abatement strategies should be most effective. This requires strong fundamental understanding of the dynamical, physical and chemical processing of air pollutants and trace gases that occurs in urban centres.

Whilst the NAQOs make good legislative targets, they do not engage with the full range of contributors to poor air quality. For example, the NAQOs for PM only specify a mass concentration not to be exceeded. They do not detail the composition of the PM, which as previously discussed, is an important factor regarding toxicity.

Benzene is one of few anthropogenic VOCs with a defined NAQO (5.00 µg m-3 as an annual mean) that is routinely monitored by the automatic and non-automatic hydrocarbon networks (Department for Environment Food & Rural Affairs 2013). Whilst the primary reason for the monitoring is to assess the concentration of this carcinogenic material in line with the NAQO, as previously described, the oxidation of VOCs such as benzene leads to the formation of secondary material such as SOA, the constituents of which are known toxins (e.g. Sekler et al. 2004).

There are many more trace gases present in ambient urban air that contribute to poor air quality that do not have any NAQOs, primarily because these species are more costly and difficult to measure. Some of these, such as hydrogen cyanide (HCN), are emitted by many different processes such as vehicular emission, domestic biomass burning or waste burning. Bonfire night (Guy Fawkes Night) represents one such large scale anthropogenic burning emission source to the UK atmosphere that is known to be highly polluting (Harrison & Shallcross 2011). Many different fuel types are burned during the event, evolving a rich mixture of primarily emitted trace gases. The full extent of anthropogenic trace gas emission from this event and the toxicity burden placed on the exposed population are largely unknown and represent major uncertainty in terms of primary emissions to the urban UK atmosphere. Many directly emitted species (primary species) have yet to be identified, in part due to their high reactivities and short lifetimes,

27 rendering many measurement techniques inappropriate leaving significant gaps in current knowledge.

1.4. Measuring the chemical composition of the urban atmosphere Measurements of trace gases in urban centres can be made by either in situ measurements or remote sensing techniques. Satellite retrievals provide good spatial and temporal coverage, however their poor resolution, especially over a city scale, make it difficult to assess finer scale variability or spatial distributions at different altitudes (Hilboll et al. 2013). The retrieval itself is not a direct measure of concentration but a parameterisation fitted to spectroscopic data which can suffer from different interferences e.g. OClO on the measurements of HCHO (Hewson et al. 2013).

In situ measurements of trace gas air pollutants are the most accurate way of determining ambient concentrations with a high temporal resolution, although the spatial coverage of such measurements is limiting. Placing the instrument on a mobile platform such as a ship or aircraft increases the spatial resolution of the measurement, although this advantage may be constrained by the lifetime and hence spatial extent of the trace gas being detected. These types of mobile platform are also limited in their spatial resolution, and for studies on the urban environment, may only detail bulk urban outflow rather than fine scale monitoring which will include the effects of micrometeorology.

Continuous deployment of instrumentation to increase temporal sampling is dependent on how practical and intensive the measurement technique is. For example, the Thermo

Scientific i42 NOx analyser measures the emission of radiation from the relaxation of excited state NO2 formed from the reaction of NO from the ambient air and internally generated O3. This instrument has been refined to a point where continuous monitoring of this pollutant is possible. Baseline procedures are automated and the drift in accuracy is minimal so calibrations are required infrequently. The instrumentation does not require any consumables e.g. a carrier gas and so can be left unattended for weeks.

For less well studied trace gases, whose urban concentrations may be orders of magnitude lower than a pollutant such as NOx, more intensive instrumentation may be required. This is most likely to be a chromatographic, spectroscopic or spectrometric technique. The justification for using any of these techniques is based on the properties of the analyte in question. For short lived species it may be necessary to perform the analysis online, i.e. at the time of collection. For example, laser induced fluorescence (LIF) is a spectroscopic technique typically used to measure extremely short lived species such as OH (e.g. Creasey et al. 1997). The high reactivity of OH does not permit

28 it to be stored, so detection must take place at the time of sampling. Conversely if the species are long lived, it may be well mixed and represent a regional concentration representative of a longer time period (e.g. Cox et al. 2003). Analyses can be performed offline where whole air samples (WAS) containing the analyte are analysed at a later date (e.g. Dyke et al. 1997). This analysis is commonly a chromatographic technique where retention time within a column is the basis for analyte separation. Gas chromatographs can be deployed for online sampling, however the response time of the instrument may be too low to capture true atmospheric variability on short time scales and so requires a reduced sampling frequency (Dunmore et al. 2015).

1.4.1. Mass spectrometry Mass spectrometry is an online measurement technique capable of detecting many species simultaneously. Its fast response time, high reproducibility and high sensitivity make it a useful technique for the detection of small concentrations of individual compounds in complex mixtures at a high temporal resolution. Whilst dominantly employed in the biosciences and proteomics industries as a laboratory based analytical tool, its versatility as a measurement technique allow it to be used in ambient sampling and detection and for deployment on a variety of different measurement platforms including aircraft (Le Breton et al. 2012) and ships (Buffaloe et al. 2014). The atmospheric pressure inlet to a mass spectrometer allows for direct sampling of ambient air with little sample preparation.

The major disadvantage of mass spectrometry is its inability to distinguish isomers, which is problematic for ambient trace gas measurements where many organic molecules are present. Also, mass spectrometers often require consumable materials e.g. a carrier gas or reagent ion gas with which the sample is mixed, making this technique difficult to deploy long term.

Broadly, mass spectrometry works by ionising compounds of interest within a sample thus enabling their trajectory and energy profile through the instrument to be manipulated by electric fields with the intention of collimating and homogenising the ions’ energy. The electric fields transmit the ions through the instrument where they are ultimately detected. Two major features of this process that can be adapted in order to suit the intended purpose of the measurements are the ion separation technique and the method of ionisation.

Ion separation Atmospheric sampling mass spectrometers typically employ either a quadrupole mass analyser (QMA) or time of flight (ToF) region as a means of separating the ions within

29 the transmitted ion packet. QMAs scan through a range of frequencies and so adjust the angle of the ion beam such that the analytes with a specific mass to charge ratio (m/z) hit the detector. Time of flight (ToF) instruments differ from their quadrupole counterparts in that the separation of ions in the ion packet is defined by the travel time of the ions accelerated by a uniform electric field.

The scanning feature of the QMA limits the number of chosen analytes to measure as the duty cycle of the instrument increases and so there is a trade-off between sensitivity and detected number of analytes. With the ToF, ion-packets are accelerated by a constant voltage and travel a uniform distance. This non selective method of separation ensures all ions are detected. Detection systems typically employ micro-channel plates or electron multipliers, both of which are designed to cause a cascade of electrons from the impact of an ion on the resistive surface of the detector. The electrical signal produced from this interaction then goes onto amplification and data acquisition.

Ionisation The method of ionisation broadly defines the selectivity of the mass spectrometer. The aerosol mass spectrometer (AMS), used to detect certain fractions of particulate matter 2- - - (SO4 , NO3 , organic, Cl ) less than 1 µm in aerodynamic diameter (PM1.0), uses electron impact (EI) to heavily fragment the target aerosol from which data products can be calculated. Whilst the reproducibility of the mass spectrum generated through EI is high, much chemical information is lost e.g. molecular composition. This makes softer ionisation techniques such as electro spray ionisation (ESI) and chemical ionisation mass spectrometry (CIMS) ideal methods of ionisation for the measurement of individual trace gases as compositional information is retained by preserving the original structure of the target compound.

ESI describes the application of a voltage to a salt solution in order to ionise a sample. The solution is atomised and mixes with the ambient sample. The soluble gases from the sample are accommodated into the atomised solution. This mixture is passed through a denuder to dry the droplets. As the droplets dry, the particles shrink and the dissolved ions are ejected by Coulomb fission producing much smaller droplets. These droplets may further evaporate and fission to form the gas phase adducts, or the adducts themselves desorb from the droplet. The disadvantage of ESI is that multiple charging of the sample leads to confusion in identification in the mass spectrum (Lu et al. 2015), which becomes increasingly more difficult when complex mixtures are sampled. It is for this reason that this technique is more commonly used for proteomic studies and not atmospheric sampling (Fenn et al. 1989).

CIMS is a very soft ionisation technique that singly charges the analyte. It relies on ionisation by the interaction of a reagent gas with a radiation source (e.g. 210Po α emission, 241Am β emission, X-ray emission, corona discharge). This ionised reagent gas mixes with ambient air where selective binding of the reagent ion and ambient molecules takes place.

Reagent ion Example of detected compound Reference

NO+ VOCs (alkanes, aldehydes, ketones, nitriles) Koss et al. (2017)

- SIF5 HNO3 Kita et al. (2006)

- SIF6 HO2NO2 Kim et al. (2007)

CO3 SO2 Speidel et al. (2007)

- SO2Cl HONO Hirokawa et al. (2009)

- NO3(HNO3)n H2SO4, VOCs (highly oxygenated) Rissanen et al. (2014)

+ H3O VOCs (hydrocarbons, aldehydes, ketones) Yuan et al. (2016)

- C2H3O2 OVOCs Brophy & Farmer (2015) I- Inorganic halogens, OVOCs (organic acids, APNs) Lee et al. (2014)

SO2 OH (as H2SO4) Muller et al. (2018) Table 3. Ionisation schemes and detectable gases used for in situ measurements. To further discriminate target analytes in an ambient air sample, different chemical ionisation schemes provide selectivity to discriminate the detection of different molecules. Table 3 summarises common reagent ions and the atmospheric compounds they are used to detect.

- Nitrate (NO3 ) - Nitrate (NO3 ) and clusters thereof are used as a reagent ion to detect atmospheric sulphuric acid by proton abstraction (Kürten et al. 2012), amines (e.g. Simon et al. 2016) and highly oxidised organic molecules (HOMs) (e.g. Mentel et al. 2015). This system is capable of detecting clustered organic molecules from dimers up to pentamers where the size of clusters begins to approach those with a mobility diameter ≥ 1.5 nm (Molteni et al. 241 - 2016). HNO3 is ionised by a Am β particle source to generate the NO3 . To minimise sample wall losses, the sample flow is delivered to the instrument inside a co-axial sheath flow in an Eisele type inlet (Eisele & Tanner 1993).

For organic molecule detection, as the nitrate cluster contains only multiples of N and O, it can be difficult to deconvolve the contribution of the reagent ion formula to that of the cluster where organic nitrogen is present. Whilst applying principles of mass spectrometric formula solving, such as the nitrogen rule, may help constrain peak identification, at higher masses the contribution from different configurations of clustered reagent ions becomes difficult to distinguish. The nitrogen rule takes advantage of the

31 odd proton number of nitrogen compared with the even proton numbers for C, H and O, making compounds that contain odd numbers of hydrogen atoms identifiable in a mass spectrum at odd m/z.

Proton transfer (H+) + Proton transfer makes use of H2O, specifically the hydronium ion H3O , as source of protons to donate H+ to less acidic species. This occurs within a low pressure drift tube in front of the ion molecule reaction region (IMR). Species commonly detected by proton transfer are unsaturated (Koss et al. 2017), aromatic (Hansel et al. 1999), low oxidation state (Sekimoto et al. 2017) and low molecular weight VOCs (Brilli et al. 2014), many of - + which originate from direct emission. Similar to the NO3 ionisation scheme, if H3O and

H2O clusters are formed, identification of molecular formulae is confused. This is typically overcome by having a large dissociation energy to fragment the clusters, but results in a reduced instrumental sensitivity (de Gouw & Warneke 2007).

Iodide (I-) The I- ionisation scheme demonstrates sensitivity towards a large range of multifunctional oxygenated VOCs including organic acids, as well as many inorganic species. I- ions are 210 generated by passing methyl iodide (CH3I) through a Po α source, which are mixed with ambient air in the IMR before they are differentially pumped through the instrument - to the mass spectrometer. The large negative mass defect of I (δm = -0.096 Th) makes adduct identification much less ambiguous when compared with Nitrate or PTR spectra. The sensitivity of the instrument towards the cluster is proportional to the binding enthalpy of the cluster (Iyer et al. 2016). I- is a weak gas phase base so proton abstraction and electron donation are broadly insignificant (Lee et al. 2014) yet the binding enthalpy of I- to a great number of adducts is large enough that cluster formation is favourable, ensuring the selectivity of I- is broad. The presence of water in the IMR is known to affect the sensitivity of the instrument by either facilitating or supressing adduct - formation with hydrated iodide ions (I.H2O ). It is suggested that the presence of the clustered water molecule generally acts to stabilise adduct formation by providing extra vibrational modes preventing cluster dissociation (Iyer et al. 2016). The ratio of I- to - I.H2O is controlled by tuning the low voltages (<100 V) applied to the quadrupoles and ion optics of the SSQ and BSQ to maximise ion transmission and prevent declustering.

For these reasons and the high sensitivities, stabilities, reproducibility and ease of operation, the I- ionisation scheme has been used extensively with quadrupole instruments for field campaigns on both the ground at static sites (Hoker et al. 2015) as well as on aircraft platforms (le Breton et al. 2014). I- has been used to measure many

32 atmospheric phenomena such as; halogen oxidation in the tropics (Le Breton et al. 2017); organic acid production from direct emission (Bannan et al. 2014) and secondary production in urban centres (Bannan et al. 2017) and over boreal forests (Jones et al. 2014); hydrogen cyanide (HCN) as a marker for North American biomass burning over the Atlantic (Le Breton et al. 2013); a range of inorganic nitrogen compounds such as nitric acid (Le Breton et al. 2014), nitrogen pentoxide (Le Breton et al. 2014) and nitryl chloride (ClNO2) (Bannan et al. 2017).

1.4.2. Data acquisition and analysis

Aerodyne ToF-CIMS The time of flight chemical ionisation mass spectrometer developed by Aerodyne Inc. (H- ToF-CIMS) (Fig. 10) has been used extensively with a variety of different ionisation schemes for different atmospheric measurement applications. It consists of a reduced pressure (102 mbar) ion molecule reaction (IMR) region where the reagent gas and sample mix to form adducts. The ions move through a critical orifice (300-1000 µm) into the short segmented quadrupole (SSQ) held at 1.5-2.0 mbar. The ions then move through another critical orifice into the back segmented quadrupole (BSQ) held at 10-3 mbar.

The SSQ and BSQ both have a direct current (DC) and radio frequency (RF) applied to pairs of opposite rods that alternate at a high frequency. This has the effect of collimating the transmitted ion beam in a helical path in order to energetically homogenise the ions. This occurs through collisional cooling of the ion beam with the ambient gas within each chamber.

Fig. 10. Schematic of the ToF-CIMS adapted from Bertram et al. (2011). The V shaped trajectory of the ions in the flight tube denotes the instrument as a V-ToF as opposed to a W-ToF which has a reflectron at both ends of the flight tube.

The ions then pass into the primary beam (PB, held at 10-5 mbar) where they travel through focusing lenses to homogenise ion acceleration and reduce the amplitude of the x and y components of the helical ion beam. An orthogonal DC pulse of opposite charge to the charge on the ions extracts ion packets from the beam and pushes them into the ToF flight tube (10-7 mbar). A reflectron at the end of the flight tube deflects the ions back in the direction of their origin where they hit a microchannel plate (MCP) detector.

The increased resolution afforded to the ToF over the quadrupole is due to the ion separation technique. The uniformly applied electric field of the orthogonal DC pulse accelerates all ions through the flight tube. This means the time taken for the individual ions to reach the detector is a function of its mass to charge ratio (eq 16).

� � � = �√( ) , where � = (16) � √� �

Where t is the time to reach the detector, m/q is the mass to charge ratio of the ion, d is a constant distance to the detector and V is a constant voltage supplied to the ion.

The reflectron at the end of the flight tube effectively doubles the path length and so increases the separation between ions of different m/z within the same ion beam and so increases the resolution. Ions of the same m/z with more kinetic energy penetrate further into the flight tube and so travel a longer path than their lower energy counterparts. This increases the arrival time of the high energy ions to match more closely with the arrival time of the lower energy ions, reducing the peak width and further increasing the resolution.

A comparison between the University of Manchester quadrupole CIMS, constructed and designed by the Georgia Institute of Technology (Nowak et al. 2007) and the Manchester aerodyne ToF-CIMS demonstrates the difference in identification, accurate quantification and fewer interferences due to increasing the resolution. For the quadrupole CIMS, the species’ it is able to detect with high confidence requires the signals’ relative isolation from interference peaks. This is necessary as the typical resolution of such an instrument is m/dm < 1000 indicating, at e.g. m/z 173, a peak separation of 0.173 Da, or less than approximately 17% would be required to fully resolve the peaks.

Fig. 11. Comparison of ToF and quadrupole mass spectra (unpublished). The resolution of the ToF instrument (black) highlights some of the interferences present on the quadrupole (blue).

For different molecules of the same unit mass e.g.CH2O2 and NO2 that are detectable with the same reagent ion, the quadrupole mass analyser cannot provide the required resolution to de-convolve signals. A ToF CIMS is capable of m/dm = 4000 and so the signals for some of these overlapping masses become possible to de-convolve (Fig. 11). An example of the increased resolving power of a ToF instrument vs. a quadrupole instrument can be demonstrated for the measurement of formic acid. The mono-isotopic mass of formic acid (HCOOH) is 172.909957 Da. The ToF-CIMS measures the HCOOH signal at 172.91130 Da indicating an accuracy of 8 ppm. As the quadrupole instrument is limited in its mass selection by the step wise increments of the QMA, it is only able to measure in units of ½ Da. The Quadrupole measurement at 173.5000 gives an accuracy of 3412 ppm. The increased resolving power of the ToF gives a peak width of 0.3 Da whereas the quadrupole peak width can be greater than 1 Da.

The data acquisition software (ToFDAQ, Tofwerk 226 AG, Aerodyne Research, Inc.) records the spectra at 222 kHz upon which statistical procedures filter incorrectly recorded spectra e.g. false positive ion counts, to collect an accurate series at typically 1Hz time resolution. The identification of peaks within the spectra relies on an accurately described peak shapes and mass axis calibration which are performed as part of the post processing of the raw spectra (Tofware Version 2.5.11, Tofwerk 226 AG, Aerodyne Research, Inc.). Representative peaks from a spectrum (5 – 10 peaks) are chosen to approximate a general peak shape for that data set. These are typically of high signal such as those found for the reagent ion as these contain the greatest number of ion counts. The average peak shape derived from those representative peaks are applied to all peaks found within the spectrum. This is required as the peaks of low signal are difficult to accurately assign as instrumental noise can make a sizeable contribution to their intensity and thus impact the shape of the peak.

The peak width is calculated as the lowest resolution (full-width of the peak at half its maximum) found for a given m/z. As the m/z range increases, peak width broadens which is a consequence of the reduced mass transmission efficiency of the instrument. Typically, mass transmission is optimised to select the m/z range of interest and is set low. The mass calibration uses representative peaks such as those from the reagent ions as well as ubiquitous instrumental background peaks to scale the m/z axis. These peaks are usually fit with an error of < 3 ppm. This allows for the accurate fitting and deconvolution of multiple peaks within the same unit mass signal anywhere on the m/z axis within the boundaries of the mass calibration peaks. In the iodide spectrum and in - this work, the highest signal used is that for I3 at m/z 381. Identified peaks are then integrated to provide a time series.

The large mass range and high resolution of the ToF-CIMS with the iodide reagent ion provides the ability to detect 102-103 species. A given unit mass signal of a ToF generated mass spectrum can contain multiple signals. The wealth of information is vast and largely unknown with the only a posteriori knowledge originating from quadrupole iodide CIMS studies of which the limitations have been discussed. The identification and assignment of peaks is a manual process and cannot currently be performed by the software with good accuracy.It is laborious, time consuming and potentially ambiguous as overlapping peaks can often be assigned formulae incorrectly. This is especially true for the software that has no knowledge of realistic chemical formulae or the selectivity of the reagent ion. Manually assigning a chemical composition to the fitted peaks can be constrained by knowledge of the selectivity of the reagent ion, the presence of isotopes at other unit masses, the expected behaviour of the compound in relation to the time and space that is being studied, and the mass defect of the suggested formula. Therefore in house methods were developed for identification, post processing and quantification as part of this study, primarily using the following principles as this functionality is not available in the commercially available analysis software (Tofware Version 2.5.11, Tofwerk 226 AG, Aerodyne Research, Inc.).

Peak identification by mass defect The mass defect of a compound is the difference between its mono-isotopic mass and its unit mass. By definition 12C has a mono-isotopic mass of 12.0000 Da and so has a mass defect equal to 0. All other elements present a difference in unit mass and monoisotopic exact mass. The mass defect of iodide is one of the largest negative values of all the elements (-0.095527 Da). This causes its appearance in a mass spectrum to be heavily shifted to the left of unit mass. This is true for other atmospherically relevant elements such as the other halogens (F, -0.001597; Cl, -0.031147; Br, 0.081663) and oxygen (-

0.005085 Da). Hydrogen and nitrogen demonstrate a positive mass defect (or mass excess of 0.007825 Da and 0.003074 Da respectively) and so the presence of these elements shifts the signal to the right of unit mass. Increasing the number of elements to a compound has an additive effect on its mass defect, therefore halogenated compounds detected as an I- adduct have a large negative mass defect, whereas a compound with a large H:C ratio that is not an I- adduct has a larger positive mass defect (excess) (Fig. 12).

Fig. 12. Representation of four peaks at a unit mass of 250 m/z. The effect of elemental composition on peak position is described by the difference of the exact peak position from unit mass (not to scale). The Iodide ion gives the greatest mass defect and hydrogen gives the greatest mass excess. An example of this can be seen in paper 3.

Where the mass defect uses 12C as the reference isotope to which all other isotope masses are relatively measured, the Kendrick mass of a species redefines the reference moiety typically to CH2 but can be extended to any other atom or set of atoms e.g. O,

CO, CO2. This indicates that increasing the number of CH2 units in the analyte will present no difference in Kendrick mass defect (KMD) as KMDCH2=0. Where many compounds share the same Kendrick mass defect it is surmised that they contain n

Kendrick base units e.g. nCH2. Therefore the atmospherically relevant compound acetic acid CH3CO2H has the same Kendrick mass defect as propionic acid (CH3CH2CO2H) as the increase in elements from the former to the latter is 1 Kendrick base unit of CH2.The normalisation of the reference isotope can be used on various moieties e.g. CO, O2, CO2 to create relationships of unique signals identified in the mass spectrum that are related to other peaks by a series of Kendrick bases.

Kendrick mass defect is used in environmental mass spectrometry (Hughey et al. 2001) and also in atmospheric mass spectrometry (Junninen et al. 2010). However, as the technique relies on ultra-high resolutions (typically m/dm > 104-105) and so it is problematic to apply to the data gathered by the ToF-CIMS at a resolution of 4x103 (Marshall & Rodgers 2004). The associated error in peak assignments with the ToF- CIMS is large giving plenty of potential overlap with other peaks. However, for another peak to be considered a Kendrick mass defect match and therefore be part of the series, the potential matching peak must have the same Kendrick mass defect (within error) and

37 be present at the correct unit mass n Kendrick bases apart from those species in the same series. Also unlike in a field such as oil exploration where non selective ionisation of extremely complex mixtures are desired (Marshall & Rodgers 2004), the selectivity of the reagent ion limits the number of potential overlaps. This provides some constraint on the potential inclusion of peaks within a Kendrick mass defect series.

A methodology has been developed in this study to use these principles to assign peaks in the ToF-CIMS spectrum. A peak list of known and unknown assignments is extracted from the post processing software. Mass defects and Kendrick mass defects of user specified Kendrick bases are calculated. If an unknown peak is n Kendrick bases apart from a known peak and its Kendrick mass defect is within error of the known peak’s Kendrick mass defect, it is labelled a match. Using each Kendrick base the potential formulas for this peak are calculated. The reason known peaks are used as a best guess starting point to which unknown peaks are matched against is that in some instances, - e.g. with increasing CH2 units, the functionality and so selectivity of the I to that compound in series is more likely to be preserved. Of course this is not always the case e.g. with the series HONO, HNO3 and HO2NO2 where nO units are added each time.

To assess the validity of the proposed formulae in terms of atomic content, the ChemSpider database (Royal Society of Chemistry 2015) is queried to return valid molecules that may further aid identification. The returned entry is tested to ensure the formula describes a molecule that is: singular, charge neutral, covalently bonded and not a hydrocarbon (in line with the selectivity criteria for iodide CIMS). Fig. 13 shows the successful identification of unknown peaks from three selected peaks as a result of the methodology applied to a subsection of a dataset obtained from a chamber study, consisting of a series of CHON compounds that are nCH2, nO and nN units apart. The methodology developed in this study using this technique has greatly enhanced the identification of species for the scientific output of this work.

Fig. 13. Demonstration of the KMD matching programme on a subset of chamber data with the reagent ion removed. (a) Mass defect. (b) Kendrick mass defect (CH2). (c) Kendrick mass defect (O). (d) Mass spectrum, inset is a portion of the mass spectrum. 27 signals are present in the spectrum but isobars reduce the number of observed peaks to 15. The clusters of signal in the mass spectrum appear regular with a +n14 repeating unit but does not provide information on what moiety is contributing to the increase. The three

identified species C2H6ON2, C3H8ON and C4H10O inset are an example of isobars. In the series, CH2 units increase as N units decrease but both have an integer mass of 14 Da making the identification of the + n14 peaks in the spectrum ambiguous. Plotting the Kendrick mass defects of the identified species exhibits the relationship between their exact peaks at any one given unit mass and their unidentified counterparts +n14 m/z.

This methodology cannot be used conclusively for the assignment of peaks as the resolution of the instrument is too low, however in conjunction with time series correlations and a greater understanding of the sources and processes that lead to the presence of the detected compounds within the spectra, it is possible to use this technique to infer the identity of the signal. For example, the iodide ToF-CIMS is

39 sensitive to HONO and HNO3 and by using this technique to interrogate the KMDO space, is shown to detect HO2NO2 (Fig. 14).

Fig. 14. Demonstrating the Kendrick mass defect with the Kendrick base O as a function of m/z for a sample of identified peaks in a mass spectrum from an urban ambient dataset (see paper1). The effect of the large - negative defect of I can be seen as a shift down in KMDO for points where m/z > 127. Inset is a subsection - - - of the figure where I.HONO and I.HNO3 (red) are used to assign the unidentified peak I.HO2NO2 (blue). No - other peak was identified at the unit mass of I.HO2NO2 with a KMDO wtihin error to match. m/dm = 3200 at 200 m/z.

Calibration The tofware post processed output produces time series’ in units of ions per second (Hz) that describe the number of ion-detector hits per second at that specific m/z. This provides a qualitative assessment of the amount of analyte detected. To quantify the amount detected as a concentration, the instrument must be calibrated such that a response of a particular magnitude can be ascribed to a known quantity of calibrant that is introduced. The method of introduction will depend on the properties of the analyte in question. Where the material is readily available and gaseous, a gas mixture can be made by serial dilution of a stock material. This is performed in the laboratory by introducing a range of known quantities of calibrant into the instrument.

Fig. 15. Example of chlorine (Cl2) raw signal from ToF-CIMS calibration. The largest quantity of Cl2 is detected first and after steady state is reached the flow is reduced to deliver less calibrant.

For example, a known pressure of Cl2 from an Aldrich cylinder (>99.5%) is introduced into a 1 L cylinder via a manifold. The cylinder is pressurised to ~3 atm with N2 to create a 1% Cl2 mixture. This is repeated two or three more times to make a 1ppm or 10 ppb calibration mixture. This mixture is then further diluted at the introduction to the instrument by setting the calibrant flow and an N2 carrier flow (Fig. 15). By varying the calibrant flow and observing the instrument response, a calibration factor is calculated from the linear least squared fit (Fig. 16).

Fig. 16. Example of chlorine (Cl2) calibration curve. The gradient of the curve is the instrument response to the introduction of calibrant (in counts per unit of concentration e.g. ppt). The coefficient a gives the intercept and coefficient b gives the gradient.

The laboratory manifold is a static instrument; calibration mixes were not possible in the field. Therefore as part of this study, a field deployable manifold was developed and utilised, allowing gas mixtures to be made during field work projects depending on the requirements of the measurements being made.

Relative humidity dependence As discussed previously, the sensitivity of a number of I- adducts is water dependent. The presence of water vapour in the IMR either disrupts or enhances adduct formation and the transmission of water adducts through the instrument is affected by the tuning of the instrument. Where necessary, water calibrations are made to ensure the change in sensitivity is accounted for (e.g. Fig. 17).

With the Manchester quadrupole CIMS measurements and past published results, all reported species have been calibrated for either by synthesis or by being commercially available. This is now not possible for the ToF-CIMS given the wide range and mass range of species that are now measurable. As part of this study therefore, developments have been made as well as automatic analysis protocols to accurately account for variations in sensitivity as a result in changes in the RH in the IMR, using the principles illustrated in Fig. 17.

Fig. 17. Example of water calibration curve for I.Cl2. the enhancement in signal when humidified relative to that when dry as a function of the partial pressure of water (PH2O) in the IMR. The interquartile range of PH2O in the IMR is very narrow and so sensitivity changes to this species as a result of PH2O are small (± 0.1).

Cross calibration by another instrument is another calibration methodology. For example, 2 - there is a strong agreement (R = 0.93) between the I2.NO2 adduct as measured on the

ToF-CIMS and the NO2 measurement of a Thermo Scientific 42i NOx analyser at the Whitworth Observatory (Fig. 18). Here we observed a non-linear increase in signal from the ToF-CIMS, indicating its susceptibility to interference at higher concentrations is greater than that of the NOx analyser. Chemiluminescence techniques used for the

42 detection of NOx species, like that employed by the Thermo Scientific 42i NOx analyser, are known to overestimate NO2 concentrations through the additional contribution of NOy

(Reed et al. 2016). The exact cause is unclear but is likely fragmentation of NOy in the IMR. At low concentrations, the cross calibration factor is 1.5 Hz ppb-1.

- Fig. 18. NO2 measured at the Whitworth observatory overlayed with I2.NO2 measured on the ToF-CIMS.

NO2 is overestimated by the ToF-CIMS at high concentrations. This is potentially due to the degredation of

NOy species in the IMR.

Where the calibrant in question is not readily available and must be synthesised in situ, the concentration of calibrant delivered to the instrument must be verified by another technique for which a calibration has already been established. If that species is not directly quantifiable, titration with a quantifiable excess of a species can often be used if the stoichiometric reaction is known. This is the case for N2O5 and ClNO2. Neither can be bought directly as they are unstable compounds at room temperature. N2O5 exists in equilibrium with NO2 and NO3 and ClNO2 photolyses to chlorine atoms and NO2.

N2O5 synthesis is performed by passing excess O3 through a volume of NO2 to generate

NO3 and N2O5 (Le Breton et al. 2014). The outflow from this reaction is cooled in a cold trap where the N2O5 is frozen. The trap is heated to room temperature and O3 is flowed through it again, oxidising any excess NO2 and NO3 to form a pure sample of N2O5. This cooling and O3 flow cycle is performed several times to ensure a high purity. The N2O5 is then sampled by the CIMS instrument to determine the instrument response. To find the concentration of N2O5 flowed to the CIMS, the flow is diverted through a heated line to decompose the N2O5 to NO2 + NO3. The NO2 is quantified by a Thermo Scientific 42i

NOx analyser and thus the N2O5 concentration is also determined.

The N2O5 can be passed over a salt (NaCl) slurry where it dissolves and decomposes in - - the aqueous phase to NO2 and NO3 . Cl reacts with the NO2 to form ClNO2 which degasses from the heterogeneous phase and can be passed into the CIMS. The drop in

N2O5 signal is directly proportional (due to the 1:1 stoichiometry) to the increase in ClNO2 signal and so a relative calibration factor can be found.

Errors and uncertainty

All calibration methodologies have an associated error. Where gas mixtures are made from the pure substance by the serial dilution method, the systematic error is controlled by the trade-off between the number of serial dilutions vs. the lowest pressures of calibrants used during the serial dilution. The pressure measurement accuracy reduces from 1% to 10% when operating below 5 torr. This requires measurements of calibrant pressure to be greater than 5 torr to reduce error, however as more material is retained, more serial dilutions are required to make gas mixtures of low concentration. The systematic error associated with a single serial dilution step is 1% but increases the chance of random error from increased operation. Typically for a standard three cycle dilution, the systematic error in the final mixture is 5%. This may increase when using less volatile materials that are difficult to measure a partial pressure of more than 5 torr in the gas phase.

For those calibrants synthesised online such as N2O5 and ClNO2 the systematic error is also low (5%). The largest source of error in this calibration is the synthesis of N2O5 and its conversion to ClNO2. The conversion to ClNO2 is assumed to be 100% and is based on previous work, however this conversion efficiency can be as low as 60% (Hoffman et al. 2003; Roberts et al. 2008). The purity of the synthesised N2O5 is difficult to unanimously quantify without direct measurement, which was not possible here. The

N2O5 purity is assumed to be high as various impurity mitigation strategies such as purging the system before synthesis with O3, evacuating the system to ensure the minimal presence of H2O vapour and cycling the purification process. The associated error of the N2O5 synthesis is propagated to the ClNO2 error.

A secondary method to directly produce ClNO2 using chlorine atoms and NO2 was devised by cross calibration with a turbulent flow tube chemical ionisation mass spectrometer (TF-CIMS) (Leather et al. 2012). The TF-CIMS quantifies NO2 that is constantly flowed into a flow tube that precedes the CIMS instrument. Increasing concentrations of Cl2 are then passed through a microwave discharge to create Cl atoms and into the NO2 stream. These Cl atoms then react with the excess NO2 to form ClNO2.

The drop in NO2 measured on the TF-CIMS is the 1:1 ratio of the increase of ClNO2 signal observed on the ToF-CIMS which subsamples the flow tube. Whilst this technique is promising in its methodology, there remain uncertainties in the differences in sampling efficiencies of the two CIMS systems. The TF-CIMS method also assumes the Cl generated from the microwave discharge forms ClNO2 with a 1:1 conversion rate. It is possible that Cl is lost elsewhere and not by the reaction with NO2. This would lead to a reduced ClNO2 signal on the ToF-CIMS and thus a greater calibration factor. This

44 method gave a calibration factor 58% greater than the established N2O5 salt slurry method and so for it to be a reliable calibration methodology, requires further work.

However, this demonstrates that reaction with excess NO2, either to form detectable products or observe a detectable loss on the instrument to be calibrated, is a viable calibration method if the loss of NO2 can be directly quantified by another instrument. This could be a TF-CIMS as in this case, or a spectroscopic method such as cavity attenuated phase shift spectroscopy (Kebabian et al. 2008).

2. Aims and Objectives The Aerodyne ToF-CIMS demonstrates a new capability of the in situ detection of ambient trace gases at low concentrations. With this tool, the composition of the urban atmosphere may be probed in greater detail to understand atmospheric processes relevant to urban air quality such as primary emission and oxidation.

The density of ambient measurements made with the ToF-CIMS is relatively sparse as this instrument is relatively new to the atmospheric measurement community. It is not as widely employed as other established instruments such as the AMS. The scarcity of specific trace gas measurements made with the ToF-CIMS is magnified when considering the environments in which they have been deployed and that different ionisation schemes alter the selectivity of the instrument. European research with the ToF-CIMS focuses strongly on the quantification and mechanistic understanding of SOA precursor formation (e.g. Rissanen et al. 2014), mainly in a laboratory context (e.g. Mentel et al. 2015) or in rural locations where biogenic emissions are high (e.g. Bianchi et al. 2017). Globally there are examples of ToF-CIMS being used for measurements of urban air (Brophy & Farmer 2015), although again this reduces when considering iodide ionisation.

A good case study for the ambient detection of many different compounds is Guy Fawkes Night (Bonfire Night, paper 1) where open fires are lit en masse representing a short lived, novel and large emission source of many different biomass burning products to the atmosphere. Trace gas mixing ratio enhancements are reported for newly identified compounds and contextualised with traditional air pollutant monitoring.

Oxidation in urban environments is perturbed relative to non-urban environments due to the presence of additional oxidant sources. Chlorinated species are of interest as they are potentially a source of the highly reactive oxidant the chlorine radical. The ability of the iodide ionisation scheme at detecting chlorinated compounds is well documented (Bannan et al. 2015; Le Breton et al. 2018). The ToF-CIMS provides the opportunity to expand the known number of detectable chlorinated compounds, including organics, due to the increased mass range and resolution afforded by the ToF (paper 2). The variability of these compounds can be explored and a comparison of chlorine radical yield from these sources is calculated.

The ToF-CIMS has been used extensively for the identification of organic compounds formed by the oxidation of a VOC precursor with the intention of understanding the formation of HOM and ultimately SOA. As BVOCs represent the majority of global VOC emission, BVOC e.g. isoprene and α-pinene oxidation are often the focus of study. In

46 urban environments, the emission of anthropogenic VOCs such as aromatics and alkanes are typically more dominant than BVOC emission, especially in deforested areas and at higher latitudes (Simpson et al. 1999). More recently, HOM formation from anthropogenic VOC precursors has been investigated (Wang et al. 2017), although to a much smaller degree. The mass range of the ToF-CIMS and rapid response of detection allow for real time sampling of high mass organic species during these experiments.

Here the presence of NOx, a near ubiquitous urban trace gas, on the product distribution of benzene oxidation products in a chamber experiment is explored (paper 3). A comparison between iodide and nitrate ionisation schemes further adds to the body of literature describing the oxidation spaces that the ionisation schemes are sensitive towards and the space where they overlap.

Blank page

3. Paper 1. Observations of isocyanate, amide, nitrate and nitro compounds from an anthropogenic biomass burning event using a ToF‐CIMS

Michael Priestley, Michael Le Breton, Thomas J. Bannan, Kimberly E. Leather, Asan Bacak, Ernesto Reyes‐Villegas, Frank De Vocht, Beth M. A. Shallcross, Toby Brazier, M. Anwar Khan, James Allan, Dudley E. Shallcross, Hugh Coe, Carl J. Percival

Published JGR: 27 February 2018 doi:10.1002/2017JD027316

Research Highlights:

Mixing ratio enhancements of 2-13 times ambient are recorded for various N containing compounds during bonfire night using a ToF-CIMS. Low wind speeds and poor mixing favour pollutant accumulation with HNCO concentrations reaching potentially harmful levels.

Whilst this event is highly polluting NO2 concentrations at this site are higher at other times most likely due to emissions from traffic.

Author Contributions:

Data collection was performed by Michael Priestley, Michael le Breton, Thomas J. Bannan and Asan Bacak. Laboratory work was performed by Michael Priestley, Kimberly E. Leather and Asan Bacak. AMS data products and interpretation were provided by Ernesto Reyes‐Villegas and James Allan. Health impact calculations were performed by Frank De Vocht. Data interpretation was aided by Beth M. A. Shallcross, Toby Brazier, M. Anwar Khan and Dudley E. Shallcross. The article was written by Michael Priestley under the supervision of Hugh Coe and Carl J. Percival.

Blank page

PUBLICATIONS

Journal of Geophysical Research: Atmospheres

RESEARCH ARTICLE Observations of Isocyanate, Amide, Nitrate, and Nitro 10.1002/2017JD027316 Compounds From an Anthropogenic Biomass Key Points: Burning Event Using a ToF-CIMS • ToF-CIMS identifies isocyanates, amides, nitro-organics, and nitrates Michael Priestley1 , Michael Le Breton1,2 , Thomas J. Bannan1 , Kimberly E. Leather1, with mixing ratio enhancements 1 1 3 4 5 between 2 and 13 times during Asan Bacak , Ernesto Reyes-Villegas , Frank De Vocht , Beth M. A. Shallcross , Toby Brazier , 5 1,6 5 1 1,7 bonfire night M. Anwar Khan , James Allan , Dudley E. Shallcross , Hugh Coe , and Carl J. Percival • Low wind speeds and poor mixing favor pollutant accumulation with 1Centre for Atmospheric Science, School of Earth and Environmental Sciences, University of Manchester, Manchester, UK, HNCO concentrations reaching 2Now at Department of Chemistry and Molecular Biology, University of Gothenburg, Göteborg, Sweden, 3School of Social potentially harmful levels and Community Medicine, University of Bristol, Bristol, UK, 4Division of Pharmacy and Optometry, University of Manchester, • While this event is highly polluting, Manchester, UK, 5School of Chemistry, University of Bristol, Bristol, UK, 6National Centre for Atmospheric Science, University NO2 concentrations at this site are 7 higher at other times most likely due of Manchester, Manchester, UK, Now at Jet Propulsion Laboratory, Pasadena, CA, USA to emissions from traffic

Abstract Anthropogenic biomass burning is poorly represented in models due to a lack of observational Supporting Information: data but represents a significant source of short-lived toxic gases. Guy Fawkes Night (bonfire night) is a • Supporting Information S1 regular UK-wide event where open fires are lit and fireworks are set off on 5 November. Previous gas Correspondence to: phase studies of bonfire night focus on persistent organic pollutants primarily using off-line techniques. C. J. Percival, Here the first simultaneous online gas phase measurements of several classes of compounds including [email protected] isocyanates, amides, nitrates, and nitro-organics are made during bonfire night (2014) in Manchester, UK, using a time-of-flight chemical ionization mass spectrometer (ToF-CIMS) using iodide reagent ions. A Citation: shallow boundary layer and low wind speeds favor pollutant buildup with typical HCN, HNCO, and Priestley, M., Le Breton, M., Bannan, T. J., CH NCO concentrations of tens of parts per thousand increasing by a factor of 13 to potentially harmful Leather, K. E., Bacak, A., Reyes-Villegas, E., 3 et al. (2018). Observations of isocyanate, levels >1 ppb. Normalized excess mixing ratios relative to CO for a range of isocyanates and amides are amide, nitrate, and nitro compounds reported for the first time. Using a HNCO:CO ratio of 0.1%, we distinguish emissions from flaming and from an anthropogenic biomass burning smoldering combustion and report more accurate normalized excess mixing ratios for the distinct burning event using a ToF-CIMS. Journal of fi Geophysical Research: Atmospheres, 123. phases. While bon re night is a highly polluting event, NO2 concentrations measured at this location https://doi.org/10.1002/2017JD027316 are higher at other times, highlighting the importance of traffic as an NO2 emission source at this location. A risk communication methodology is used to equate enhancements in hourly averaged black carbon and Received 30 JUN 2017 NO concentrations caused by bonfire night as an equivalent of 26.1 passively smoked cigarettes. Accepted 14 FEB 2018 2 Accepted article online 27 FEB 2018 1. Introduction Biomass burning (BB) is a major source of gas and carbonaceous aerosol emission to the atmosphere (Andreae & Merlet, 2001), both of which act to reduce air quality worldwide (Molina et al., 2007). Solid biofuel burning makes up part of the anthropogenic contribution to BB, being the primary source of heating and cooking for 3 billion people worldwide (World Health Organization, 2015). Emissions from BB affect large population centers across the globe. For example, in New Delhi, India, a mega city with a population of

26,454,000 (United Nations, 2016), approximately 99% of inhabitants are exposed to more PM2.5 than the 3 10 μgmÀ WHO air quality guideline (World Health Organization, 2006). An estimated 20% of the PM2.5, which contributes to this exposure originates from open BB (Amann et al., 2017). While this is not necessarily as important in the UK, solid biofuel burning is becoming more popular for financial, esthetic, and environmental reasons (Caird et al., 2008) and is significant in its contribution to reducing air quality (Fuller et al., 2014). One example of a regular, nationwide BB event in the UK is Guy Fawkes Night, or bonfire night, which is celebrated annually on and around 5 November by lighting open fires and fireworks as part of community events and at individual households. These bonfires are lit at roughly the same time during the evening and are designed to have a strong flaming phase that lasts for 1–2 h. After flaming, the fires are not refueled and ©2018. The Authors. fi This is an open access article under the so there is an extended period of smoldering as the res are left to die away. The UK Environment Agency terms of the Creative Commons permits the open burning of untreated wood and garden waste (UK Government 2015) and states that trea- Attribution License, which permits use, ted materials and household waste (solvents, plastics, etc.) should not be burnt, although it is likely that these distribution and reproduction in any fi fi medium, provided the original work is types of materials do contribute to the composition of bon re night open res. This mixed fuel source is dif- properly cited. ficult to categorize and most likely represents a mixture of residential biofuel combustion and garbage

PRIESTLEY ET AL. 1 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

burning, the latter of which is poorly characterized across the globe and known to emit many toxic compounds (Akagi et al., 2011, and references therein). It is known that bonfire night is among the most polluted days in terms of air quality in the UK (Dyke et al., 1997; Mari et al., 2010; Pongpiachan et al., 2015). Bonfire night exhibits significantly elevated particle levels compared with the year average at all urban sites in the UK (Harrison & Shallcross, 2011), and inspection of long-term measurement sites such as the Marylebone Road site in London shows that the night associated with bonfire night (it may be the Friday, Saturday, or Sunday closest to 5 November) would appear to be the highest in terms of pollution with associated implications for human health. The effects of open fire and firework events on enhancing metalliferous particles (Moreno et al., 2007), aerosol (Vassura et al., 2014), and trace gas concentrations (Drewnick et al., 2006) including volatile organic compounds (VOCs) are well documented. Much of the data on gaseous pollutants collected during previous bonfire nights in the UK have focused on persistent organic pollutants (Farrar et al., 2004; Harrad & Laurie, 2005), which have been linked to climate change (Nadal et al., 2015) and are linked to adverse human health effects including cancer (Mouly & Toms, 2016) and reproductive diseases (Bonde et al., 2016). Sampling of specific pollutants has mainly been off-line, using whole air sampling or filter collection, reducing the temporal resolution of the data sets. Advances in measurement techniques such as the development of the time-of-flight chemical ionization mass spectrometer (ToF-CIMS) (Bertram et al., 2009) with its high selectivity, sensitivity, resolution, and data acquisition rate permit enhanced detectability of short-lived, toxic, nonpersistent organic pollutants trace gases in real time. Hydrogen cyanide (HCN) is one such highly toxic gas and known BB tracer (e.g., Yokelson et al., 2007) with a typical lifetime of 2–4 months (Li et al., 2009) that has previously been measured by iodide CIMS (e.g., Le Breton et al., 2013). Globally, the greatest sources of HCN to the atmosphere are biogenic, via cyanogen- esis, and BB (Li et al., 2003; Shim et al., 2007), which is known to be highly variable and strongly dependent on fuel type (Akagi et al., 2011; Coggon et al., 2016). In urban locations, a significant contribution to ambient HCN originates from vehicles (Moussa et al., 2016). Isocyanic acid (HNCO) is another highly toxic, long-lived gas (lifetime of days to decades; Borduas et al., 2016) emitted from BB with similar anthropogenic and biogenic sources as HCN. Urban sources of HNCO are attributed to primary activity such as automotive emission (Jathar et al., 2017), residential heating (BB) (Woodward-Massey et al., 2014), and industrial processes, for example, from brick kiln emissions (Sarkar et al., 2016). A secondary source of HNCO is amide oxidation (e.g., Borduas et al., 2015), which has been observed at a suburban site in Mohali, India (Chandra & Sinha, 2016), and in an urban environment in Pasadena, California (Roberts et al., 2014). Mean urban concentrations of HNCO are variable having previously been measured to be on the order of 10–100 ppt by acetate CIMS in Pasadena (Roberts et al., 2014) and ~1 ppb using a proton transfer reaction-mass spectrometer (MS) in Mohali (Chandra & Sinha, 2016). This higher level is attributed to a strong regional background potentially caused by the oxidation of another set of precursors, alkyl amines, originating from agricultural BB (Sarkar et al., 2016, 2017). In addition to its toxicological importance, HNCO measurements are useful as the criteria of [HNCO]/[CO] has been used to define the separation of flaming from smoldering combustion phases (Roberts et al., 2011). Methyl isocyanate (MIC) is a homologue of HNCO and a known secondary pollutant with precursors originating from both biogenic and anthropogenic sources (Lu et al., 2014; Woodrow et al., 2014) MIC has been measured as a direct emission from industrial processes and the burning of common building materials (Blomqvist et al., 2003; Henriks-Eckerman et al., 2002). Amides are known products of BB (e.g., Stockwell et al., 2015) and from the burning of household materials (Kim et al., 2015). These compounds are known toxins (Gescher, 1990), which, as previously mentioned, are also precursors to isocyanic compounds (Borduas et al., 2015). In Shanghai, Yao et al. (2016) found that amides with greater than three carbon atoms were prevalent over lower mass amides, with concentrations on the order of parts per billion (ppbs) suggesting both secondary and industrial primary sources. Organic nitrates are another class of toxic compounds with wide implications for atmospheric chemistry.

They are a reservoir of NO2, thus enabling the transport of NOx and subsequent ozone formation at remote locations, and are typically semivolatile, contributing to secondary organic aerosol formation and thus particulate matter (PM) with further associated air quality issues. Other nitrates such as peroxynitric acid have previously been detected by iodide CIMS (Veres et al., 2015) and peroxyacetyl nitrate (PAN), which has previously

PRIESTLEY ET AL. 2 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

been detected by PAN iodide CIMS (Veres & Roberts, 2015). Nitro-organic compounds such as nitrophenol and its degradation products, which have been shown to be genotoxic (Sekler et al., 2004), have many sources including vehicular emission, degradation of pesticides (Lüttke et al., 1997, and references therein), secondary gas phase reactions (Berndt & Böge, 2006), and aqueous aerosol phase reactions (Yuan et al., 2015) and have vapor pressures low enough to readily partition into the condensed phase (Bannan et al., 2017). Nitrophenols, methyl nitrophenols, and nitrocatechols have been detected in laboratory BB smoke using ToF-MS (Iinuma et al., 2010) and also in urban plumes (Mohr et al., 2013). Measurements were carried out from 29 October 2014 to 11 November 2014 at the University of Manchester to assess the impact of bonﬁre night activity on local air quality and to probe the complex composition of an urban BB plume using novel identiﬁcation methods to identify toxic species that currently do not feature in air quality health assessments.

2. Methodology 2.1. Site Description Manchester is located in the center of Greater Manchester Metropolitan County, an administrative area encompassing 10 different metropolitan boroughs of mostly urban districts, with a collective area of 1,276 km3 and a population of 2.6 million inhabitants. The Whitworth Observatory is an urban rooftop measurement site approximately 15 m above street level and 100 m from the nearest road, located in the Simon Building at the University of Manchester’s south campus, approximately 1.5 km south of Manchester City Centre (53.467°N, 2.232°W). For a map of the measurement location and sites of large-scale public bonfires, see Reyes-Villegas et al. (2017). 2.2. Instrument Description 2.2.1. Time-of-Flight Chemical Ionization Mass Spectrometer A high-resolution ToF-CIMS as described by Lee et al. (2014) was deployed to measure a vast suite of atmospheric species using the iodide reagent ion. It consists of a reduced pressure ion molecule reaction (IMR) region coupled to a Tofwerk atmospheric pressure interface high-resolution time-of-flight mass spectrometer (Junninen et al., 2010). The instrument has been described in detail elsewhere (Lee et al., 2014); here we describe the specific details related to its setup for this campaign. The IMR was held at a constant pressure of 100 mbar by a scroll pump (Agilent SH-112) controlled with a servo control valve placed between the scroll pump and IMR. Ambient air was drawn into the IMR via a critical orifice at 2.2 standard liters per minute (slm). The reagent IÀ ions were created by flowing 10 standard cubic

centimeters per minute of methyl iodide (CH3I) emitted from a permeation tube held at 40°C through 1/8″ perfluoro alkoxy (PFA), which is carried by a 2 slm nitrogen (N2) flow through a Po-210, 10 mCi, alpha emitting reactive ion source (NRD Inc Static Solutions Limited) orthogonal to the ambient flow into the IMR. Once in the IMR, the IÀ ions then ionize species of interest. The ionizer and ambient air flows mixed for approximately 30 ms until a fraction of the flow was sampled through an orifice into the first of four differentially pumped chambers in the ToF-CIMS. The first chamber was held at 150 mbar by a scroll pump (Triscroll 600), and the second stage was pumped by a split flow turbo molecular drag pump and held at 1.50 mbar. Quadrupole ion guides transmit the ions through these stages while simultaneously providing extracollisional cooling and energetic homogenization of the ions as they enter the extractor region. The electric field strengths in the axial direction (<2 V/cm) were set to optimize the total ion signal and transmission of the iodized species

of interest (E/N ratio of 65 Td). Optimization of the I·H2OÀ cluster signal is considered essential to optimize system sensitivity since the detection of particular compounds, for example, formic acid, is dependent on

the number of I·H2OÀ adducts. The ions were subsequently pulsed into the drift region of the ToF-CIMS at 22.22 kHz where the arrival time is detected with a pair of microchannel plate detectors. The mass calibration

was performed for seven known masses: NO3À,IÀ,IÀ·H2O, IÀ·HCOOH, IÀ·HNO3,I2À, and I3À, which covers a range of 62 to 381 m/z. The mass calibration was ﬁtted to a third-order polynomial and was accurate to within 2 ppm; ensuring that peak identiﬁcation was accurate below 20 ppm. The resolution was 3,560 at 127 m/z and 3,845 at 381 m/z. All species reported are observed as adducts with IÀ. 2.2.1.1. Sampling, Calibration, and Backgrounds To minimize sampling losses through the inlet tubing, a fast inlet pump is implemented to sample at 15 slm through 1 m long ¾″ PFA tubing. This translates to an inlet residence time of 1.2 s. Approximately 2.2 slm of

PRIESTLEY ET AL. 3 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

the centerline flow is subsampled into the IMR of the ToF-CIMS through a conical shaped critical orifice, while the remaining flow is exhausted by the scroll pump. Backgrounds were taken every 6 h for 20 min by over-

flowing the inlet to the IMR with dry N2. As the ToF-CIMS is able to detect 102–103 species, calibration becomes difficult. One methodology to simplify calibration is to apply a uniform calibration factor using a maximum sensitivity based on adduct formation at the collision limit and assuming maximum transmission efficiency (Ehn et al., 2014; Lopez-Hilfiker et al., 2015). Another methodology makes use of an average organic acid calibration factor (Chhabra et al., 2015), while others aim to derive sensitivities from bind- Figure 1. Time-of-flight chemical ionization mass spectrometer C1–C5 organic acid calibration factors. (inset) C2–C5 calibration factors show ing enthalpies of the reagent ion (Iyer et al., 2016). Here we apply the similar sensitivities. Error bars represent ±2σ. The gray area represents upper laboratory-determined calibration factors for a series of C1 to C5 alipha- and lower errors of C2–C5 calibration factors. tic organic acids to species with an equivalent number of carbon atoms, assuming that the carbon number dictates the collision limit of adduct formation. Above a carbon number of 2 within this particular series, the calibration factor is independent of m/z 1 (Figure 1). We use this assumption to apply a uniform calibration factor of 6.20 Hz pptÀ , derived from for- 1 mic acid, for C1 compounds and a calibration factor of 1.28 Hz pptÀ , derived from acetic acid, for Cn compounds (n > 1).

Formic acid was both measured and calibrated for throughout the campaign by adding known concentrations of HCOOH as previously described in Le Breton et al. (2012). The ToF-CIMS HCN signal was cross calibrated with the calibrated quad-CIMS. Known concentrations of HCN were produced by ﬂowing from a HCN calibration cylinder (BW Technologies) that was diluted from 10 ppm mix with an accuracy of ±10% as described by Le Breton et al. (2013) (supporting information S1). The sensitivity of the quad- CIMS to HNCO is assumed to be the same as HCN. The sensitivity of the quad-CIMS and ToF-CIMS is shown in Table 1.

Sensitivity changes due to reagent ions are minimal. Mean average IÀ + I·H2OÀ counts were high (3.52 ×106 ± 5.2 ×105 (1σ)) and well above the threshold where we observe a much reduced dependency

of sensitivity on reagent ion count (Le Breton et al., 2014). We normalize to IÀ, I·H2OÀ, or the sum of both depending on which has the best correlation with the signal of the species of interest. If there is no discernable correlation, no normalization takes place. None of the signals for the species reported in this manuscript

were normalized to I·H2OÀ or the sum of IÀ + I·H2O.À The calibration standard stock solutions for each acid were made with 95–97% reagent grade organic acids (Sigma-Aldrich) to produce 1% volume per volume solutions in water. These stock solutions were subsampled to make calibration standards ranging from 0.1 to 1.1 ppm. About 100 μL of calibration standard was injected into an evacuated Pyrex impinger connected to the evacuated 118 L Extreme Range Reaction Chamber (Leatheret al., 2010). The calibration standard volatilizes when exposed to the evacuated system, and partitioning to the gas phase is Table 1 aided by passing N2 carrier gas through the impinger until the chamber Calibration Factors for the ToF-CIMS Used in This Study is ﬁlled to ~760 torr (1.74% error). Calibration mixes of 500, 1,000, and 1 2,000 ppt are subsampled by the ToF-CIMS through 70 cm ¼“ PFA tub- Calibration factor (Hz pptÀ ) ing. Blank water backgrounds were performed before every calibration Compound Quad-CIMS ToF-CIMS ToF-CIMS 2σ error by injecting 100 μL of water and the lines cleaned with N2 between sam- HCN 4 1.93 - pling. These calibrations were performed postcampaign and compared HNCO 4 2.65 - with in-house made gas mixtures as described in Bannan et al. (2015) Formic acid (C1) - 6.20 2.48 showing the same results to within 5% error. Acetic acid (>C1) - 1.28 0.60 Propanoic acid - 0.61 0.08 A literature survey (Brophy & Farmer, 2015; Iyer et al., 2016; Lee et al., Butanoic acid - 1.61 0.33 2014; Veres & Roberts, 2015) of formic acid calibration factors (normal- Pentanoic acid - 1.21 0.33 ized to 1 ×106 Hz reagent ion) provides a range between 2.9 and ﬂ 1 1 Note. ToF-CIMS = time-of- ight chemical ionization mass spectrometer. 13 Hz pptÀ (with reported errors of 0.6–5.0 Hz pptÀ ) highlighting

PRIESTLEY ET AL. 4 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Table 2 Measurements, Resolutions, and Accuracies of Meteorological Instruments at the Whitworth Observatory Parameter Instrument Resolution Accuracy 1 Wind speed Gill Windmaster Pro Sonic Anemometer 0.01 msÀ 1.50% Wind direction Gill Windmaster Pro Sonic Anemometer 0.10° 2.00% Temperature Rotronics MP100-H mounted in Rotronics Aspirated Radiation Shield (RS12T) 0.10°C 0.30°C Relative humidity Rotronics Hygroclip 0.10% 1.50% Barometric pressure Vaisala PTB10 Digital Barometer with Vaisala SPH10 Static Pressure Head 0.01 hPa 0.30 hPa Direct solar radiation Kipp and Zonen CMP-11 Pyranometer — 2.00% NOx (NO + NO2) Thermo Scientific 42i 0.40 ppb 0.40 ppb O3 Casella ML2010 10.00 ppb 10.00 ppb CO Thermo Scientific 48i 0.10 ppm 0.04 ppm SO2 Thermo Scientific 43i-TLE 0.20 ppb 0.05 ppb

the variation in sensitivity that is associated with the instruments’ condition and circumstance for this 1 1 particular compound. Our formic acid calibration factor of 6.20 ± 1.28 Hz pptÀ (4.13 ± 0.85 Hz pptÀ normalized to 1 ×106 Hz reagent ion) is well within this range. Of 61 compounds surveyed in the 5 1 literature, a range of calibration factors from 7.6 ×10À to 22 Hz pptÀ with a mean sensitivity and error 1 1 of 5.20 ± 0.97 Hz pptÀ and 1σ = 6.61 Hz pptÀ further demonstrate the variability of the detection capabilities of the ToF-CIMS. 2.2.2. Meteorology and Air Quality Meteorological measurements (pressure, temperature, relative humidity, wind speed and direction, precipi- tation, visibility, and actinic ﬂux) are made approximately 100 m away on the nearby George Kenyon Building approximately 40 m above ground level in the University’s south campus (approximately 53.466°N, 2.232°W). For a map of the measurement site and large-scale public bonﬁres, see Reyes-Villegas et al. (2017). Temperature, humidity, pressure, and wind speed and direction instruments are located approximately 3–5 m above the surface of the station to reduce the impact of the building below on the measurements being made. A summary of meteorological measurements, instruments, ranges, resolutions, accuracies, and data collection frequencies is summarized in Table 2. Automated high-frequency (0.1 Hz)

long-term trace gas measurements of NOx,O3, CO, and SO2 are made at the same location as the ToF- CIMS measurements. It is noted that the NOx measurement technique uses a molybdenum catalyst that is known to cause an overestimation of NO2 concentrations by the nonselective conversion of NOy species as well as NO2 to NO, which is actively detected by chemiluminescence. However, as the UK automatic urban rural network also uses this technique to measure NOx, the results presented here should be comparable. A lack of CO2 measurements prevents the calculation of modified combustion efficiencies (MCEs) (Yokelson, Griffith, & Ward, 1996).

2.3. Kendrick Mass Defect Kendrick mass defect (KMD) analysis is a mass spectrometric data analysis technique used to identify individual species in complex mixtures. While KMD analysis has been used to study the chemical composition of secondary organic aerosol (SOA) (Walser et al., 2008), it relies on a high-resolution (m/Δm > 50,000) typical of magnetic sector instruments with high accuracies, for example, <1 ppm (Bristow & Webb, 2003). The average resolution of the ToF-CIMS was 3,800, and accuracy was <20 ppm. While these metrics indicate that the instrument is not able to unequivocally distinguish individual species, it is possible to use KMD analysis to suggest potential chemical formulae for unknown species if we account for the error in exact mass peak assignment and the selectivity of the IÀ reagent ion. From the ToF-CIMS high-resolution mass spectra recorded over the campaign, a total of 652 peaks was manually identified; 75 peaks were first identified by a priori knowledge and by using peak fitting software, leaving 577 unknown assignments. The peak list, containing the 75 known assignments, was then run through a program that matches known and unknown species that share the same KMD (within error) when normalized to different moieties (e.g., 12 CH2, O, and CHO) as well as C. The procedure is as follows: Where the m/z of the unknown assignment is

PRIESTLEY ET AL. 5 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

an integer n times the m/z of the known assignment, it is interpreted that the assignments are related in KMD space and the formula of the unknown assignment can be determined from the known assignment (±n moieties). These potential assignments are assessed for their validity in terms of realistic structures and detectability by the ionization method. Figure 2 shows the total number of peak assignments as well as the known assignment of formic acid and the ability of the program to identify four “unknowns”in the same series, that is, the

C2-C5 organic acids. As this technique relies heavily on high mass accuracy, accurate mass calibration and peak shape are vitally important. Peak assignments are treated as indicative and cannot account for isomeric or isobaric compounds. A sample of the chemical formula and suggested chemical species identified by this technique in conjunction with time series correlation analysis is discussed below, and isocyanate, nitrate, nitro, and Figure 2. Kendrick mass defect plot of the identified peaks in this data set amide compounds are tentatively assigned. (gray). (inset) Exact masses (red) and measured masses (blue) of the C1–C5 organic acids in CH2 Kendrick mass defect space. Error bars show propagated 20 ppm error. All measured peaks are well within error. The exact masses 3. Results and Discussion (red) show where the measured masses should be found on this plot, 3.1. Pollution Events connected by a straight line. Where the assignment is measured (picked, blue), the error in the exact mass is visible as a deviation from the straight Diurnal profiles of CO and NOx concentrations show a prominent daily line. The transition from formic to acetic acid shows a deviation in the bimodal cycle associated with traffic rush hours (07:00–10:00 UTC and straight line, which is a manifestation of the systematic error in the mass axis. 16:30–18:00 UTC). At the peak of the morning rush hour (approximately All subsequent measured peaks sit on a straight line, indicating that there is no more deviation, and so the error in the mass axis is propagated but 08:30) average CO and NOx concentrations peak at 225 ppb and 31 ppb not compounded. and again at 250 ppb and 30 ppb during the evening rush hour at 16:45. Minima for both species are observed during the early morning (04:35), with average values of 146 ppb and 5 ppb. Periodic stagnation events 1 occur when wind speeds are low (<2.0 msÀ ) with the highest concentrations of NOx and CO measured when these events coincide with a rush hour period. At these times, CO and NOx concentrations reach, at their maximum, 400 ppb and 508 ppb, respectively. Figure 3 summarizes the measurement period. All measurements are reported as 1 min averages.

3.2. Bonfire Night Bonfire night (5 November) is defined as the period between 16:30 on 5 November and 07:30 on 6 November. During this period, the mean pressure was 998 hPa, mean temperature was 5.4°C (lowest during the 1 campaign), and mean wind speed was 1.89 m sÀ . This combination of high pressure, low temperature, and low wind speed is indicative of a stable, shallow boundary layer and poor dispersion, increasing the buildup of pollutants. No one bonfire was directly sampled; instead, a mixture from multiple sources (both private and public) was accumulated and mixed together forming a homogenous air mass dominating the

conventional background. NOx measurements from the UK automatic urban rural network monitoring network (Manchester Piccadilly, Manchester South, Salford Eccles, Glazebury) recorded maxima with varying wind directions and show little correlation with nearby large-scale events further suggesting that over a wider geographical area, the air mass composition cannot be attributed to any one single event (S2). 3.2.1. Identifying the Bonfire Burn Period Using BB Markers CIMS has previously been used to identify BB plumes using a statistical approach (Le Breton et al., 2013) where the concentration of a BB marker (e.g., HCN) above 6 times the standard deviation of its median background concentration is used to define a BB plume. As previously stated, no one plume during bonfire night could be identified individually. However, as bonfire night burning broadly follows a regimented schedule of prescribed activity (with flaming phase combustion occurring for 1–2 h during the early evening and smoldering thereafter lasting well into the next morning), this approach to defining in-plume sampling is applic- able in identifying the temporal bounds of bonfire night burning activity, from here termed the bonfire burn period (BBP). HCN and isocyanic acid (HNCO) were used as BB markers to which the 6 sigma methodology was applied. The BBP started at approximately 16:00 and ended between 04:00 and 05:00. As HCN is more

PRIESTLEY ET AL. 6 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Figure 3. Time series of key measurements during the campaign (30 min averaged). (a) HCN and HNCO; (b) NO, NO2, and NOx; (c) CO and SO2; (d) temperature and relative humidity; and (e) wind speed. Nongray background indicates daylight (approximately 07:30–16:30).

Table 3 Top 20 Correlations of HCN and HNCO With Other Measured Species During the BBP HNCO HCN

Formula Identiﬁed as R2 Formula Identiﬁed as R2

1C4H6O6 Tartaric acid 0.95 HONO Nitrous acid 0.97 2C3H4O2 Acrylic acid 0.95 NOx NOx 0.94 3C4H6O2 Butanoic acid 0.94 C2H5NO N-Methylformamide 0.94 4C3H3NO3 2,4-Oxazolidinedione 0.94 C2H7NO3 — 0.94 5 HCN Hydrogen cyanide 0.93 C4H6O2 Butanoic acid 0.93 6C2H3NO2 N-Formylformamide 0.93 C2H5N3O3 — 0.93 7 HONO Nitrous acid 0.92 HNCO Isocyanic acid 0.93 8 CH3NCO Methyl isocyanate 0.92 NO2 Nitrogen dioxide 0.93 9 CH4O2 Methanediol 0.91 NO Nitric oxide 0.93 10 CH2O2 Formic acid 0.91 C7H14O6 — 0.92 11 C7H14O6 — 0.90 C3H7NO N,N-Dimethylformamide 0.92 12 NO2 Nitrogen dioxide 0.90 C3H7NO Acrylamide 0.92 13 C5H6O2 — 0.89 C5H6O2 — 0.91 14 C3H7NO Acrylamide 0.89 C4H6O6 Tartaric acid 0.91 15 C2H5NO N-Methylformamide 0.88 C6H6O Phenol 0.91 16 C6H6O Phenol 0.88 C2H3NO2 N-Formylformamide 0.90 17 NOx NOx 0.88 C3H4O2 Acrylic acid 0.90 18 C2H7NO3 — 0.87 C7H6O2 Benzoic acid 0.90 19 C5H4O4 — 0.86 C5H4O4 — 0.90 20 C3H6O2 Propanoic acid 0.86 C5H8O4 Glutaric acid 0.87 Note. BBP = bonﬁre burn period.

PRIESTLEY ET AL. 7 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Table 4 Mean and Maximum Concentrations of Selected Species With Bonfire Plume Removed, Maximum Concentration Measured During the BBP Only, and Limit of Detections BBP removed BBP only fi Formula Mass (Da) Identi ed as Compound Cmean (ppt) Cmax (ppt) Cmax (ppt) LOD (ppt) HCN 154 Hydrogen cyanide 35 132 1,235 2.59 HNCO 170 Isocyanic acid Isocyanate 12 144 1,639 2.70 CH3NCO 184 Methyl isocyanates Isocyanate 61 327 4,299 14.00 C3H3NO2 212 Cyano acetic acid Isocyanate 10 65 84 1.41 CH3NO4 220 Nitroperoxy methane Nitrate 8 20 31 2.86 C2H3NO5 248 PAN Nitrate 54 221 15 4.90 C6H5NO3 266 Nitrophenol Nitro 131 530 630 15.62 C7H7NO3 280 Methyl nitrophenol Nitro 96 299 550 7.28 C6H9N3O6 346 Trinitrocyclohexane Nitro 14 133 73 2.82 CH3NO 170 Formamide Amide 10 25 189 2.05 C2H5NO 186 N-Methylformamide Amide 9 23 275 3.75 C3H5NO 198 Acrylamide Amide 15 52 148 2.43 C2H3NO2 200 N-Formylformamide Amide 26 95 100 3.30 C3H7NO 200 N,N-Dimethylformamide Amide 8 16 102 3.72 Note. BBP = bonfire burn period; LOD = limit of detections.

routinely measured, with elevated concentrations suggested unique to BB, it is chosen preferentially to define the BBP. Within the BBP, HCN and HNCO correlate with 184 and 140 species where 82 common species have R2 > 0.75. The correlated species are detected at elevated concentrations at an early stage during the evolution of the BBP. A summary of the 20 identified species with the strongest correlations with HCN and HNCO is shown in Table 3. Table 4 summarizes mean and maximum concentrations of nitrogen-containing compounds during the measurement period and the maximum concentration during the BBP. The limit of detection (LOD) reported is 3 times the standard deviation of the background measured at 1 min time resolution. Generally, concentrations of all species are highest during the BBP than at any other point. Two exceptions are PAN and trinitrocylco- hexane. LODs range from one to tens of ppt. 3 Not including the BBP, 19 ± 13% (1σ) of the 0.32 μgmÀ mean total detected mass (including all unknown peaks) can be attributed to the 75 identified peaks. This percentage is conserved at the highest mass loading 3 (1.00 μgmÀ ) recorded for this period. However, during the BBP the fraction of identified material increases 3 to 24 ± 21%, and the mean total detected mass increases to 0.70 μgmÀ . At the highest mass loading of 3 2.0 μgmÀ during the BBP, the fraction of identified material further increases to 41%. The five largest

organic concentration enhancements between the BBP and non-BBP are identified as C6H6O2,C6H6O3, CH2O2,C3H6O3, and C7H8O2. 3.2.2. Identification and Behavior of Key Species Excluding the BBP, the mean ambient HCN concentration of 35 ppt and maximum concentration of 132 ppt is comparable with other ground-based HCN measurements, both rural and urban (Ambrose et al., 2012; Knighton et al., 2009). The species with the highest correlations with HCN with the BBP removed are propanoic acid (R2 = 0.79), HONO (R2 = 0.76), MIC (R2 = 0.75), formamide (R2 = 0.71), and N-methylformamide (R2 = 0.68). Figure 4 summarizes diurnal profiles of selected species with the bonfire night period overlaid. Low HCN concentrations <10 ppt were more commonly measured during weekends and during the nighttime when anthropogenic activity is at a minimum. Low concentrations were also measured when wind speeds were high and the wind direction was southwesterly, associated with cleaner, inflowing air masses. Typical HCN concentrations of 30–50 ppt were most commonly observed during weekdays, when wind 1 speeds were greater than 3.0 msÀ (with no dependence on wind direction). Combined with a diurnal

PRIESTLEY ET AL. 8 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Figure 4. Five minute averaged ambient diurnal profiles of identified species throughout the campaign with the bonfire burn period removed (colored, left axis) and the shaded area representing 95% variation. Right axis and black trace show concentration and profile of the bonfire burn period starting at 16:30 on 5 November (nightfall) and extending to 16:30 on 6 November demonstrating the elevation and nontypical behavior of these compounds.

profile that does not exhibit a local emission signal (Figure 4), this range of concentrations is suggestive of a regional rather than local signal. For the weekdays of 30 October and 4, 5, and 7 November between the hours of 07:00 and 09:00, HCN concentrations >50 ppt and a positive linear relationship with NO (R2 = 0.67, 0.94, 0.73, and 0.59, respectively) demonstrates that there is a local vehicular component to the HCN measurement. Maximum HCN concentrations of 82 ppt on 4 November are higher than those associated with the regional signal (30–50 ppt). As the lifetime of NO is much less than HCN, the HCN/NO relationship is most obvious when wind speeds are low 1 (typically <3.0 msÀ ). As the wind speeds are higher on other weekdays, this is the most likely reason for poor correlations at those times. The example from 4 November indicates that a local enhancement of ~20 ppt is possible above the regional background. The maximum concentration associated with the BBP is 1.24 ppb, a factor of 10 higher than ambient. For HNCO, the nonbonfire mean concentration is 12 ppt with a maximum of 144 ppt. Similarly to HCN, the strongest relationship between HNCO and NO is during rush hour on the low wind speed days of 4 and 5 November (R2 = 0.71 and 0.72) but, unlike HCN, is poor on the other low wind speed days of 30 October and 7 November. A secondary source component to HNCO is evident as the HNCO/HCN ratio is higher in the presence of the photochemical marker ozone (Figure 5). One such formation pathway may be the oxidation of formamide (Borduas et al., 2015) with which HNCO has the highest correlation over the data set (R2 = 0.71) (not including BBP). Further species with high correlations with HNCO are phenol (R2 = 0.71), toluic 2 2 2 acid (R = 0.71), tartaric acid (R = 0.70), and C5H6O2 (R = 0.70). The correlation between HCN and HNCO is good R2 = 0.64. The maximum concentration measured during the bonfire plume of 1.64 ppb is comparable to concentrations recorded in agricultural burning plumes (Chandra & Sinha, 2016; Roberts et al., 2014) and is above the 1.0 ppb threshold, which is considered detrimental to human health (Roberts et al., 2011). Maximum concentrations of MIC during the BBP of 4.3 ppb and 327 ppt outside of the BBP were observed. An R2 of 0.75 with HCN and an of 0.53 with HNCO indicates that the behavior of MIC is more akin to HCN and is

PRIESTLEY ET AL. 9 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Figure 5. (a) HNCO versus HCN colored by levels of O3. The photochemical component of the HNCO/HCN relationship is highlighted at high levels of O3, which is indicative of photochemistry. (b) Positive relationship between formamide (HNCO precursor) and HNCO by levels of O3. All data from the bonﬁre burn period has been removed (30 min average).

therefore more likely to be primary in origin. The diurnal profile of MIC shows a minimum in the early morning fi and large variation at rush hours consistent with typical NOx diurnal pro les, indicating that MIC has a local, potentially traffic source. We detect a large number of nitrates and nitrated species, some of which have been previously detected with CIMS (peroxynitric acid and PAN) and some that have not: nitroformic acid, nitroperoxy methane, and trinitrocyclohexane. The nitrates and nitrated species exhibit maxima during the day, yet elevated concentrations measured during the BBP are evident. Sharp increases in concentrations of all nitrates and nitrated species, similar to those of the isocyanate compounds, are observed, although the increase in ambient levels is not as high (factor of 3 compared with a factor of 10 for isocyanate compounds and HCN). Concentrations then increase again, later in the early morning of 6 November. Maximum PAN and trinitrocyclohexane concentrations of 221 and 133 ppt are measured at 16:00 on 31 October 2014, while during the bonfire plume the maximum concentrations are depleted to 15 and 73 ppt, respectively. Conversely, while high nitrophenol and methyl nitrophenol concentrations of 530 and 229 ppt are also recorded at other points in the measurement period (afternoon of 31 October 2014), the highest concentrations of 630 and 550 ppt are observed during the BBP. This BBP enhancement is short lived, however, with concentrations returning to ~300 ppt at 22:00. Additional nitrogen-containing compounds were identified as amides: acrylamide, formamide, N- methylformamide, N,N-dimethylformamide, and N-formylformamide. Excluding the BBP, mean ambient concentrations are typically approximately tens of parts per thousand. Formamide, N-methylformamide, and N, N-dimethylformamide all exhibit maxima during the early afternoon (13:00–14:00) and minima during the morning rush hour (approximately 07:00–09:00). This is consistent with photochemical formation during

the day and loss when NOx concentrations are high. During the BBP, concentrations of all amide species increase to hundreds of parts per thousand very quickly. Unlike the nitrates, no secondary peaks in concentrations are detected later in the morning. Instead, concentrations return to ambient levels indicating that emission is short lived and loss processes are fast. These loss processes are unlikely to be homogeneous gas phase reactions as the lifetimes of these compounds, controlled by reaction with OH, are typically 1–2 days (Borduas et al., 2015; Bunkan et al., 2015).

3.3. Distinguishing Combustion Regimes and Emission Ratios. Laboratory and field studies indicate that the ratio of [HNCOCO] decreases from 0.6–0.1% during flaming combustion to values 5–10 times lower during smoldering combustion (Roberts et al., 2010, 2011). By applying a 0.1% threshold, the criteria for flaming combustion is met between 17:30 and 20:00 on bonfire night.

PRIESTLEY ET AL. 10 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Figure 6. Normalized excess mixing ratios of selected compounds to CO. Gray points are smoldering phase emission described by the blue line, nongray points are ﬂaming emission described by the red line, and black lines include all data. Slopes and R2 are reported in Table 5.

This is the approximate time that most bonfire flaming combustion occurs, as previously stated. These times are indicative rather than exact, as there will be some variation in flaming time between events and air mass mixing will affect the temporal coincidence at the measurement site.

ΔX X X NEMR plume À background (1) X ¼ ΔCO ¼ CO CO plume À background The normalized excess mixing ratios (NEMRs) for a range of identiﬁed species relative to CO are calculated during the BBP using equation (1), many of which to the authors’ knowledge have not previously been reported (Figure 6). The background concentrations of the species of interest and CO are taken as the mean value from the diurnal proﬁle of that species and CO at 5 min time steps from the entire measurement period with the BBP removed. “Plume”concentrations of the species of interest and CO are taken

Table 5 1 Emission Ratios Relative to CO (ppt ppbÀ ) BBP Flaming Smoldering 1 2 1 2 1 2 Compound NEMR (ppt ppbÀ ) R NEMR (ppt ppbÀ ) R NEMR (ppt ppbÀ ) R

Hydrogen cyanide 0.58 ± 0.68 0.57 1.11 ± 0.62 0.78 0.61 ± 0.31 0.80 Isocyanic acid 0.45 ± 0.96 0.28 1.44 ± 0.74 0.80 0.48 ± 0.40 0.59 Methyl isocyanate 0.97 ± 2.52 0.21 4.19 ± 2.81 0.71 1.00 ± 1.34 0.36 N-Formylformamide 0.03 ± 0.05 0.32 0.07 ± 0.03 0.85 0.03 ± 0.02 0.61 N,N-Dimethylformamide 0.04 ± 0.04 0.60 0.10 ± 0.06 0.79 0.03 ± 0.02 0.77 Acrylamide 0.04 ± 0.07 0.43 0.15 ± 0.05 0.90 0.04 ± 0.03 0.65 Formamide 0.07 ± 0.09 0.52 0.19 ± 0.07 0.88 0.07 ± 0.04 0.74 N-Methylformamide 0.11 ± 0.12 0.63 0.31 ± 0.15 0.83 0.11 ± 0.06 0.78 Note. Error reported is 2σ. BBP = bonﬁre burn period; NEMR = normalized excess mixing ratios.

PRIESTLEY ET AL. 11 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

from the BBP. As bonfires are only lit after dark, photochemical production cannot occur during this time and so cannot confound any increases in measured concentrations. By applying the [HNCOCO] metric to the NEMRs, it is possible to distinguish two distinct burning phases, flaming and smoldering (Figure 6). Table 5 and Figure 7 summarize the NEMRs for the identified species. Comparing the emission factor (EF) of a species as a function of the MCE (equation (2)) is best used to define the magnitude of emission during smoldering and flaming phase combustion (e.g., Yokelson et al., 1999).

Δ CO MCE 2 (2) ¼ ΔCO Δ CO þ 2

Calculating the MCE here is not possible as CO2 was not measured; however, by using typical literature values of MCE where ﬂaming and smol-

Figure 7. Normalized excess mixing ratio for HCN and seven selected isocya- dering EFHCN have been measured, it is possible to estimate the nate and amide compounds. concentrations of CO2 present due to BB. Using MCEs of 0.95 for ﬂaming emission and 0.88 for smoldering emission (Akagi et al., 2013), we esti-

mate that during flaming emission, CO2 concentrations are 19 times greater than CO concentrations but only 7.3 times greater during smoldering emission. Using CO2 + CO as a proxy for mass of material burned, 2.4 times more material was burned during the flaming phase than during smoldering. So while the NEMR appears greater for flaming emission than smoldering, it does not take into account the greater mass burned during the flaming emission stage. Normalizing the HCN NEMR to

[CO2] + [CO] (as a proxy for mass burned) for the different phases shows that the smoldering NEMR is higher than the flaming NEMR by about a third (32%) (S3). In all instances, the correlation of the compound and CO during the entire BBP is less good than the correlations observed when the BBP is separated into the two different burn phases. This indicates that separating the BBP produces more accurate NEMRs. Errors during the smoldering emission phase are likely to be overestimated as they include the transition period from flaming to smoldering. Together, this indicates that the NEMR is underestimated, when considering the BBP as a whole, if flaming combustion is occurring and may lead to an artificial reduction in the perceived relationship of a species to a specific combustion phase. For example, the correlation of MIC and CO over the entire BBP is poor (R2 = 0.21) indicating that there is a poor evidence for MIC emission during the BBP. However, if the burning phase is accounted for, the correlation during flaming emission is much higher (R2 = 0.71) suggesting that a strong source of MIC emission is occurring during this flaming emission phase. The 2σ error on the MIC BBP regression is 259% of the slope value, yet the flaming phase error is 67%, again indicating that by considering the burn phase, the emission ratio is more accurate. Figure 7 shows that the NEMRs for HCN, HNCO, and MIC are approximately an order of magnitude higher than the amides, which to our knowledge represent the first reported NEMRs for these compounds. 1 The NEMR for HCN during flaming emission is 1.11 ± 0.62 ppt ppbÀ and is comparable to NEMRs from boreal forest fires in North America and Siberia (Tereszchuk et al., 2013); an agricultural building fire (Brilli et al., 2014); and African savannah, tropical, and extratropical forest fires (Hornbrook et al., 2011) but is low com- 1 pared with other studies (Hornbrook et al., 2011, and references therein). Higher HCN NEMRs (5–12 ppt pptÀ )

have been detected in regions containing large cities where deposited NOx contributes to the nitrogen enrichment of the emission from local forest ﬁres (Yokelson et al., 2007), but as the fuels used in this instance

are unlikely to have accumulated NOx, this mechanism should not contribute to enhancing the nitrogen content of the fuel being burned. Emissions of nitrogen-containing VOCs such as HCN and HNCO are known to be highly variable depending on the type and origin of the fuel. Increasing the nitrogen content of a biogenic fuel type by 1% can increase the emission of nitrogen-containing VOCs by 2–6% (Coggon et al., 2016), and the inclusion of a diesel oxidation catalyst to a diesel engine can increase the HNCO NEMR by a factor of 30 (Jathar et al., 2017). Total isocyanate concentrations measured after burning various plastics can vary by 3 orders of magnitude

PRIESTLEY ET AL. 12 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

Table 6 Maximum Concentrations of Routinely Measured Pollutants Measured at the Whitworth Observatory With Their Respective NAQO Bonﬁre measurements 29/10/2014 to 11/11/2014 From 1/6/2014 to 28/2/2016 (includes bonﬁre night)

Pollutant NAQO (ppb) Sample period Cmax (ppb) Time Cmax (ppb) Time

NO2 105 1 h mean 46.76 5/11/2014 19:00 60.79 3/12/2014 17:00 SO2 132 1 h mean 1.78 5/11/2014 20:00 14.82 21/9/2014 12:00 SO2 47 24 h mean 0.69 9/11/2014 3.32 21/9/2014 CO 8,377 8 h running mean 1,551.78 5/11/2014 21:00 1,551.78 5/11/2014 21:00 Note. Dates are formatted as day/month/year. NAQO = UK National Air Quality Objectives.

depending on the precursor fuel nitrogen content, and accordingly, plastics containing no nitrogen do not produce any isocyanates (Blomqvist et al., 2003). This variability in fuel type is one potential reason that the HCN NEMR is low. Another reason may be the assumption that the enhancement in CO above the calculated background is entirely due to bonfire burning is not accurate enough, as other sources of CO, for example, fireworks and or enhanced contribution from cars, cause an underestimation in the NEMR. Unfortunately, no discernable tracer of firework activity was found using iodide CIMS. During street to city- scale dispersion experiments over (approximately) 2–5 km, using inert perflurocarbon gas tracers, the ratio of different tracers released at the same point remained constant over all distances measured (Martin et al., 2010, 2011; Wood et al., 2009). Depending on the prevailing wind speed, the tracer ratio did not return to background levels following cessation of tracer release for 30–60 min at the release point itself. In other experiments where inert tracer was heated to produce a buoyant plume, not only was it detected at much further distances downwind, the ratio was also preserved (Britter et al., 2002). While multiple sources will inevitably have different initial ratios (e.g., [HNCOCO]), those ratios should be preserved over at least kilometer-scale distances. It is impossible of course to categorically define the transition from a flaming phase to a smoldering phase, but given the very long lifetime of HNCO and CO with respect to travel time (based on when fires are likely to have been lit) the two distinct phases identified should be robust.

3.4. Air Quality The highest concentration of CO recorded during the BBP is 1,551 ppb (as an 8 h running mean) which is

comparable with summertime pollution events in London (Bannan et al., 2014). A maximum NO2 concentration of 43.76 ppb is measured during the BBP. Of the pollutants routinely measured at the Whitworth Observatory, none of the UK National Air Quality Objectives (NAQOs) (Defra, 2012) were breached during the measurement period at this measurement site. The highest maximum recorded concentrations of the pollutants at their NAQO sample periods were between 19:00 and 21:00 on bonﬁre night except the 24 h

mean SO2, which occurred 4 days later. As the Whitworth Observatory is approximately 15 m above ground level and 100 m from the nearest road, it is not considered a roadside/kerbside site, and so these data cannot convey any enhancement in concentration due to local vehicular emission at roadside where pollution levels and human exposure are often highest.

However, this analysis does demonstrate that NO2 and SO2 concentrations are higher at times in the year other than bonﬁre night due to other phenomena, typically local trafﬁc emission at times of low wind speed.

Although the SO2 concentrations at their highest are much lower than the NAQOs, NO2 concentrations are noteworthy (Table 6).

3.5. Health Impacts It is difﬁcult to translate concentrations of the trace gas species measured with the ToF-CIMS into impacts

on human health. Attention has mainly focused on particulate matter, NO2, CO, SO2, and O3. In a comprehensive review of the literature on the human impact of wildland fire smoke, which shares some of its source materials with what is burnt during bonfire night, consistent associations with mortality and respiratory morbidity were observed (Liu et al., 2015), although most of these wildfires would occur for longer periods than the one night evaluated here. Adetona et al. (2016) similarly concluded that for the general (exposed) public there was strong evidence of an association with acute respiratory effects,

PRIESTLEY ET AL. 13 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

weak evidence of acute cardiovascular effects and insufficient evidence for any conclusions regarding birth outcomes. In an attempt to put the findings of this study in context, we use the risk communication methodology of van der Zee et al. (2016) to communicate the risk from specific air pollutants in equivalent numbers of passively smoked cigarettes.

The maximum 1 h mean NO2 concentration encountered during the 2014 Bonfire night was 46.76 ppb, which, although below air quality limits, is approximately 13.82 ppb above the estimated regional concentration above the surface layer for 2014 in Manchester (Defra, 2017). The 1 h mean concentration of black 3 carbon fluctuated between 0 and 5 μgmÀ during the measurement period but increased to its maximum 3 of 21.11 μgmÀ during bonfire night. The average concentrations of these two pollutants from 16:00 to 04:00 over bonfire night are estimated to be the equivalent to 26.1 passively smoked cigarettes. The methodology described here, however, additionally identified trace amounts of many other chemicals generally not measured, many of which are known irritants and/or toxicants at higher (ppm) exposure levels: for example, in terms of workplace limit values MIC has a threshold limit value of 0.2 ppm (National Institute for Occupational Safety and Health, 2017), the Occupational Safety and Health Administration permissible exposure limit for HCN is 10 ppm (time weighted average) or 5 mg/m3 averaged over 15 min (U.S. Department of Labor, 2005) and may be carcinogenic or teratogenic. However, the health impact of these chemicals at the low parts per billion level at which they are observed here remains unknown, if any given the likely duration during which human exposure would occur.

4. Conclusions Here we present the first simultaneous online gas phase measurements of isocyanates, amides, and nitrates and nitro-organics using the ToF-CIMS during a 2 week period including bonfire night (5 November) and present, to our knowledge, the first reported NEMRs of amides to CO. Typical HCN concentrations of 0–50 ppt were measured before and after bonfire night with a maximum of 82 ppt measured during rush hours when wind speeds were low. Mean HNCO concentrations were 12 ppt, and a maximum of 144 ppt was measured at the same time as high HCN concentrations. While HNCO concentrations show evidence of a photochemical source, MIC does not and behaves more like HCN as a primary pollutant. Maximum bonfire night concentrations of HCN and HNCO were 1.24 ppb and 1.64 ppb, respectively. Nitrates, nitrated species, and amide enhancements were lower than those for HCN and HNCO at ~2–3.5 times higher than ambient levels. No one bonfire was directly sampled as a shallow inversion layer, and low wind speeds caused a pooling of outflow from many large- and small-scale emission sources. We identify that the bonfire burn period (BBP) lasted between approximately 16:00 and 04:00–05:00 when using HCN as a tracer. No tracer for fireworks was found using iodide CIMS.

In the absence of CO2 measurements and therefore MCE calculations, we use a HNCO:CO ratio of 0.1% to distinguish flaming and smoldering combustion (Roberts et al., 2011) and report normalized excess mixing ratios (NEMRs) relative to CO for HCN and a range of isocyanates and amides, many of which have not been reported before. The NEMRs separated by combustion phase are more accurate than when treating the BBP as a whole, with flaming phase showing greater enhancements than smoldering. We note that this does not 1 take into account the amount of mass burned. The flaming phase NEMR of HCN = 1.11 ± 0.62 ppt ppbÀ , while low is consistent with other studies of biogenic BB (e.g., Tereszchuk et al., 2013). The uncertainty of the mixed fuel types used in bonfire construction prevents a detailed analysis of whether fuel nitrogen content is consistent with this lower HCN NEMR. Future work should attempt to better understand the composition of the mixed fuel types used in anthropogenic open burning. The highest concentration of CO at this measurement site within a year and a half period (inclusive of this data) was measured during bonfire night, further confirming that this is a high-pollution event, yet match

those measured in London during summer (Bannan et al., 2014). Conversely, NO2 concentrations were higher at other times within that year and a half period at this site, indicating that other phenomena can contribute ﬁ more than bon re night to increased levels of NO2. NO2 concentrations measured here do not exceed current national air quality objectives (NAQOs) but equally do not represent kerbside measurements where the

majority of NO2 exceedances occur (McLean & Drabble, 2015).

PRIESTLEY ET AL. 14 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

The health impact of the exposures to the many identiﬁed chemicals in the parts per thousand range is unknown, although maximum 1 min average concentrations of HNCO during the BBP exceed 1 ppb, which has previously been considered above levels harmful to humans (Roberts et al., 2011). Initial assessment of a

combination of black carbon and NO2 indicates that this exposure, at the measurement location, would be equivalent to an average of 26.1 passively smoked cigarettes between 16:00 and 04:00 during bonﬁre night. As detection of toxic compounds becomes more routine, their inclusion in health burden quantiﬁcation methodologies would further contribute to understanding the true impact of BB events on human health.

Acknowledgments References This work was supported by the EAO NERC DTP. D. E. S. and M. A. H. K. Adetona, O., Reinhardt, T. E., Domitrovich, J., Broyles, G., Adetona, A. M., Kleinman, M. T., et al. (2016). Review of the health effects of wildland fi fi fi would like to thank NERC (NE/ re smoke on wildland re ghters and the public. Inhalation Toxicology, 28(3), 95–139. https://doi.org/10.3109/08958378.2016.1145771 K01501X). Processed data are available Akagi, S. K., Yokelson, R. J., Burling, I. R., Meinardi, S., Simpson, I., Blake, D. R., et al. (2013). Measurements of reactive trace gases and variable via the British Atmospheric Data Centre O3 formation rates in some South Carolina biomass burning plumes. Atmospheric Chemistry and Physics, 13(3), 1141–1165. https://doi.org/ (BADC) http://catalogue.ceda.ac.uk/ 10.5194/acp-13-1141-2013 uuid/71a34def1d104f1e925f1a6f7- Akagi, S. K., Yokelson, R. J., Wiedinmyer, C., Alvarado, M. J., Reid, J. S., Karl, T., et al. (2011). Emission factors for open and domestic biomass d12ac21, are archived at the University burning for use in atmospheric models. Atmospheric Chemistry and Physics, 11(9), 4039–4072. Retrieved from http://www.atmos-chem- of Manchester, and are available phys.net/11/4039/2011/, https://doi.org/10.5194/acp-11-4039-2011 on request. Amann, M., Purohit, P., Bhanarkar, A. D., Bertok, I., Borken-Kleefeld, J., Cofala, J., et al. (2017). Managing future air quality in megacities: A case study for Delhi. Atmospheric Environment, 161, 99–111. Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S1352231017302856, https://doi.org/10.1016/j.atmosenv.2017.04.041 Ambrose, J. L., Zhou, Y., Haase, K., Mayne, H. R., Talbot, R., & Sive, B. C. (2012). A gas chromatographic instrument for measurement of hydrogen cyanide in the lower atmosphere. Atmospheric Measurement Techniques, 5(6), 1229–1240. Retrieved from http://www.atmos- meas-tech.net/5/1229/2012/, https://doi.org/10.5194/amt-5-1229-2012 Andreae, M. O., & Merlet, P. (2001). Emission of trace gases and aerosols from biomass burning. Global Biogeochemical Cycles, 15(4), 955–966. https://doi.org/10.1029/2000GB001382 Bannan, T. J., Bacak, A., Muller, J. B. A., Booth, A. M., Jones, B., le Breton, M., et al. (2014). Importance of direct anthropogenic emissions of formic acid measured by a chemical ionisation mass spectrometer (CIMS) during the winter ClearfLo campaign in London, January 2012. Atmospheric Environment, 83, 301–310. https://doi.org/10.1016/j.atmosenv.2013.10.029 Bannan, T. J., Booth, A. M., Bacak, A., Muller, J. B. A., Leather, K. E., Le Breton, M., et al. (2015). The first UK measurements of nitryl chloride using a chemical ionisation mass spectrometer in central London in the summer of 2012, and an investigation of the role of Cl atom oxidation. Journal of Geophysical Research: Atmospheres, 120, 5638–5657. https://doi.org/10.1002/2014JD022629 Bannan, T. J., Booth A. M., Jones B. T., O’Meara S., Barley M. H., Riipinen I., et al. (2017). Measured saturation vapor pressures of phenolic and nitro-aromatic compounds. Environmental Science & Technology, 51(7) 3922–3928. https://doi.org/10.1021/acs.est.6b06364 Berndt, T., & Böge, O. (2006). Formation of phenol and carbonyls from the atmospheric reaction of OH radicals with benzene. Physical Chemistry Chemical Physics (PCCP), 8(10), 1205–1214. https://doi.org/10.1039/b514148f Bertram, T. H., Thornton, J. A., & Riedel, T. P. (2009). An experimental technique for the direct measurement of N2O5 reactivity on ambient particles. Atmospheric Measurement Techniques, 2(1), 231–242. Retrieved from www.atmos-meas-tech.net/2/231/2009/, https://doi.org/ 10.5194/amt-2-231-2009 Blomqvist, P., Hertzberg, T., Dalene, M., & Skarping, G. (2003). Isocyanates, aminoisocyanates and amines from fires—A screening of common materials found in buildings. Fire and Materials, 27 (6), 275–294. https://doi.org/10.1002/fam.836 Bonde, J. P., Flachs, E. M., Rimborg, S., Glazer, C. H., Giwercman, A., Ramlau-Hansen, C. H., et al. (2016). The epidemiologic evidence linking prenatal and postnatal exposure to endocrine disrupting chemicals with male reproductive disorders: A systematic review and meta-analysis. Human Reproduction Update (Hypothesis ), 4,1–22. Borduas, N., da Silva, G., Murphy, J. G., & Abbatt, J. P. D. (2015). Experimental and theoretical understanding of the gas phase oxidation of atmospheric amides with OH radicals: Kinetics, products, and mechanisms. Journal of Physical Chemistry A, 119(19), 4298–4308. https:// doi.org/10.1021/jp503759f Borduas, N., Place, B., Wentworth, G. R., Abbatt, J. P. D., & Murphy, J. G. (2016). Solubility and reactivity of HNCO in water: Insights into HNCO’s fate in the atmosphere. Atmospheric Chemistry and Physics, 16(2), 703–714. https://doi.org/10.5194/acp-16-703-2016 Brilli, F., Gioli, B., Ciccioli, P., Zona, D., Loreto, F., Janssens, I. A., & Ceulemans, R. (2014). Proton transfer reaction time-of-flight mass spectrometric (PTR-TOF-MS) determination of volatile organic compounds (VOCs) emitted from a biomass fire developed under stable nocturnal conditions. Atmospheric Environment, 97 , 54–67. https://doi.org/10.1016/j.atmosenv.2014.08.007 Bristow, A. W. T., & Webb, K. S. (2003). Intercomparison study on accurate mass measurement of small molecules in mass spectrometry. Journal of the American Society for Mass Spectrometry, 14(10), 1086–1098. https://doi.org/10.1016/S1044-0305(03)00403-3 Brophy, P., & Farmer, D. K. (2015). A switchable reagent ion high resolution time-of-flight chemical ionization mass spectrometer for real-time measurement of gas phase oxidized species: Characterization from the 2013 southern oxidant and aerosol study. Atmospheric Measurement Techniques, 8(7), 2945–2959. Retrieved from http://www.atmos-meas-tech.net/8/2945/2015/, https://doi.org/10.5194/ amt-8-2945-2015 Bunkan, A. J. C., Hetzler, J., Mikoviny, T., Wisthaler, A., Nielsen, C. J., & Olzmann, M. (2015). The reactions of N-methylformamide and N,N-dimethylformamide with OH and their photo-oxidation under atmospheric conditions: Experimental and theoretical studies. Physical Chemistry Chemical Physics (PCCP), 17 (10), 7046–7059. Retrieved from http://pubs.rsc.org/en/content/articlehtml/2015/cp/ c4cp05805d, https://doi.org/10.1039/C4CP05805D Caird, S., Roy, R., & Herring, H. (2008). Improving the energy performance of UK households: Results from surveys of consumer adoption and use of low- and zero-carbon technologies. Energy Efficiency, 1(2), 149–166. https://doi.org/10.1007/s12053-008-9013-y Chandra, B. P., & Sinha, V. (2016). Contribution of post-harvest agricultural paddy residue fires in the N.W. Indo-Gangetic Plain to ambient carcinogenic benzenoids, toxic isocyanic acid and carbon monoxide. Environment International, 88, 187–197. https://doi.org/10.1016/ j.envint.2015.12.025 Chhabra, P. S., Lambe, A. T., Canagaratna, M. R., Stark, H., Jayne, J. T., Onasch, T. B., et al. (2015). Application of high-resolution time-of-flight chemical ionization mass spectrometry measurements to estimate volatility distributions of α-pinene and naphthalene oxidation

PRIESTLEY ET AL. 15 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

products. Atmospheric Measurement Techniques, 8(1), 1–18. Retrieved from http://www.atmos-meas-tech.net/8/1/2015/, https://doi.org/ 10.5194/amt-8-1-2015 Coggon, M. M., Veres, P. R., Yuan, B., Koss, A., Warneke, C., Gilman, J. B., et al. (2016). Emissions of nitrogen-containing organic compounds from the burning of herbaceous and arboraceous biomass: Fuel composition dependence and the variability of commonly used nitrile tracers. Geophysical Research Letters, 43, 9903–9912. https://doi.org/10.1002/2016GL070562 Defra (2012). National air quality objectives. Retrieved from https://uk-air.defra.gov.uk/assets/documents/National_air_quality_objectives. pdf, Accessed April 20, 2017. Defra (2017). Background maps. Retrieved from https://uk-air.defra.gov.uk/data/laqm-background-home, Accessed February 23, 2017. Drewnick, F., Hings, S. S., Curtius, J., Eerdekens, G., & Williams, J. (2006). Measurement of fine particulate and gas-phase species during the new year’s fireworks 2005 in Mainz, Germany. Atmospheric Environment, 40(23), 4316–4327. Retrieved from http://www.sciencedirect. com/science/article/pii/S1352231006003499, https://doi.org/10.1016/j.atmosenv.2006.03.040 Dyke, P., Coleman, P., & James, R. (1997). Dioxins in ambient air, bonfire night 1994. Chemosphere, 34(5–7), 1191–1201. https://doi.org/ 10.1016/S0045-6535(97)00418-9 Ehn, M., Thornton, J. A., Kleist, E., Sipilä, M., Junninen, H., Pullinen, I., et al. (2014). A large source of low-volatility secondary organic aerosol. Nature, 506(7489), 476–479. https://doi.org/10.1038/nature13032/npapers2://publication/doi/10.1038/nature13032 Farrar, N. J., Smith, K. E. C., Lee, R. G. M., Thomas, G. O., Sweetman, A. J., & Jones, K. C. (2004). Atmospheric emissions of polybrominated diphenyl ethers and other persistent organic pollutants during a major anthropogenic combustion event. Environmental Science and Technology, 38(6), 1681–1685. https://doi.org/10.1021/es035127d fi Fuller, G. W., Tremper, A. H., Baker, T. D., Yttri, K. E., & Butter eld, D. (2014). Contribution of wood burning to PM10 in London. Atmospheric Environment, 87 , 87–94. https://doi.org/10.1016/j.atmosenv.2013.12.037 Gescher, A. (1990). 2.3 - Formamide A2 - BUHLER, D.R. In D. J. B. T.-N. And P. S (Ed.), Ethel Browning’s toxicity and metabolism of industrial solvents, (Second E. REED ed.pp. 160–164). Amsterdam: Elsevier. Retrieved from http://www.sciencedirect.com/science/article/pii/ B9780444813169500311 Harrad, S. J., & Laurie, L. (2005). Concentrations, sources and temporal trends in atmospheric polycyclic aromatic hydrocarbons in a major conurbation. Journal of Environmental Monitoring, 7 (7), 722–727. https://doi.org/10.1039/b503983e Harrison, T., & Shallcross, D. E. (2011). Smoke is in the air: How fireworks affect air quality Did you realise that fireworks cause measurable air pollution? Science in School, 21, 47–51. Henriks-Eckerman, M.-L., Välimaa, J., Rosenberg, C., Peltonen, K., & Engström, K. (2002). Exposure to airborne isocyanates and other thermal degradation products at polyurethane-processing workplaces. Journal of Environmental Monitoring, 4(5), 717–721. Retrieved from http:// xlink.rsc.org/?DOI=b206339p. https://doi.org/10.1039/B206339P Hornbrook, R. S., Blake, D. R., Diskin, G. S., Fried, A., Fuelberg, H. E., Meinardi, S., et al. (2011). Observations of nonmethane organic compounds during ARCTAS—Part 1: Biomass burning emissions and plume enhancements. Atmospheric Chemistry and Physics, 11(21), 11,103–11,130. https://doi.org/10.5194/acp-11-11103-2011 Iinuma, Y., Böge, O., & Herrmann, H. (2010). Methyl-nitrocatechols: Atmospheric tracer compounds for biomass burning secondary organic aerosols. Environmental Science and Technology, 44(22), 8453–8459. https://doi.org/10.1021/es102938a Iyer, S., Lopez-Hilfiker F., Lee B. H., Thornton, J. A., & Kurtén T. (2016). Modeling the detection of organic and inorganic compounds using iodide-based chemical ionization. The Journal of Physical Chemistry A, 120(4), 576–587. https://doi.org/10.1021/acs.jpca.5b09837 Jathar, S. H., Heppding, C., Link, M. F., Farmer, D. K., Akherati, A., Kleeman, M. J., et al. (2017). Investigating diesel engines as an atmospheric source of isocyanic acid in urban areas. Atmospheric Chemistry and Physics, 17 (14), 8959–8970. https://doi.org/10.5194/acp-17- 8959-2017 Junninen, H., Ehn, M., Petäjä, T., Luosujärvi, L., Kotiaho, T., Kostiainen, R., et al. (2010). A high-resolution mass spectrometer to measure atmospheric ion composition. Atmospheric Measurement Techniques, 3(4), 1039–1053. Retrieved from http://www.atmos-meas-tech.net/ 3/1039/2010/, https://doi.org/10.5194/amt-3-1039-2010 Kim, K. H., Pandey, S. K., Kim, Y. H., Sohn, J. R., & Oh, J. M. (2015). Emissions of amides (N,N-dimethylformamide and formamide) and other obnoxious volatile organic compounds from different mattress textile products. Ecotoxicology and Environmental Safety, 114, 350–356. https://doi.org/10.1016/j.ecoenv.2014.07.008 Knighton, W. B., Fortner, E. C., Midey, A. J., Viggiano, A. A., Herndon, S. C., Wood, E. C., & Kolb, C. E. (2009). HCN detection with a proton transfer reaction mass spectrometer. International Journal of Mass Spectrometry, 283(1–3), 112–121. https://doi.org/10.1016/j.ijms.2009.02.013 Le Breton, M., Bacak, A., Muller, J. B. A., O’Shea, S. J., Xiao, P., Ashfold, M. N. R., et al. (2013). Airborne hydrogen cyanide measurements using a chemical ionisation mass spectrometer for the plume identification of biomass burning forest fires. Atmospheric Chemistry and Physics, 13(18), 9217–9232. https://doi.org/10.5194/acp-13-9217-2013 fi Le Breton, M., Bacak, A., Muller, J. B. A., Bannan, T. J., Kennedy, O., Ouyang, B., et al. (2014). The rst airborne comparison of N2O5 measurements over the UK using a CIMS and BBCEAS during the RONOCO campaign. Analytical Methods, 6(24), 9731–9743. Retrieved from http://pubs.rsc.org/en/content/articlehtml/2014/ay/c4ay02273d, https://doi.org/10.1039/C4AY02273D Le Breton, M., McGillen, M. R., Muller, J. B. A., Bacak, A., Shallcross, D. E., Xiao, P., et al. (2012). Airborne observations of formic acid using a chemical ionization mass spectrometer. Atmospheric Measurement Techniques, 5(12), 3029–3039. Leather, K. E., McGillen, M. R., & Percival, C. J. (2010). Temperature-dependent ozonolysis kinetics of selected alkenes in the gas phase: An experimental and structure-activity relationship (SAR) study. Physical Chemistry Chemical Physics (PCCP), 12(12), 2935–2943. https://doi. org/10.1039/b919731a Lee, B. H., Lopez-Hilfiker, F. D., Mohr, C., Kurtén, T., Worsnop, D. R., & Thornton, J. A. (2014). An iodide-adduct high-resolution time-of-flight chemical-ionization mass spectrometer: Application to atmospheric inorganic and organic compounds. Environmental Science & Technology, 48(11), 6309–6317. https://doi.org/10.1021/es500362a Li, Q., Jacob, D. J., Yantosca, R. M., Heald, C. L., Singh, H. B., Koike, M., et al. (2003). A global three-dimensional model analysis of the atmospheric budgets of HCN and CH3CN: Constraints from aircraft and ground measurements. Journal of Geophysical Research, 108(D21), 8827. https://doi.org/10.1029/2002JD003075 Li, Q., Palmer, P. I., Pumphrey, H. C., Bernath, P., & Mahieu, E. (2009). What drives the observed variability of HCN in the troposphere and lower stratosphere? Atmospheric Chemistry and Physics, 9(21), 8531–8543. Retrieved from http://www.atmos-chem-phys.net/9/8531/2009/, https://doi.org/10.5194/acp-9-8531-2009 Liu, J. C., Pereira, G., Uhl, S. A., Bravo, M. A., & Bell, M. L. (2015). A systematic review of the physical health impacts from non-occupational exposure to wildfire smoke. Environmental Research, 136, 120–132. https://doi.org/10.1016/j.envres.2014.10.015 Lopez-Hilfiker, F. D., Iyer, S., Mohr, C., Lee, B. H., D’Ambro, E. L., Kurtén, T., & Thornton, J. A. (2015). Constraining the sensitivity of iodide adduct chemical ionization mass spectrometry to multifunctional organic molecules using the collision limit and thermodynamic stability of

PRIESTLEY ET AL. 16 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

iodide ion adducts. Atmospheric Measurement Techniques Discussions, 8(10), 10,875–10,896. Retrieved from http://www.atmos-meas-tech- discuss.net/8/10875/2015/, https://doi.org/10.5194/amtd-8-10875-2015 Lu, Z., Hebert, V. R., & Miller, G. C. (2014). Gas-phase reaction of methyl isothiocyanate and methyl isocyanate with hydroxyl radicals under static relative rate conditions. Journal of Agricultural and Food Chemistry, 62(8), 1792–1795. Retrieved from http://www.ncbi.nlm.nih.gov/ pubmed/24483206 Lüttke, J., Scheer, V., Levsen, K., Wünsch, G., Neil Cape, J., Hargreaves, K. J., et al. (1997). Occurrence and formation of nitrated phenols in and out of cloud. Atmospheric Environment, 31(16), 2637–2648. https://doi.org/10.1016/S1352-2310(96)00229-4 Mari, M., Harrison, R. M., Schuhmacher, M., Domingo, J. L., & Pongpiachan, S. (2010). Inferences over the sources and processes affecting polycyclic aromatic hydrocarbons in the atmosphere derived from measured data. Science of the Total Environment, 408(11), 2387–2393. https://doi.org/10.1016/j.scitotenv.2010.01.054 Martin, D., Petersson, K. F., & Shallcross, D. E. (2011). The use of cyclic perfluoroalkanes and SF6 in atmospheric dispersion experiments. Quarterly Journal of the Royal Meteorological Society, 137 (661), 2047–2063. Martin, D., Price, C. S., White, I. R., Nickless, G., Petersson, K. F., Britter, R. E., et al. (2010). Urban tracer dispersion experiments during the second DAPPLE field campaign in London 2004. Atmospheric Environment, 44(25), 3043–3052. McLean, A. & Drabble, J. (2015). 2014 detailed air quality assessment for Greater Manchester The Greater Manchester Combined Authority. Retrieved from http://www.manchester.gov.uk/downloads/download/4166/air_quality_reports Mohr, C., Lopez-Hilfiker, F. D., Zotter, P., Prévôt, A. S., Xu, L., Ng, N. L., et al. (2013). Contribution of nitrated phenols to wood burning Brown carbon light absorption in Detling, United Kingdom during winter time. Environmental Science & Technology, 47 (12), 6316–6324. https:// doi.org/10.1021/es400683v Molina, L. T., Kolb, C. E., de Foy, B., Lamb, B. K., Brune, W. H., Jimenez, J. L., et al. (2007). Air quality in North America’s most populous city–overview of the MCMA-2003 campaign. Atmospheric Chemistry and Physics, 7 (10), 2447–2473. https://doi.org/10.5194/ acp-7-2447-2007 Moreno, T., Querol, X., Alastuey, A., Cruz Minguillón, M., Pey, J., Rodriguez, S., et al. (2007). Recreational atmospheric pollution episodes: Inhalable metalliferous particles from firework displays. Atmospheric Environment, 41(5), 913–922. Retrieved from http://linkinghub.elsevier.com/retrieve/pii/S1352231006009745, https://doi.org/10.1016/j.atmosenv.2006.09.019 Mouly, T. A., & Toms, L. L. (2016). Breast cancer and persistent organic pollutants (excluding DDT): A systematic literature review. Environmental Science and Pollution Research, 23(22), 22,385–22,407. https://doi.org/10.1007/s11356-016-7577-1 Moussa, S. G., Leithead, A., Li, S.-M., Chan, T. W., Wentzell, J. J. B., Stroud, C., et al. (2016). Emissions of hydrogen cyanide from on-road gasoline and diesel vehicles. Atmospheric Environment, 131, 185–195. https://doi.org/10.1016/j.atmosenv.2016.01.050 Nadal, M., Marquès, M., Mari, M., & Domingo, J. L. (2015). Climate change and environmental concentrations of POPs: A review. Environmental Research, 143, 177–185. https://doi.org/10.1016/j.envres.2015.10.012 National Institute for Occupational Safety and Health (2017). CDC—Benzoic acid—International Chemical Safety Cards - NIOSH. Retrieved from https://www.cdc.gov/niosh/ipcsneng/neng0103.html Pongpiachan, S., Tipmanee, D., Khumsup, C., Kittikoon, I., & Hirunyatrakul, P. (2015). Assessing risks to adults and preschool children posed by PM2.5-bound polycyclic aromatic hydrocarbons (PAHs) during a biomass burning episode in northern Thailand. Science of the Total Environment, 508, 435–444. https://doi.org/10.1016/j.scitotenv.2014.12.019 Reyes-Villegas, E., Priestley, M., Ting, Y. C., Haslett, S., Bannan, T., le Breton, M., et al. (2017). Supplemental information for: Simultaneous aerosol mass spectrometry and chemical ionisation mass spectrometry measurements during a biomass burning event in the UK: Insights into nitrate chemistry. Atmospheric Chemistry and Physics Discussions, (July), 1–22. https://doi.org/10.5194/acp-2017-605 Roberts, J. M., Veres, P., Warneke, C., Neuman, J. A., Washenfelder, R. A., Brown, S. S., et al. (2010). Measurement of HONO, HNCO, and other inorganic acids by negative-ion proton-transfer chemical-ionization mass spectrometry (NI-PT-CIMS): Application to biomass burning emissions. Atmospheric Measurement Techniques, 3(4), 981–990. https://doi.org/10.5194/amt-3-981-2010 Roberts, J. M., Veres, P. R., Cochran, A. K., Warneke, C., Burling, I. R., Yokelson, R. J., et al. (2011). Isocyanic acid in the atmosphere and its possible link to smoke-related health effects. Proceedings of the National Academy of Sciences of the United States of America, 108(22), 8966–8971. https://doi.org/10.1073/pnas.1103352108 Roberts, J. M., Veres, P. R., VandenBoer, T. C., Warneke, C., Graus, M., Williams, E. J., et al. (2014). New insights into atmospheric sources and sinks of isocyanic acid, HNCO, from recent urban and regional observations. Journal of Geophysical Research: Atmospheres, 119, 1060–1072. https://doi.org/10.1002/2013JD019931 Sarkar, C., Sinha, V., Kumar, V., Rupakheti, M., Panday, A., Mahata, K. S., et al. (2016). Overview of VOC emissions and chemistry from PTR-TOF-MS measurements during theSusKat-ABCcampaign:Highacetaldehyde, isoprene and isocyanic acid in wintertime air of the Kathmandu Valley. Atmospheric Chemistry and Physics, 16(6), 3979–4003. https://doi.org/10.5194/acp-16- 3979-2016 Sarkar, C., Sinha, V., Sinha, B., Panday, A. K., Rupakheti, M., & Lawrence, M. G. (2017). Source apportionment of NMVOCs in the Kathmandu Valley during the SusKat-ABC international field campaign using positive matrix factorization. Atmospheric Chemistry and Physics, 17 (13), 8129–8156. https://doi.org/10.5194/acp-17-8129-2017 Sekler, M. S., Levi, Y., Polyak, B., Novoa, A., Dunlop, P. S. M., Byrne, J. A., & Marks, R. S. (2004). Monitoring genotoxicity during the photocatalytic degradation of p-nitrophenol. Journal of Applied Toxicology, 24(5), 395–400. https://doi.org/10.1002/jat.1029 Shim, C., Wang, Y., Singh, H. B., Blake, D. R., & Guenther, A. B. (2007). Source characteristics of oxygenated volatile organic compounds and hydrogen cyanide. Journal of Geophysical Research, 112, D10305. https://doi.org/10.1029/2006JD007543 Stockwell, C. E., Veres, P. R., Williams, J., & Yokelson, R. J. (2015). Characterization of biomass burning emissions from cooking fires, peat, crop residue, and other fuels with high-resolution proton-transfer-reaction time-of-flight mass spectrometry. Atmospheric Chemistry and Physics, 15(2), 845–865. Retrieved from http://www.atmos-chem-phys.net/15/845/2015/, https://doi.org/10.5194/ acp-15-845-2015 Tereszchuk, K. A., González Abad, G., Clerbaux, C., Hadji-Lazaro, J., Hurtmans, D., Coheur, P. F., & Bernath, P. F. (2013). ACE-FTS observations of pyrogenic trace species in boreal biomass burning plumes during BORTAS. Atmospheric Chemistry and Physics, 13(9), 4529–4541. https:// doi.org/10.5194/acp-13-4529-2013 U.S. Department of Labor (2005). Chemical sampling information | chlorobenzene | occupational safety and health administration., P. Accessed 23/02/2017. Retrieved from https://www.osha.gov/dts/chemicalsampling/data/CH_227000.html UK Government, D. for C. and L. (2015). Celebrating with bonfires and fireworks. Retrieved from https://www.gov.uk/government/publications/burning-of-waste-on-campfires-and-bonfires United Nations (2016). The world’s cities in 2016—Data booklet (ST/ESA/ SER.A/392). The World’s Cities in 2016. Retrieved from www. unpopulation.org.

PRIESTLEY ET AL. 17 Journal of Geophysical Research: Atmospheres 10.1002/2017JD027316

van der Zee, S. C., Fischer, P. H., & Hoek, G. (2016). Air pollution in perspective: Health risks of air pollution expressed in equivalent numbers of passively smoked cigarettes. Environmental Research, 148, 475–483. https://doi.org/10.1016/j.envres.2016.04.001 Vassura, I., Venturini, E., Marchetti, S., Piazzalunga, A., Bernardi, E., Fermo, P., & Passarini, F. (2014). Markers and influence of open biomass burning on atmospheric particulate size and composition during a major bonfire event. Atmospheric Environment, 82, 218–225. Retrieved from http://www.sciencedirect.com/science/article/pii/S1352231013007905, https://doi.org/10.1016/j.atmosenv.2013.10.037 Veres, P. R., & Roberts, J. M. (2015). Development of a photochemical source for the production and calibration of acyl peroxynitrate compounds. Atmospheric Measurement Techniques, 8(5), 2225–2231. Retrieved from http://www.atmos-meas-tech.net/8/2225/2015/, https://doi.org/10.5194/amt-8-2225-2015 Veres, P. R., Roberts, J. M., Wild, R. J., Edwards, P. M., Brown, S. S., Bates, T. S., et al. (2015). Peroxynitric acid (HO2NO2) measurements during the UBWOS 2013 and 2014 studies using iodide ion chemical ionization mass spectrometry. Atmospheric Chemistry and Physics, 15(14), 8101–8114. Retrieved from http://www.atmos-chem-phys.net/15/8101/2015/, https://doi.org/10.5194/acp-15-8101-2015 Walser, M. L., Desyaterik, Y., Laskin, J., Laskin, A., & Nizkorodov, S. A. (2008). High-resolution mass spectrometric analysis of secondary organic aerosol produced by ozonation of limonene. Physical Chemistry Chemical Physics, 10(7), 1009–1022. Wood, C. R., Arnold, S. J., Balogun, A. A., Barlow, J. F., Belcher, S. E., Britter, R. E., et al. (2009). Dispersion experiments in central London: The 2007 DAPPLE project. Bulletin of the American Meteorological Society, 90(7), 955–969. Woodrow, J. E., LePage, J. T., Miller, G. C., & Hebert, V. R. (2014). Determination of methyl isocyanate in outdoor residential air near metam-sodium soil fumigations. Journal of Agricultural and Food Chemistry, 62(36), 8921–8927. https://doi.org/10.1021/jf501696a Woodward-Massey, R., Taha, Y. M., Moussa, S. G., & Osthoff, H. D. (2014). Comparison of negative-ion proton-transfer with iodide ion chemical ionization mass spectrometry for quantification of isocyanic acid in ambient air. Atmospheric Environment, 98, 693–703. Retrieved from http://www.sciencedirect.com/science/article/pii/S1352231014007079, https://doi.org/10.1016/j.atmosenv.2014.09.014 World Health Organization (2006). WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide: global update 2005: summary of risk assessment (pp.1–22). Geneva: World Health Organization. World Health Organization (2015). Residential heating with wood and coal: Health impacts and policy options in Europe and North America. Retrieved from http://ehsdiv.sph.berkeley.edu/krsmith/publications/2015/ResidentialHeatWoodCoal.pdf Yao, L., Wang, M.-Y., Wang, X.-K., Liu, Y.-J., Chen, H.-F., Zheng, J., et al. (2016). Detection of atmospheric gaseous amines and amides by a high resolution time-of-flight chemical ionization mass spectrometer with protonated ethanol reagent ions. Atmospheric Chemistry and Physics Discussions,1–32. Retrieved from http://www.atmos-chem-phys-discuss.net/acp-2016-484/ Yokelson, R. J., Goode, J. G., Ward, D. E., Susott, R. A., Babbitt, R. E., Wade, D. D., et al. (1999). Emissions of formaldehyde, acetic acid, methanol, and other trace gases from biomass fires in North Carolina measured by airborne Fourier transform infrared spectroscopy. Journal of Geophysical Research, 104(D23), 30,109–30,125. https://doi.org/10.1029/1999JD900817 Yokelson, R. J., Griffith, D. W. T., & Ward, D. E. (1996). Open-path Fourier transform infrared studies of large-scale laboratory biomass fires. Journal of Geophysical Research, 101(D15), 21,067–21,080. https://doi.org/10.1029/96JD01800 Yokelson, R. J., Urbanski, S. P., Atlas, E. L., Toohey, D. W., Alvarado, E. C., Crounse, J. D., et al. (2007). Emissions from forest fires near Mexico City. Atmospheric Chemistry and Physics, 7 (21), 5569–5584. https://doi.org/10.5194/acp-7-5569-2007 Yuan, B., Liggio, J., Wentzell, J., Li, S. M., Stark, H., Roberts, J. M., et al. (2015). Secondary formation of nitrated phenols: Insights from observations during the Uintah Basin Winter Ozone Study (UBWOS) 2014. Atmospheric Chemistry and Physics Discussions, 15(20), 28,659–28,697. Retrieved from http://www.atmos-chem-phys-discuss.net/15/28659/2015/, https://doi.org/10.5194/acpd-15-28659-2015

PRIESTLEY ET AL. 18

Journal of Geophysical Research: Atmospheres

Supporting Information for

Observations of isocyanate, amide, nitrate and nitro compounds from an anthropogenic biomass burning event using a TOF-CIMS

1 1† 1 1 Michael Priestley , Michael Le Breton , Thomas J. Bannan , Kimberly E. Leather , Asan Bacak1, Ernesto Reyes-Villegas1, Frank De Vocht4, Beth M. A. Shallcross,2 Toby Brazier3, M. Anwar Khan3, James Allan1,5, Dudley E. Shallcross3, Hugh Coe1, Carl J. Percival1‖.

1Centre for Atmospheric Science, School of Earth and Environmental Sciences, University of Manchester, Manchester, M13 9PL, UK. 2Division of Pharmacy and Optometry, University of Manchester, Manchester, M13 9PL, UK. 3School of Chemistry, The University of Bristol, Cantock’s Close BS8 1TS, UK. 4School of Social and Community Medicine, The University of Bristol, Canynge Hall, 39 Whatley Road, BS8 2PS, UK. 5National Centre for Atmospheric Science, University of Manchester, Manchester, M13 9PL, UK. †Now at Department of Chemistry and Molecular Biology, University of Gothenburg, 412 96 Göteborg, Sweden. ‖Now at Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109. Corresponding author: Carl Percival ([email protected])

Contents of this file Figures S1 to S2 Tables S3

Figure S1. Cross calibration of HCN on the TOF-CIMS from the Quad-CIMS. First plot shows time series of HCN on both instruments. The second plot shows HCN from each instrument plotted against each other indicating excellent linearity between the two measurements.

Figure S2. Time series of NOx concentration (ppb) during the bonfire burn period (BBP) as a function of wind direction (deg) coloured by wind speed (m s-1). Little correlation between wind direction and NOx concentration is observable at any of the sites.

HCN ΔCO + ΔCO NEMR / ΔCO + ΔCO 2 2 Emission NEMR ΔCO Derived as proxy as proxy for MCE+ Phase / ppt / ppb= ΔCO for mass mass burned / 2 ppb-1 = / ppb burned / ppb ppt ppb-1 / ppb Flaming 1.11 100 0.95 1900 2000 9.99 Smouldering 0.61 100 0.88 733 833 13.18 =From this work. +From Akagi et al., 2013.

Table S3. Table showing the HCN NEMR during flaming emission is higher than during smouldering emission. Using literature values of MCE typical for flaming and smouldering emission where the emission factor for HCN has been determined, a derived excess CO2 concentration (ΔCO2) is calculated and summed to excess CO

(ΔCO) and used as a proxy for mass burned (ΔCO + ΔCO2). During flaming emission,

ΔCO2 is 19 times greater than ΔCO but only 7.3 times greater during smouldering emission. During flaming ΔCO + ΔCO2 is 2.4 times greater than during smouldering, indicating 2.4 times more material is burned. Normalising the NEMR’s to ΔCO + ΔCO2 shows smouldering emission is greater than flaming by ~30%, which is consistent with other studies (Akagi et al. 2013). Derived �CO2 was calculated using the following equation: �� = � � ( ) − � �� 4. Paper 2. Observations of organic and inorganic chlorinated compounds and their contribution to chlorine radical concentrations in an urban environment in Northern Europe during the wintertime

Michael Priestley, Michael le Breton, Thomas J. Bannan, Stephen D. Worrall, Asan Bacak, Andrew R. D. Smedley, Ernesto Reyes-Villegas, Archit Mehra, James Allan, Ann R. Webb, Dudley E. Shallcross, Hugh Coe, Carl J. Percival

Published ACP: 21 September 2018 doi: 10.5194/acp-18-13481-2018

Research Highlights:

A range of organic and inorganic chlorine containing compounds is identified in ambient urban air.

Their contribution to the steady state Cl radical concentration indicates that Cl2

and ClNO2 are a much greater source than HOCl or ClOVOCs.

The Cl2 contribution to Cl is highly variable and dependent on shortwave radiation

to both form and photolyse Cl2.

Author Contributions:

Data collection was performed by Michael Priestley, Michael le Breton, Thomas J. Bannan and Asan Bacak. Laboratory work was performed by Michael Priestley, Stephen D. Worrall, Archit Mehra and Asan Bacak. AMS data products and interpretation were provided by Ernesto Reyes‐Villegas and James Allan. Photolysis data was provided by Andrew R. D. Smedley and Ann R. Webb. Data interpretation was aided by Dudley E. Shallcross. The article was written by Michael Priestley under the supervision of Hugh Coe and Carl J. Percival.

Blank page

Atmos. Chem. Phys., 18, 13481–13493, 2018 https://doi.org/10.5194/acp-18-13481-2018 © Author(s) 2018. This work is distributed under the Creative Commons Attribution 4.0 License.

Observations of organic and inorganic chlorinated compounds and their contribution to chlorine radical concentrations in an urban environment in northern Europe during the wintertime

Michael Priestley1, Michael le Breton1,a, Thomas J. Bannan1, Stephen D. Worrall1,b, Asan Bacak1, Andrew R. D. Smedley1,c, Ernesto Reyes-Villegas1, Archit Mehra1, James Allan1,2, Ann R. Webb1, Dudley E. Shallcross3, Hugh Coe1, and Carl J. Percival1,d 1Centre for Atmospheric Science, School of Earth and Environmental Sciences, University of Manchester, Manchester, M13 9PL, UK 2National Centre for Atmospheric Science, University of Manchester, Manchester, M13 9PL, UK 3School of Chemistry, The University of Bristol, Cantock’s Close BS8 1TS, UK anow at: Department of Chemistry and Molecular Biology, University of Gothenburg, 412 96 Gothenburg, Sweden bnow at: School of Materials, University of Manchester, Manchester, M13 9PL, UK cnow at: School of Mathematics, University of Manchester, Manchester, M13 9PL, UK dnow at: Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109, USA

Correspondence: Carl Percival ([email protected])

Received: 6 March 2018 – Discussion started: 7 March 2018 Revised: 27 July 2018 – Accepted: 28 August 2018 – Published: 21 September 2018

Abstract. A number of inorganic (nitryl chloride, ClNO2; tion and loss pathways are inhibited by reduced photolysis chlorine, Cl2; and hypochlorous acid, HOCl) and chlori- rates. This results in ClNO2 making up the dominant frac- nated, oxygenated volatile organic compounds (ClOVOCs) tion (83 %) on low radiant-flux days, as its concentrations have been measured in Manchester, UK during October are still high. As most ClOVOCs appear to be formed photo- and November 2014 using time-of-flight chemical ionisa- chemically, they exhibit a similar dependence on photolysis, tion mass spectrometry (ToF-CIMS) with the I reagent ion. contributing 3 % of the Cl radical budget observed here. ClOVOCs appear to be mostly photochemical in origin, although direct emission from vehicles is also suggested. Peak concentrations of ClNO2, Cl2 and HOCl reach 506, 16 and 9 ppt respectively. The concentrations of ClNO2 are comparable to measurements made in London, but measurements 1 Introduction of ClOVOCs, Cl2 and HOCl by this method are the first reported in the UK. Maximum HOCl and Cl2 concentrations Oxidation controls the fate of many atmospheric trace gases. are found during the day and ClNO2 concentrations remain For example, increasing the oxidation state of a given species elevated into the afternoon if photolysis rates are low. Cl2 may increase its deposition velocity (Nguyen et al., 2015) exhibits a strong dependency on shortwave radiation, further or solubility (Carlton et al., 2006) and reduce its volatility adding to the growing body of evidence that it is a product of (Carlton et al., 2006), all of which act to reduce the atmo- secondary chemistry. However, night-time emission is also spheric lifetime of that species and can lead to the forma- observed. The contribution of ClNO2, Cl2 and ClOVOCs to tion of secondary material such as secondary organic aerosol the chlorine radical budget suggests that Cl2 can be a greater (SOA) or ozone (O3). As the identity of the chemical species source of Cl than ClNO2, contributing 74 % of the Cl rad- change with oxidation, intrinsic and diverse properties of icals produced on a high radiant-flux day. In contrast, on a the chemical species are altered, influencing their toxicity low radiant-flux day, this drops to 14 %, as both Cl2 produc- (Borduas et al., 2015) and their impact on the environment,

Published by Copernicus Publications on behalf of the European Geosciences Union. 13482 M. Priestley et al.: Observations of organic and inorganic chlorinated compounds e.g. cloud-particle nucleating efficiency (Ma et al., 2013) or Bannan et al., 2017). These processes may be represented by global warming potential (Boucher et al., 2009). the following equations; The hydroxyl radical (OH) is considered the most important daytime atmospheric oxidant due to its ubiquity and NO3 NO 2NO2 , (3) high reactivity, with an average tropospheric concentration + ! 6 3 NO3 NO2 N2O5(g) N2O5(aq) , (4) of 10 molecules cm (Heal et al., 1995). However, rate co- + ! ! Cl N O ClNO NO, (5) efficients for the reaction of the chlorine radical (Cl) can + 2 5 ! 2 + 3 J be 2 orders of magnitude larger than those for OH (Spicer ClNO2 ClNO ( ) ClNO (g) Cl NO . (6) et al., 1998), indicating that lower Cl concentrations of 2 aq ! 2 ! + 2 1 104 atoms cm 3 that are estimated to exist in urban areas ⇥ (e.g. Bannan et al., 2015) can be just as significant in their The anthropogenic emission of molecular chlorine is identi- contribution to oxidation. fied as another inland source of Cl in the US (e.g. Thornton Cl-initiated oxidation of volatile organic compounds et al., 2010; Riedel et al., 2012) and in China (e.g. Wang et (VOCs) forms chlorinated analogues of the OH-initiated oxi- al., 2017; Liu et al., 2017), where some of the highest concen- dation products, via addition Eq. (1) or hydrogen abstraction trations 3.0–4.7 ppb have been recorded. As well as industrial Eq. (2), forming HCl that may react with OH to regenerate processes, the suspension of road salt used to melt ice on Cl. Subsequent peroxy radicals formed through Cl oxidation roads during the winter has been suggested as a large source can take part in the HOx cycle and contribute to the enhanced of anthropogenic Cl Mielke et al., 2016). This wintertime- formation of O3 and SOA (Wang and Ruiz, 2017). This is only source, combined with reduced nitrate radical photoly- represented by the following; sis, is expected to yield greater ClNO2 concentrations at this time of the year (Mielke et al., 2016). O R X 2 R(X)OO , (1) The photolysis of molecular chlorine (Cl2) is another po- + ! · tential source of Cl. Numerous heterogeneous formation O2 RH X ROO HX, (2) mechanisms leading to Cl2 from particles containing Cl are + ! ·+ known. These include the reaction of Cl and OH (Vogt et where X is OH or Cl. al., 1996), which may originate from the photolysis of O3(aq) Nitryl chloride (ClNO2) is a major reservoir of Cl that is (Oum, 1998) or from the reactive uptake of ClNO2 (Leu et produced by aqueous reactions between particulate chloride al., 1995), ClONO2 (Deiber et al., 2004) or HOCl (Eigen and (Cl) and nitrogen pentoxide (N2O5), as seen in Eq. (4) and Kustin, 1962) to acidic Cl containing particles. Thornton et Eq. (6). Gaseous ClNO2 is produced throughout the night al. (2010) also suggest that inorganic Cl reservoirs such as and is typically photolysed at dawn before OH concentra- HOCl and ClONO2 may also enhance the Cl concentration, tions reach their peak, as in Eq. (6). This early morning re- potentially accounting for the shortfall in the global burden 1 1 lease of Cl induces oxidation earlier in the day and has been (8–22 Tg yr source from ClNO2 and 25–35 Tg yr as cal- shown to increase maximum 8 h mean O3 concentrations by culated from methane isotopes). This may be direct through up to 7 ppb under moderately elevated NOx levels (Sarwar photolysis or indirect through heterogenous reactions with et al., 2014). Typical ClNO2 concentrations measured in ur- Cl on acidic aerosol. ban regions range from 10 s of ppt to 1000 s of ppt. Mielke et Globally, Cl2 concentrations are highly variable. In the al. (2013) measured a maximum of 3.6 ppb (0.04 Hz) during marine atmosphere, concentrations of up to 35 ppt have been the summertime in Los Angeles, with maximum sunrise con- recorded (Lawler et al., 2011), whereas at urban costal sites centrations of 800 ppt. Bannan et al. (2015) measured a maxi- in the US, concentrations on the order of 100s ppt have been mum concentration of 724 ppt (1 Hz) at an urban background measured (Keene et al., 1993; Spicer et al., 1998). Sampling site in London during summer. They state that in some in- urban outflow, Riedel et al. (2012) measure a maximum of stances, ClNO2 concentrations increase after sunrise and at- 200 ppt Cl2 from plumes and mean concentrations of 10 ppt tribute this to the influx of air masses with higher ClNO2 on a ship in the LA basin. Maximum mixing ratios of up to concentrations by either advection or from the collapse of 65 ppt have also been observed in the continental US (Mielke the residual mixing layer. In urban environments where NOx et al., 2011). emission and subsequent N2O5 production is likely, Cl may More interestingly, these studies (Keene et al., 1993; be the limiting reagent in the formation of ClNO2 if excess Lawler et al., 2011; Mielke et al., 2011; Spicer et al., 1998) NO does not reduce NO3 as seen in Eq. (3) before N2O5 is report maximum Cl2 concentrations at night and minima produced (e.g. Bannan et al., 2015). Whilst the distance from during the day. However, there is a growing body of evi- a marine source of Cl may explain low, inland concentra- dence suggesting that daytime Cl2 may also be observed. Al- tions Faxon et al., 2015), the long-range transport of marine though the primary emission may be one source of daytime air can elevate inland ClNO2 concentrations (Phillips et al., Cl2 (Mielke et al., 2011), others demonstrate that the diurnal 2012) and the long-range transport of polluted plumes to a characteristics of the Cl2 time series have a broader signal marine location can also elevate ClNO2 concentrations (e.g. suggestive of continuous processes rather than intermittent

Atmos. Chem. Phys., 18, 13481–13493, 2018 www.atmos-chem-phys.net/18/13481/2018/ M. Priestley et al.: Observations of organic and inorganic chlorinated compounds 13483 signals typically associated with sampling emission sources ions was used to sample ambient air between 29 October and under turbulent conditions. 11 November 2014 at the University of Manchester’s south- In a clean marine environment Liao et al. (2014) observe ern campus, approximately 1.5 km south of Manchester city maximum Cl2 concentrations of 400 ppt attributed to emis- centre, UK (53.467 N, 2.232 W) and 55 km east of the Irish sions from a local snow pack source. A maximum was mea- Sea. The sample loss to the 1 m long 3/400 Perfluoroalkoxy sured during the morning and evening with a local mini- alkane (PFA) inlet was minimised by using a fast inlet pump mum during midday caused by photolysis. They also de- inducing a flow rate of 15 standard litres per minute (slm) scribe negligible night-time concentrations, with significant which was subsampled by the ToF-CIMS. Backgrounds were loss attributed to deposition. Faxon et al. (2015) measured taken every 6 h for 20 min by overflowing dry N2 and were Cl2 with a time-of-flight chemical ionisation mass spectrom- applied consecutively. The overflowing of dry N2 will have etry (ToF-CIMS) recording a maximum during the afternoon a small effect on the sensitivity of the instrument to those of 4.8 ppt (0.0016 Hz) and suggesting a local precursor pri- compounds whose detection is water dependent. Here we mary source of Cl2 that is potentially soil emission, with fur- find that due to the low instrumental backgrounds, the ab- ther heterogeneous chemistry producing Cl2. At a rural site solute error remains small and is an acceptable limitation in northern China, Liu et al. (2017) measured mean concen- in order to measure a vast suite of different compounds for trations of Cl2 of 100 ppt and a maximum of 450 ppt, peaking which no best practice backgrounding method has been es- during the day; they also report 480 ppt observed in an urban tablished. Whilst backgrounds were taken infrequently, they environment in the US during summer. They attribute power- are of a comparable frequency to those used in previous stud- generation facilities burning coal as the source. ies where similar species are measured (Lawler et al., 2011; Another potential source of Cl to the atmosphere is Osthoff et al., 2008; Phillips et al., 2012). The stability of the photolysis of chlorinated organic compounds (ClVOCs, the background responses, i.e. for Cl 0.16 0.07 (1) ppt, 2 ± chlorocarbons, organochlorides) that are emitted from both and the stability of the instrument diagnostics with respect to natural (biomass burning, oceanic and biogenic emission) the measured species suggest that they effectively capture the (e.g. Yokouchi et al., 2000) and anthropogenic sources true instrumental background. (e.g. Butler, 2000). Whilst many ClVOCs are only consid- Formic acid was calibrated throughout the campaign and ered chemically important in the stratosphere, those that post campaign. Very little deviation in the formic acid cal- are photochemically labile in the troposphere, e.g. methyl ibrations was observed. The mean average sensitivity was hypochlorite (CH OCl), whose absorption cross section is 30.66 1.90 (1) Hz ppt 1. A number of chlorinated species 3 ± non-negligible at wavelengths as long as 460 nm (Crowley et were calibrated post campaign using a variety of different al., 1994), can act as a source of Cl and take part in oxidative methods, and relative calibration factors were applied based chemistry. on measured instrument sensitivity to formic acid as has been Photolysis of ClVOCs have been postulated to contribute performed previously (e.g. Le Breton et al., 2014a, 2017; 0.1 0.5 103 atoms cm 3 globally to the Cl budget of the Bannan et al., 2015). A summary of calibration procedures ⇥ boundary layer (Hossaini et al., 2016), although on much and species calibrated are described below. All data from be- smaller spatial and temporal scales, the variance in this esti- tween 16:30 LT on 5 November and 00:00 LT on 7 November mate is likely to be large. Very few data exist on the concen- have been removed to prevent the interference of a large- trations, sources and spatial extent of oxygenated ClVOCs scale anthropogenic biomass burning event (Guy Fawkes (ClOVOCs) and their contribution to the Cl budget. Night) on these analyses. The ToF-CIMS is a highly selective and sensitive instrument with high mass accuracy and a resolution (m/dm 2.1 Calibrations 4000) that is capable of detecting a suite of chlorinated ⇠ compounds, including HOCl and organic chlorine (Le Bre- We calibrate a number of species by overflowing the inlet ton et al., 2018) as well as other oxygenated chlorine species with various known concentrations of gas mixtures (Le Bre- and chloroamines (Wong et al., 2017). Here we use the ToF- ton et al., 2012), including molecular chlorine (Cl2, 99.5 % CIMS with the I reagent ion to characterise the sources of purity, Aldrich), formic acid (98/100 %, Fisher) and acetic chlorine and estimate their contribution to Cl concentrations acid (glacial, Fisher) by making known mixtures (in N2) and in the wintertime in Manchester, UK. flowing 0–20 standard cubic centimetres per minute (sccm) into a 3 slm N2 dilution flow that is subsampled. The Cl2 1 calibration factor is 4.6 Hz ppt . 2 Methodology/experiment As all chlorinated VOCs we observe are oxygenated we assume the same sensitivity found for 3-chloropropionic acid 1 Full experimental details and a description of meteorologi- (10.32 Hz ppt ) for the rest of the organic chlorine species cal and air quality measurements can be found in Priestley detected. Chloropropionic acid (Aldrich) was calibrated fol- et al. (2018). A time-of-flight chemical ionisation mass spec- lowing the methodology of Lee et al. (2014). A known trometer (ToF-CIMS) (Lee et al., 2014b) using iodide reagent quantity of chloropropionic acid was dissolved in methanol www.atmos-chem-phys.net/18/13481/2018/ Atmos. Chem. Phys., 18, 13481–13493, 2018 13484 M. Priestley et al.: Observations of organic and inorganic chlorinated compounds

(Aldrich) and a known volume was doped onto a filter. The CIMS is able to quantify the concentration of ClNO2 gen- filter was slowly heated to 200 C to ensure the total desorp- erated in the flow tube as the equivalent drop in NO2 sig- tion of the calibrant whilst 3 slm N2 flowed over it. This was nal. This indirect measurement of ClNO2 is similar in its repeated several times. A blank filter was first used to deter- methodology to ClNO2 calibration by quantifying the loss mine the background. of N2O5 reacted with Cl (e.g. Kercher et al., 2009). We do ClNO2 was calibrated by the method described by Kercher not detect an increase in I.Cl2 signal from this calibration et al. (2009) with N2O5 synthesised following the method- and so rule out the formation of Cl2 from inorganic species ology described by Le Breton et al. (2014)a, giving a cal- in our inlet due to unknown chemistry occurring in the IMR. 1 ibration factor of 4.6 Hz ppt . Excess O3 is generated by The TF-CIMS method gives a calibration factor 58 % greater flowing 200 sccm O2 (BOC) through an ozone generator than that of the N2O5 synthesis method. The Cl atom titra- (BMT, 802N) and into a 5 L glass volume containing NO2 (, tion method assumes a 100 % conversion to ClNO2 and does > 99.5 %). The outflow from this reaction vessel is cooled in not take into account any Cl atom loss, which will lead to a cold trap held at 78 C (195 K) by a dry ice and glycerol a reduced ClNO concentration and thus a greater calibra- 2 mixture where N2O5 is condensed and frozen. The trap is al- tion factor. Also, the method assumes a 100 % sampling ef- lowed to reach room temperature and the flow is reversed, ficiency between the TF-CIMS and ToF-CIMS; again this where it is then condensed in a second trap held at 195 K. could possibly lead to an increased calibration factor. Whilst This process is repeated several times to purify the mixture. the new method of calibration is promising, we assume that The system is first purged by flowing O3 for 10 min before the proven method developed by Kercher et al. (2009) is the use. To ascertain the N2O5 concentration on the line, the flow correct calibration factor and assign an error of 50 % to that is diverted through heated line to decompose the N2O5 and calibration factor. We feel that the difference between the two into to a Thermo Scientific 42i NOx analyser, where it is de- methods is taken into account by our measurement uncer- tected as NO2. It is known that the Thermo Scientific 42i tainty. NOx analyser suffers from interferences from NOy species, We calibrate HOCl using the methodology described indicating that this method could cause an underestimation by Foster et al. (1999) giving a calibration factor of 1 of the ClNO2 concentrations reported here. Based on previ- 9.22 Hz ppt . 100 sccm N2 is flowed through a fritted bub- ous studies (e.g. Le Breton et al., 2014; Bannan et al., 2017) bler filled with NaOCl solution (min 8 % chlorine, Fisher) where comparisons with a broad-beam cavity-enhancement that meets a dry 1.5 slm N2 flow, with the remaining flow absorption spectrometer (BBCEAS) have been made, good made up of humidified ambient air, generating the HOCl and agreement has been found between co-located N2O5 mea- Cl2 signal measured on the ToF-CIMS. The flow from the surements. We feel that this calibration method works well, bubbler is diverted through a condensed HCl () scrubber likely in part due to the high purity of the N2O5 synthe- (condensed HCl on the wall of 20 cm PFA tubing) where sised and because the possible interference of NOy on the HOCl is titrated to form Cl2. The increase in Cl2 concen- NOx analyser during this calibration is considered negligible. tration when the flow is sent through the scrubber is equal to The N2O5 is passed over a salt slurry where excess chloride the loss of HOCl signal and as the calibration factor for Cl2 may react to produce ClNO2. The drop in the N2O5 signal is is known, the relative calibration factor for HOCl to Cl2 is equated to the rise in ClNO2, as the stoichiometry of the re- found. action is 1 1. The conversion efficiency of N O to ClNO Additionally, several atmospherically relevant ClVOCs : 2 5 2 over wet NaCl is known to vary by 60 %–100 % (Hoffman et were sampled in the laboratory to assess their detectabil- al., 2003; Roberts et al., 2008). Here we follow the method- ity by the ToF-CIMS with I. The instrument was able to ology of Osthoff et al. (2008) and Kercher et al. (2009) that detect dichloromethane (DCM, VWR), chloroform (CHCl3, ensures that conversion is 100 % efficient, so we assume a 99.8 %, Aldrich) and methyl chloride (CH3Cl, synthesised), 100 % yield in this study. although the instrument response was poor. The response to We developed a secondary novel method to quantify 3-chloropropionic acid was orders of magnitude greater than ClNO2 by cross-calibration with a turbulent flow tube chem- for the ClVOCs suggesting that the role of the chlorine atom ical ionisation mass spectrometer (TF-CIMS) (Leather et is negligible compared with the carboxylic acid group in de- al., 2012). Chlorine atoms were produced by combining a termining the I sensitivity in this case. 2.0 slm flow of He with a 0–20 sccm flow of 1 % Cl2, which was then passed through a microwave discharge produced by 2.2 Cl radical budget calculations a surfatron (Sairem) cavity operating at 100 W. The Cl atoms were titrated via a constant flow of 20 sccm NO2 (99.5 % pu- Within this system, we designate ClNO2, HOCl and organic rity NO2 cylinder, Aldrich) from a diluted (in N2) gas mix to chlorine as sources of Cl. As HCl was not detected, it is which the TF-CIMS has been calibrated. This flow is car- not possible to quantify the contribution of Cl from the re- ried in 52 slm N that is purified by flowing through two action of HCl OH. Loss processes of Cl are Cl O and 2 + + 3 heated molecular sieve traps. This flow is subsampled by the Cl CH (7). Photolysis rates for the Cl sources are taken + 4 ToF-CIMS where the I.ClNO2 adduct is measured. The TF- from the US National Center for Atmospheric Research

Atmos. Chem. Phys., 18, 13481–13493, 2018 www.atmos-chem-phys.net/18/13481/2018/ M. Priestley et al.: Observations of organic and inorganic chlorinated compounds 13485

(NCAR) Tropospheric Ultraviolet and Visible TUV radia- 3 Results tion model (Mandronich, 1987) assuming a 100 % quantum yield at our latitude and longitude with a column overhead Concentrations of all chlorinated species are higher at the O3 measured by the Brewer spectrophotometer #172 (Smed- beginning of the measurement campaign, when air masses ley et al., 2012) and assuming zero optical depth. To account originating from continental Europe were sampled (Reyes- for the effective optical depth of the atmosphere, includ- Villegas et al., 2018). Toward the end of the measurement ing clouds and other optical components, we scale our ide- campaign, ClNO2 and ClOVOCs concentrations were low, alised photolysis rate coefficient (J ) by the observed trans- which is consistent with the pollution during this period hav- mittance values in the UV-A waveband (325 to 400 nm). ing a high fraction of primary components (Reyes-Villegas These transmittance values are calculated from UV spectral et al., 2018), see Fig. 1. scans of global irradiance, measured at half-hourly intervals by the Brewer spectrophotometer and provided as an output 3.1 Inorganic chlorine of the SHICrivm analysis routine (Slaper et al., 1995). The Cl rate coefficient for the reaction with O3 is kCl O3 1.20 11 3 1 1 + = ⇥ We detect a range of inorganic chlorine species and frag- 10 cm molecule s (Atkinson et al., 2006b) and CH4 13 3 1 1 ments including I.Cl, I.ClO, I.HOCl, I.Cl2, I.ClNO2 is kCl CH4 1.03 10 cm molecule s (Atkinson et + = ⇥ and I.ClONO, however we do not detect I.ClO, I.Cl2O, al., 2006a). The individual kCl VOC are taken from the NIST 2 2 + I.Cl O , I.ClNO or I.HCl . Laboratory studies have shown chemical kinetics database. This is represented by the fol- 2 2 lowing equation; that the ToF-CIMS is sensitive to detection of I.HCl, however under this configuration, the I.HCl adduct was not observed. The statistics of the concentrations reported below do not take into account the limits of detection (LOD), so for [Cl] SS = some of the measurements, values may be reported below the 2J [Cl ] J [ClNO ] Cl2 2 + ClNO2 2 + LOD. JHOCl[HOCl] JClOVOC6 ClVOCs + [n ] (7) kO3 Cl [O3] kCH4 Cl [CH4] i kCl VOCi VOC i. + + + + + [ ] 3.1.1 ClNO2 P

As methane was not measured, an average concentration was ClNO2 (m/z 208) was detected every night of the campaign taken from the European Centre for Medium-Range Weather with an LOD (3 standard deviation of the background) of ⇥ Forecasts (ECMWF) Copernicus atmosphere monitoring ser- 3.8 ppt. The 1 Hz mean night-time concentration of ClNO2 vice (CAMS). VOC concentrations were approximated by was 58 ppt (not accounting for the LOD), and a maximum applying representative VOC : benzene ratios for the UK ur- of 506 ppt (not accounting for the LOD) was measured as ban environment (Derwent et al., 2000) and applying those a large spike on the evening of 30 October. These concen- to a typical urban UK benzene : CO ratio (Derwent et al., trations are comparable to other urban UK measured values, 1995), where CO was measured at the Whitworth observa- although the maximum concentration reported here is 30 % tory. The VOC : benzene ratios are scaled to the year of this lower than that measured in London (Bannan et al., 2015) study to best approximate ambient levels (Derwent et al., but is consistent with high concentrations expected during 2014). The calculated benzene : CO ratio is in good agree- the winter, as discussed in the introduction. ment with a Non-Automatic Hydrocarbon Network monitor- The diurnal profile of ClNO2 increases through the ing site (Manchester Piccadilly) approximately 1.5 km from evening to a local morning maximum, with rapid loss after the measurement location, indicating that the approximation sunrise. Although we observe a rapid build-up after sunset made here is reasonably accurate. The ratios assume that traf- (ca. 16:30 LT) and loss after sunrise (ca. 07:30 LT), the maxi- fic emissions are the dominant source of the VOCs, as is as- mum concentration measured within a given 24 h period typi- sumed here. cally peaks at around 22:00 LT and halves by 03:00 LT, where The photosensitivity of the ClOVOCs to wavelengths it is maintained. The reasons for the early onset in peak con- longer than 280 nm dictates their ability to contribute to centration and loss throughout the night is unclear, although the Cl budget in the troposphere. As many of the identified on 1 November, a sharp decrease in ClNO2 is a consequence species here do not have known photolysis rates, we approx- of a change in wind direction, indicating that the source of imate the photolysis of methyl hypochlorite JCH3OCl for all ClNO2 is directional. A minimum concentration of < LOD is ClOVOCs, as it is the only available photolysis rate for an reached by 15:00 LT, indicating that concentrations can per- oxygenated organic compound containing a chlorine atom sist for much of the day. On 7 November ClNO2 concen- provided by the TUV model and no other more suitable pho- trations grow throughout the morning, even after photolysis tolysis rate could be found elsewhere, e.g. the JPL kinetics begins, until 11:00 LT. Correlated high wind speeds suggest database. The same quantum yield and actinic flux assump- that the long-range transport and downward mixing is a likely tions are made. cause for this daytime increase. www.atmos-chem-phys.net/18/13481/2018/ Atmos. Chem. Phys., 18, 13481–13493, 2018 13486 M. Priestley et al.: Observations of organic and inorganic chlorinated compounds

2 Figure 1. Time series of (a) ClNO2 (ppt), (b) HOCl (ppt) and O3 (ppb), and (c) Cl2 (ppt) and direct solar radiation (Wm ). (d) NO (ppb) and NO2 (ppb). (e) Relative humidity (%). Data is removed during bonﬁre night (5–6) and HOCl data is discounted thereafter due to a persistent interference that was not present earlier.

Typically, elevated concentrations of ClNO2 are measured similarly to Cl2, but remain elevated for longer, dropping af- when the wind direction is easterly and wind speeds are low ter sunset. The diurnal profile is similar to that for O3, with 1 (2–4 ms ), also during periods of southerly winds between a maximum during the day and minima during morning and 1 3–9 ms . The potential sources of Cl precursor from these evening rush hours when NOx is emitted locally. The strong directions are industrial sites, including waste water treat- correlation with O (R2 0.67) is expected, as the route to 3 = ment facilities (8.5 km east and 7.0 km south) that may use the formation of HOCl is the oxidation of Cl with O3 to form salt water as part of the chemical disinfection process (Gher- ClO and then the oxidation by HO2 to form HOCl. Non- naout and Ghernaout, 2010). Another source of the ClNO2 negligible night-time concentrations of a maximum 8.1 ppt precursor is found from the southwest at wind speeds of are only measured when concentrations of other inorganic Cl 1 9 ms , indicating a more distant source that is also likely containing species are high. The HOCl signal is artificially to be industrial or marine. The correlation between ClNO2 elevated after the night of the 5 due to a persistent interfer- and Cl2 is poor at most times, apart from the night of the 30 ence from a large-scale biomass burning event (Guy Fawkes where a strong linear relationship is observed. This is consis- Night, Priestley et al., 2018), which cannot be de-convolved tent with polluted continental air masses advecting a variety from the dataset due to the small difference in their mass- of trace gases. Throughout the measurement campaign the to-charge ratios and insufficient instrument resolution. For relationship between ClNO2 and Cl2 is poor, so it is unlikely this reason HOCl data after this date are discounted from the they share the same source. analysis.

3.1.2 HOCl 3.1.3 ClO

HOCl concentrations average 2.18 ppt (not accounting for We detect the I.ClO adduct at m/z 178, which strongly cor- the LOD) and reach a daytime maximum of 9.28 ppt with an relates with I.ClNO2 and I.Cl signals, all of which show LOD of 3.8 ppt. Concentrations peak in the early afternoon, night-time maxima. This is inconsistent with the ClO photo-

Atmos. Chem. Phys., 18, 13481–13493, 2018 www.atmos-chem-phys.net/18/13481/2018/ M. Priestley et al.: Observations of organic and inorganic chlorinated compounds 13487 chemical production pathway of Cl O , suggesting that its C H O exhibits a similar diurnal profile and radiation de- + 3 2 4 5 maximum concentration should be measured during the day, pendency (Fig. 2). Also, the production of O3 increases and as was observed for HOCl. It is not possible to confirm if the decreases with direct solar radiation at the same times we ob- I.ClO is a fragment of a larger ClO containing molecule, serve the enhancements in concentrations of Cl2 and C2H4O5 however, as the fragmentation of multiple larger molecules (Fig. 2). The changes in O3 production are observed when are detected as a single adduct, e.g. the I.Cl cluster is a NO concentrations are near zero, indicating that O3 produc- known fragment from ClNO2 and HOCl, it is reasonable to tion is VOC limited. Finally, other large organic molecules suspect that I.ClO may be a fragment as well. e.g. C10H14O4 do not exhibit this strong coupling with direct solar radiation. This evidence suggests that a local photolytic 3.1.4 Cl2 daytime mechanism is responsible for the increase in daytime concentrations as has previously been suggested (e.g. We observe concentrations of Cl2 during the day ranging Finley and Saltzman, 2006). from 0–16.6 ppt with a mean value of 2.3 ppt (not account- Although peak concentrations of Cl2 are observed in the ing for the LOD) and night-time concentrations of 0–4.7 ppt daytime, high levels of Cl2 are also observed during the with mean concentrations of 0.4 ppt (not accounting for the night. At the beginning of the measurement period, which LOD), see Fig. 1. The LOD is 0.5 ppt and a calibration fac- has previously been characterised using an aerosol mass 1 tor of 4.5 Hz ppt was found. These concentrations are of spectrometer (AMS) as a period of high secondary activity the same order of magnitude as measured at an urban site (Reyes-Villegas et al., 2018), there are persistent, non-zero in the US but up to 2 orders of magnitude smaller than at concentrations of Cl ( 4 ppt) after sunset. On 4 Novem- 2  US urban costal sites (Keene et al., 1993; Spicer et al., 1998) ber, after the period of high secondary activity, intermittent and a megacity impacted rural site in northern China (Liu et elevations in night-time Cl2 concentrations when the wind is al., 2017). Although the maximum measured value here is northerly suggest that a local emission source, with concen- an order of magnitude greater than that measured in Hous- trations reaching a maximum of 4.6 ppt. Two more distinct ton (Faxon et al., 2015), the photolysis rate of Cl2 here is 2 night-time sources, ranging from the south west through to orders of magnitude smaller compared with Houston at that the east of the measurement site, indicate a likely origin in time. industrial areas, some of which contain chemical production The diurnal profile of Cl2 exhibits a maximum at midday and water treatment facilities. and a minimum at night (early morning) consistent with other studies (Faxon et al., 2015; Liao et al., 2014; Liu et al., 2017). 3.2 Organic chlorine The days with the greatest concentration are those where direct shortwave radiation is at its highest. On 5 Novem- We detected seven C C ClOVOCs of the 2 6 ber, the incidence of direct shortwave radiation is unhin- forms CnH2n 1O1Cl, CnH2n 1O2Cl, CnH2n 1O3Cl, + + + dered throughout the day and a similarly uniform profile for CnH2n 1O2Cl, CnH2n 1O3Cl and CnH2n 3O2Cl (Fig. 3), Cl2 is also observed. On 1 November, Cl2 concentrations of which only C2H3O2Cl has been previously reported (Le increase unhindered as direct radiation increases but when Breton et al., 2018). We find no evidence for the detection of cloud cover reduces radiation transmission efficiency, a cor- small chlorohydrocarbons, e.g. poly-chloromethanes, such responding drop in Cl2 is also observed (Fig. 2). Also, when as methyl chloride, dimethyl chloride and chloroform, or global radiation is low throughout the day, e.g. 7 November, poly-chloroethanes such as those described by Huang et we observe very low concentrations of Cl2. al. (2014) in the ambient data, but qualitative testing and There is the potential that the Cl2 signal detected is an laboratory calibrations show that the iodide reagent ion can instrumental artefact generated either by chemistry in the detect CH Cl (not calibrated), CH Cl (LOD 143 ppb) 3 2 2 = IMR or from displacement reactions or degassing on the in- and CHCl (LOD 11 ppb). We find no discernible evi- 3 = let walls. We believe none of these to be the case. First, the dence for the detection of 4-chlorocrotonaldehyde, the Cl correlation between the signal used for labile chlorine in the oxidation product of 1,3-butadiene and unique marker of IMR 35Cl (m/z 35) is high with ClNO (R2 0.98) yet is chlorine chemistry (Wang and Finlayson-Pitts, 2001) due 2 = non-existent with Cl (R2 0.01) indicating Cl concentra- to interferences from other CHO compounds. We do not 2 = 2 tion is independent of 35Cl concentrations. Second, there is believe that these species are products of inlet reactions no correlation between HNO and Cl (R2 0.07) which as there is a poor correlation (R2 0.039) with labile 3 2 = = suggests that acid displacement reactions are not occurring chlorine 35Cl. on the inlet walls. Third, there is no correlation between tem- The maximum hourly averaged total ClOVOCs con- perature and Cl (R2 0.08), indicating that localised am- centration is 28 ppt at 12:00 LT and at a minimum of 2 = bient inlet heating is also not a contributing factor to in- 5 ppt at 07:00 LT, when NOx concentrations are highest at creased Cl concentrations. Fourth, we observe a similar di- 30 ppb. Concentrations of C H O Cl (tentatively identi- 2 ⇠ 2 3 2 rect radiation dependency for other photochemical species as fied as chloroacetic acid) and C6H13OCl (tentatively identi- we observe for Cl2. For example, the temporal behaviour of fied as chloro-hexanol) are the highest of any ClOVOCs, ac- www.atmos-chem-phys.net/18/13481/2018/ Atmos. Chem. Phys., 18, 13481–13493, 2018 13488 M. Priestley et al.: Observations of organic and inorganic chlorinated compounds

Figure 2. Time series for 1 November 2014, with (a) Cl2, (b) solar radiation (global, direct and indirect), (c) photochemical marker C2H4O5, 1 O3 and (d) O3 and NOx, where highlighted boxes demonstrate that 1[ t ] is increasing. The increase in concentration of Cl2,C2H4O5 and O3 production when VOC limited are strongly coupled with direct solar radiation. Greyed areas are night time.

C2H3O2Cl, they do not enhance as much as those photochemical species or return to nominal levels after the solar maximum. Instead, they increase again during the night, with C3H5O3Cl reaching a maximum concentration of 8 ppt at 20:00 LT. This trend suggests that concentration changes could be a function of boundary layer height. C3H7O2Cl and C4H7O2Cl (yellow in Fig. 3) are the only ClOVOCs that show a positive correlation with NOx (R2 0.42, R2 0.41) and negative correlation with O = = 3 Figure 3. Diurnal profiles of Cl VOCs. (a) Stacked plot showing (R2 0.58, R2 0.53). Their correlation is stronger = = total Cl VOC concentration. (b) The first data point of each diurnal with NO (R2 0.55, R2 0.48), a product of traffic 2 = = trace is mean normalised to 1.0. Reds show photochemical dom- emission. This suggests that at least some of the time, inated signals with maxima at midday, whereas yellow and blue they accumulate at low wind speeds, indicating their ori- traces show a more typical diurnal concentration profile associated with changes in boundary layer height, indicating that these species gins as local, primary emissions or as thermal degrada- have longer lifetimes. tion products that have a traffic source, e.g. polychlorinated dibenzo p dioxins/dibenzofurans (PCDD/F) and their oxidation products (Fuentes et al., 2007; Heeb et al., 2013). The diurnal profile shows maxima during midday consistent with counting for between 20 % and 30 % of total ClOVOCs con- other photochemical species, which is expected of secondary centrations measured. All concentrations rise towards mid- formation. It is possible that these compounds are isobaric day, with C3H7O2Cl and C2H3O2Cl rising the most by a or isomeric with the other compounds that interfere with the factor of 4 and returning to nominal levels by the early perceived signals recorded here. evening (red in Fig. 3). C3H7O2Cl and C2H3O2Cl correlate The diurnal profile of C3H5O3Cl (green in Fig. 3) exhibits 2 well with Cl2 (R 0.77 and 0.75, respectively), which is con- a similar shape to the bimodal distribution observed for NOx. sistent with a photochemical formation mechanism identify- Cross-correlation indicates that a time lag of 3 h provides ing these species as secondary products, potentially chloro- the best correlation with NO of R2 0.80. This suggests 2 = propanediol and chloro-acetic acid. that local oxidation chemistry, which takes place over long Whilst the diurnal profiles of C6H13OCl and C5H7O2Cl (blue in Fig. 3) are similar to those of C3H7O2Cl and

Atmos. Chem. Phys., 18, 13481–13493, 2018 www.atmos-chem-phys.net/18/13481/2018/ M. Priestley et al.: Observations of organic and inorganic chlorinated compounds 13489

Figure 5. Diurnal proﬁle for 1 November of (a) idealised JCl2 and Figure 4. Transmission scaled J values for Cl2, ClNO2 and HOCl P (Cl), (b) scaled JCl2 and P (Cl), and (c) the difference between for 1, 5 and 7 November, where 1 November had high photoly- (a) and (b). Transmission efﬁciency scaled photolysis reduce P (Cl) sis rates in the morning that were reduced during the afternoon, from Cl2 photolysis. 5 November is the closest to a full day’s ideal photolysis and 7 November shows very weak photolysis.

ClNO2 are much greater (Fig. 6). During the morning of 5 November, ClNO is the dominant source of Cl, con- periods in the day and is sensitive to traffic emission, is the 2 tributing 95 % of the total Cl concentration, a maximum of source of this ClOVOC. 3.0 103 Cl radicals cm 3, to the steady-state concentration, ⇥ which is approximately a factor of 3 lower than the estimated maximum concentration of 9.5 103 Cl radicals cm3 4 Discussion ⇥ produced by ClNO2 photolysis in London during the sum- 4.1 Effect of global radiation transmission efficiency on mer (Bannan et al., 2015) and a factor of 22 lower than the maximum concentration of 85.0 103 Cl radicals cm 3 cal- Cl radical production ⇥ culated from measurements of ClNO2 in Houston (Faxon et Three days are selected based on their different solar short al., 2015). In both instances, this is due to a combination of wave transmission efficiencies to quantify the variation in lower JClNO2 and lower ClNO2 concentrations. Cl2 formation and photolysis and thus the influence of Cl2 As the day progresses, concentrations of Cl2 increase and on producing Cl. The average transmission of global radia- it becomes the dominant and more sustained source of Cl by contributing 95 % of Cl (12.5 103 Cl radicals cm 3) tion on 5 November was high with 84 14 % (1), whereas ⇥ ± by the early afternoon, which is approximately 4 that of on 7 November it was very low with 21 14 %, sometimes ⇥ ± the ClNO measured in the early morning and 1.3 higher dropping below 10 % in the middle of the day. The middle 2 ⇥ case is represented by 1 November, where the transmission than the maximum estimated concentration calculated from efficiency in the morning was high with 88 11 %, but in the the ClNO2 photolysis in London (Bannan et al., 2015). The ± afternoon it was highly variable and dropped to 55 20 % maximum Cl concentration produced from Cl2 and ClNO2 ± photolysis on 5 November reached 14.2 103 Cl radicals cm3 (see Fig. 4). These 3 days provide good case studies for the ⇥ investigation of the effect of global radiation on molecular at 11:30 LT which is approximately 16 % of the 85.0 3 3 ⇥ chlorine concentrations and therefore the production of Cl. 10 Cl,radicals cm maximum calculated value from the pho- The reduced transmission efficiency inhibits Cl2 forma- tolysis of these two species in Houston in summer (Faxon et tion, thereby reducing the contribution of Cl2 to Cl pro- al., 2015). This is dominated by the contribution of Cl2, in- duction. The lower transmission efficiency also reduces the dicating that Cl2 can be a much more significant source of photolysis of Cl2 and so reduces the production of Cl Cl than ClNO2. On this high-flux day, when hourly mean even further. Figure 5 shows the divergence between the Cl2 concentrations range between 0–7 ppt, the source term is 1 calculated between 4–21 ppt Cl2 h , which is slightly lower, ideal JCl2 without transmission efficiency correction (a), the although consistent with previous studies (Faxon et al., 2015; JCl2 value scaled by transmission efficiency (b) and subsequent Cl formation. Cl production rates are similar un- Finley and Saltzman, 2006; Spicer et al., 1998). til 11:00 LT, when the scaled production then becomes an A day with low photolysis rates and high daytime ClNO2 average 47 % lower. This is most prominent at 13:00 LT, concentrations has been highlighted as 7 November. On this when the difference between ideal and scaled production is day, ClNO2 is the dominant Cl source (95 %) reaching a 4 3 1 maximum of 3.4 103 Cl radicals cm 3 at 09:30 LT, which is 8.4 10 Cl radicals cm s . ⇥ ⇥ 87 % of that calculated for London (Bannan et al., 2015). ⇠ 4.2 Contribution of inorganic chlorine to Cl radical A mean Cl2 concentration of 0.3 ppt (less than the LOD of production 0.5 ppt) on this day is very low, as production of Cl2 at its 1 maximum, calculated as 0.6 ppt h , is also low. This com- The contribution of HOCl and ClOVOCs to Cl forma- bined with a low maximum of J 1.13 10 4 h 1 means Cl2 = ⇥ tion is negligible due to low photolysis rates and low that maximum Cl production from Cl2 photolysis on this day concentrations, whereas the contributions from Cl and is very low, generating 0.9 103 Cl radicals cm 3 at 10:00 LT 2 ⇥ www.atmos-chem-phys.net/18/13481/2018/ Atmos. Chem. Phys., 18, 13481–13493, 2018 13490 M. Priestley et al.: Observations of organic and inorganic chlorinated compounds

5 Conclusions

A large suite of inorganic and organic, oxygenated, chlorinated compounds has been identified in ambient, urban air during the wintertime in the UK. Of the seven organic chlorinated compounds (ClOVOCs) identified here, only C2H3O2ClO (tentatively assigned as chloroacetic acid) has previously been reported. Although the ToF-CIMS with I Figure 6. Steady state concentration of Cl from ClNO2, Cl2, HOCl and total ClOVOC photolysis for (a) 1 November, (b) 5 November, is sensitive towards chlorinated and polychlorinated aliphatic and (c) 7 November. The importance of ClNO2 during the morning compounds, e.g. methyl chloride (CH3Cl) dimethyl chlo- is most evident on the 5, with a diminishing contribution throughout ride (CH2Cl2) and chloroform (CHCl3), their concentrations the day. On the high-flux days, Cl2 is the most important source of were below the detection limit. The sources of ClOVOCs are Cl, but on the low-flux day, ClNO2 is most important. mostly photochemical with maxima of up to 28 ppt observed at midday, although C3H7O3Cl and C4H7O2Cl concentrations correlate with NOx accumulating at low wind speeds, indicating they are produced locally, potentially as the ther- or a quarter of the maximum contributed by ClNO2 on this mal breakdown products of higher-mass chlorinated species day (see Fig. 6). This is represented by the following equa- such as polychlorinated dibenzo p dioxins/dibenzofurans tion; (PCDD/F) from car exhausts or the oxidation products thereof. C3H5O3Cl shows a good diurnal cross-correlation with NO with a time lag of 3 h, suggesting that its produc- J 2 J 1 Cl2 Cl Cl2(g) Cl. (8) tion is sensitive to NOx concentrations on that time scale. (aq) ! 2 ! Alongside ClOVOCs, daytime concentrations of Cl2 and ClNO2 are measured, reaching maxima of 17 and 506 ppt, respectively. ClNO is a source of Cl in every daytime pe- The dependency of Cl formation on Cl2 production and loss 2 highlights the sensitivity of this reaction channel to the pho- riod measured. Cl2 shows strong evidence of a daytime pro- tolysis that is demonstrated on these 2 days. The production duction pathway limited by photolysis as well as emission sources evident during the evening and night time. of Cl from ClNO2 is less sensitive, relatively speaking, to the On a day of high radiant flux (84 14 % of ide- solar flux, as the production of ClNO2 does not rely on pho- ± tochemistry but chemical composition cf. Eqs. (6) and (8). alised values), Cl2 is the dominant source of Cl, gen- This further highlights the role of photolytic mechanisms in erating a maximum steady state concentration of 12.5 3 3 ⇥ the re-activation of particulate chloride to gaseous chlorine 10 Cl radicals cm or 74 % of the total Cl produced by the radicals. photolysis of Cl2, ClNO2, HOCl and ClOVOC, with the latter three contributing 19 %, 4 % and 3 %, respectively. This contrasts with a share of 14 % for Cl2, 83 % for ClNO2 and 4.3 Organic vs. inorganic contribution to Cl radical 3 % for ClOVOCs on a low radiant-flux day (21 14 % of ± production idealised values). On the low radiance day, not only is the photolysis of all Cl species inhibited, reducing Cl concentra- Summing the concentrations of the ClOVOCs and assum- tions, but also the formation of Cl2 and some ClOVOCs by ing a uniform photolysis rate JCH3OCl as detailed in the photochemical mechanisms is inhibited, thus the variability above section, we derive the contribution of total measured in contribution between days is highly sensitive to the inci- ClOVOC to the Cl budget and compare it to the contribution dence of sunlight. This further highlights the importance of from inorganic Cl measured here (Fig. 6). On the high-flux photochemistry in the re-activation of particulate chloride to day, the Cl concentration reaches 4.0 102 Cl radicals cm 3 gaseous chlorine radicals. Similarly to Cl , ClOVOCs can be ⇥ 2 at midday, which is 30 % of the contribution by ClNO2, 3.6 % an important source of Cl, although the behaviour of their of the contribution from Cl2 and 2 % of the HOCl contribu- contribution is similar to Cl2, relying on high rates of photol- tion for the same day. On the low-flux day, the ClOVOC con- ysis rather than high concentrations as is the case for ClNO2. tribution is 11.0 102 Cl radicals cm 3, which is 2.8 % of The contribution of the ClOVOCs to the Cl budget would ⇥ ⇠ the ClNO contribution on that day and 57 % of the Cl be better determined if more specific photolysis rates for each 2 ⇠ 2 contribution. Like Cl2, the production of most ClOVOC re- compound were available and so would further improve the quires a photolytic step to generate concentrations that can accuracy of the contribution they make to the Cl budget. In then go on to decompose, providing the Cl. Here it is sug- addition, future work should aim to identify the processes gested that the organic contribution to Cl production is neg- leading to the formation of these compounds to better con- ligible at 15 % on the low radiant-flux day and 3 % on the strain the Cl budget in the urban atmosphere. Further ambi- high-flux day. ent measurements of a broader suite of chlorinated species as

Atmos. Chem. Phys., 18, 13481–13493, 2018 www.atmos-chem-phys.net/18/13481/2018/ M. Priestley et al.: Observations of organic and inorganic chlorinated compounds 13491 shown here in different chemical environments would help to C. J.: Ground and Airborne U.K. Measurements of Nitryl Chlo- better constrain the contribution that chlorine-initiated chem- ride: An Investigation of the Role of Cl Atom Oxidation at Wey- istry has on a global scale. bourne Atmospheric Observatory, J. Geophys. Res.-Atmos., 122, 11154–11165, https://doi.org/10.1002/2017JD026624, 2017. Borduas, N., Da Silva, G., Murphy, J. G., and Abbatt, J. P. D.: Data availability. The data are to be uploaded to http://catalogue. Experimental and theoretical understanding of the gas phase ceda.ac.uk/ (last access: 20 September 2018). oxidation of atmospheric amides with OH radicals: Kinetics, products, and mechanisms, J. Phys. Chem. A, 119, 4298–4308, https://doi.org/10.1021/jp503759f, 2015. Boucher, O., Friedlingstein, P., Collins, B., and Shine, K. P.: Author contributions. MP, MlB, TJB and AB operated the CIMS The indirect global warming potential and global temperature during for the measurement period. Data analysis was performed change potential due to methane oxidation, Environ. Res. Lett., by MP, MIB and TJB. ERV and JA aided with data analysis. The 4, 044007, https://doi.org/10.1088/1748-9326/4/4/044007, 2009. manuscript was written by MP. MP, SDW and AM performed labo- Butler, J.: Better budgets for methyl halides?, Nature, 403, 260–261, ratory calibrations. ARDS and AW provided transmittance data and https://doi.org/10.1038/35002232, 2000. contributed to that section of the manuscript. DES and CJP provided Carlton, A. G., Turpin, B. J., Lim, H. J., Altieri, K. E., and assistance with the steady-state calculation. DES, CJP and HC sup- Seitzinger, S.: Link between isoprene and secondary organic ported with project management and manuscript refinement. aerosol (SOA): Pyruvic acid oxidation yields low volatility organic acids in clouds, Geophys. Res. Lett., 33, 2–5, https://doi.org/10.1029/2005GL025374, 2006. Competing interests. The authors declare that they have no conflict Crowley, N., Helleis, F., Miiller, R., Moortgat, G. K., Crutzen, P. of interest. J., and Orlando, J. J.: CH3OCI: UV/visible absorption cross sections, J values and atmospheric significance, J. Geophys. Res., 99, 20683–20688, 1994. Acknowledgements. This work was conducted during a PhD Deiber, G., George, Ch., Le Calvé, S., Schweitzer, F., and study supported by the Natural Environment Research Council Mirabel, Ph.: Uptake study of ClONO2 and BrONO2 by (NERC) EAO Doctoral Training Partnership and is fully funded Halide containing droplets, Atmos. Chem. Phys., 4, 1291–1299, by NERC, whose support is gratefully acknowledged (Grant ref https://doi.org/10.5194/acp-4-1291-2004, 2004. no. NE/L002469/1). Derwent, R. G., Middleton, D. R., Field, R. A., Goldstone, M. E., Lester, J. N., and Perry, R.: Analysis and interpretation of air Edited by: Steven Brown quality data from an urban roadside location in Central Lon- Reviewed by: two anonymous referees don over the period from July 1991 to July 1992, Atmos. Envi- ron., 29, 923–946, https://doi.org/10.1016/1352-2310(94)00219- B, 1995. Derwent, R. G., Davies, T. J., Delaney, M., Dollard, G. J., Field, R. References A., Dumitrean, P., Nason, P. D., Jones, B. M. R., and Pepler, S. A.: Analysis and interpretation of the continuous hourly moni- Atkinson, R., Baulch, D., Cox, R., and Crowley, J.: Summary toring data for 26 C 2 -C 8 hydrocarbons at 12 United Kingdom of Evaluated Kinetic and Photochemical Data for Atmospheric sites during, 1996. Atmos. Environ., 34, 297–312, 2000. Chemistry, Web Version, February 2006, IUPAC Subcomm. Gas Derwent, R. G., Dernie, J. I. R., Dollard, G. J., Dumitrean, Kinet. Data Eval. Atmos. Chem., 1–60, 2006a. P., Mitchell, R. F., Murrells, T. P., Telling, S. P., and Field, Atkinson, R., Baulch, D. L., Cox, R. A., Crowley, J. N., Hamp- R. A.: Twenty years of continuous high time resolution son, R. F., Hynes, R. G., Jenkin, M. E., Rossi, M. J., Troe, J., volatile organic compound monitoring in the United King- and IUPAC Subcommittee: Evaluated kinetic and photochemi- dom from 1993 to 2012, Atmos. Environ., 99, 239–247, cal data for atmospheric chemistry: Volume II – gas phase re- https://doi.org/10.1016/j.atmosenv.2014.10.001, 2014. actions of organic species, Atmos. Chem. Phys., 6, 3625–4055, Eigen, M. and Kustin, K.: The Kinetics of Halogen https://doi.org/10.5194/acp-6-3625-2006, 2006b. Hydrolysis, J. Am. Chem. Soc., 84, 1355–1361, Bannan, T. J., Murray Booth, A., Bacak, A., Muller, J. B. A., https://doi.org/10.1021/ja00867a005, 1962. Leather, K. E., Le Breton, M., Jones, B., Young, D., Coe, H., Al- Faxon, C., Bean, J., and Ruiz, L.: Inland Concentrations of Cl2 and lan, J., Visser, S., Slowik, J. G., Furger, M., Prévôt, A. S. H., Lee, ClNO2 in Southeast Texas Suggest Chlorine Chemistry Signif- J., Dunmore, R. E., Hopkins, J. R., Hamilton, J. F., Lewis, A. C., icantly Contributes to Atmospheric Reactivity, Atmosphere, 6, Whalley, L. K., Sharp, T., Stone, D., Heard, D. E., Fleming, Z. 1487–1506, https://doi.org/10.3390/atmos6101487, 2015. L., Leigh, R., Shallcross, D. E., and Percival, C. J.: The first UK Finley, B. D. and Saltzman, E. S.: Measurement of Cl2 measurements of nitryl chloride using a chemical ionization mass in coastal urban air, Geophys. Res. Lett., 33, 6–9, spectrometer in central London in the summer of 2012, and an https://doi.org/10.1029/2006GL025799, 2006. investigation of the role of Cl atom oxidation, J. Geophys. Res., Foster, K. L., Caldwell, T. E., Benter, T., Langer, S., Hemminger, 120, 5638–5657, https://doi.org/10.1002/2014JD022629, 2015. J. C., Finlayson-pitts, B. J., and November, A.: Techniques for Bannan, T. J., Bacak, A., Le Breton, M., Flynn, M., Ouyang, B., quantifying gaseous HOCl using atmospheric pressure ionization McLeod, M., Jones, R., Malkin, T. L., Whalley, L. K., Heard, D. E., Bandy, B., Khan, M. A. H., Shallcross, D. E., and Percival, www.atmos-chem-phys.net/18/13481/2018/ Atmos. Chem. Phys., 18, 13481–13493, 2018 13492 M. Priestley et al.: Observations of organic and inorganic chlorinated compounds

mass spectrometry, Phys. Chem. Chem. Phys., 1, 5615–5621, cal ionization mass spectrometer, Atmos. Meas. Tech., 5, 3029– doi:Doi 10.1039/A907362k, 1999. 3039, https://doi.org/10.5194/amt-5-3029-2012, 2012. Fuentes, M. J., Font, R., Gómez-Rico, M. F., and Martín-Gullón, Le Breton, M., Bacak, A., Muller, J. B. A., Xiao, P., Shallcross, I.: Pyrolysis and combustion of waste lubricant oil from diesel B. M. A., Batt, R., Cooke, M. C., Shallcross, D. E., Baugui- cars: Decomposition and pollutants, J. Anal. Appl. Pyrol., 79, tte, S. J.-B. and Percival, C. J.: Simultaneous airborne nitric 215–226, https://doi.org/10.1016/j.jaap.2006.12.004, 2007. acid and formic acid measurements using a chemical ioniza- Ghernaout, D. and Ghernaout, B.: From chemical disinfection tion mass spectrometer around the UK: Analysis of primary and to electrodisinfection: The obligatory itinerary?, Desalin. Wa- secondary production pathways, Atmos. Environ., 83, 166–175, ter Treat., 16, 156–175, https://doi.org/10.5004/dwt.2010.1085, https://doi.org/10.1016/j.atmosenv.2013.10.008, 2014a. 2010. Le Breton, M., Bacak, A., Muller, J. B. A., Bannan, T. J., Heal, M., Heard, D., Pilling, M., and Whitaker, B.: On Kennedy, O., Ouyang, B., Xiao, P., Bauguitte, S. J.-B., Shall- the Development and Validation of FAGE for Local cross, D. E., Jones, R. L., Daniels, M. J. S., Ball, S. M., Measurement of Tropospheric OH and HO2, J. At- and Percival, C. J.: The first airborne comparison of N2O5 mos. Sci., 52, 3428–3441, https://doi.org/10.1175/1520- measurements over the UK using a CIMS and BBCEAS dur- 0469(1995)052<3428:OTDAVO>2.0.CO;2, 1995. ing the RONOCO campaign, Anal. Methods, 6, 9731–9743, Heeb, N. V., Zennegg, M., Haag, R., Wichser, A., Schmid, P., https://doi.org/10.1039/C4AY02273D, 2014b. Seiler, C., Ulrich, A., Honegger, P., Zeyer, K., Emmenegger, Le Breton, M., Bannan, T. J., Shallcross, D. E., Khan, M. A., L., Bonsack, P., Zimmerli, Y., Czerwinski, J., Kasper, M., and Evans, M. J., Lee, J., Lidster, R., Andrews, S., Carpenter, Mayer, A.: PCDD/F formation in an iron/potassium-catalyzed L. J., Schmidt, J., Jacob, D., Harris, N. R. P., Bauguitte, S., diesel particle filter, Environ. Sci. Technol., 47, 6510–6517, Gallagher, M., Bacak, A., Leather, K. E., and Percival, C. https://doi.org/10.1021/es400760h, 2013. J.: Enhanced ozone loss by active inorganic bromine chem- Hoffman, R. C., Gebel, M. E., Fox, B. S., and Finlayson- istry in the tropical troposphere, Atmos. Environ., 155, 21–28, Pitts, B. J.: Knudsen cell studies of the reactions of N2O5 https://doi.org/10.1016/j.atmosenv.2017.02.003, 2017. and ClONO2 with NaCl: development and application of a Le Breton, M., Hallquist, Å. M., Pathak, R. K., Simpson, D., Wang, model for estimating available surface areas and corrected up- Y., Johansson, J., Zheng, J., Yang, Y., Shang, D., Wang, H., take coefficients, Phys. Chem. Chem. Phys., 5, 1780–1789, Liu, Q., Chan, C., Wang, T., Bannan, T. J., Priestley, M., Per- https://doi.org/10.1039/b301126g, 2003. cival, C. J., Shallcross, D. E., Lu, K., Guo, S., Hu, M., and Hal- Hossaini, R., Chipperfield, M. P., Saiz-Lopez, A., Fernan- lquist, M.: Chlorine oxidation of VOCs at a semi-rural site in dez, R., Monks, S., Feng, W., Brauer, P., and von Glasow, Beijing: significant chlorine liberation from ClNO2 and subse- R.: A global model of tropospheric chlorine chemistry: quent gas- and particle-phase Cl-VOC production, Atmos. Chem. Organic versus inorganic sources and impact on methane Phys., 18, 13013–13030, https://doi.org/10.5194/acp-18-13013- oxidation, J. Geophys. Res.-Atmos., 121, 14271–14297, 2018, 2018. https://doi.org/10.1002/2016JD025756, 2016. Lee, B. H., Lopez-Hilfiker, F. D., Mohr, C., Kurtén, T., Worsnop, D. Huang, B., Lei, C., Wei, C., and Zeng, G.: Chlorinated R., and Thornton, J. A.: An iodide-adduct high-resolution time- volatile organic compounds (Cl-VOCs) in environment of-flight chemical-ionization mass spectrometer: Application to – sources, potential human health impacts, and current atmospheric inorganic and organic compounds, Environ. Sci. remediation technologies, Environ. Int., 71, 118–138, Technol., 48, 6309–6317, https://doi.org/10.1021/es500362a, https://doi.org/10.1016/j.envint.2014.06.013, 2014. 2014. Keene, W. C., Maben, J. R., Pszenny, A. A. P., and Galloway, J. Leu, M.-T., Timonen, R. S., Keyser, L. F., and Yung, Y. N.: Measurement Technique for Inorganic Chlorine Gases in the L.: Heterogeneous Reactions of HNO3(g) NaCl(s) .fw- + Marine Boundary Layer, Environ. Sci. Technol., 27, 866–874, darw. HCl(g) NaNO3(s) and N2O5(g) NaCl(s) .fwdarw. + + https://doi.org/10.1021/es00042a008, 1993. ClNO2(g) NaNO3(s), J. Phys. Chem., 99, 13203–13212, + Kercher, J. P., Riedel, T. P., and Thornton, J. A.: Chlorine activation https://doi.org/10.1021/j100035a026, 1995. by N2O5: simultaneous, in situ detection of ClNO2 and N2O5 Liao, J., Huey, L. G., Liu, Z., Tanner, D. J., Cantrell, C. A., Orlando, by chemical ionization mass spectrometry, Atmos. Meas. Tech., J. J., Flocke, F. M., Shepson, P. B., Weinheimer, A. J., Hall, S. R., 2, 193–204, https://doi.org/10.5194/amt-2-193-2009, 2009. Ullmann, K., Beine, H. J., Wang, Y., Ingall, E. D., Stephens, C. Lawler, M. J., Sander, R., Carpenter, L. J., Lee, J. D., von Glasow, R., Hornbrook, R. S., Apel, E. C., Riemer, D., Fried, A., Mauldin, R., Sommariva, R., and Saltzman, E. S.: HOCl and Cl2 ob- R. L., Smith, J. N., Staebler, R. M., Neuman, J. A., and Nowak, servations in marine air, Atmos. Chem. Phys., 11, 7617–7628, J. B.: High levels of molecular chlorine in the Arctic atmosphere, https://doi.org/10.5194/acp-11-7617-2011, 2011. Nat. Geosci., 7, 91–94, https://doi.org/10.1038/ngeo2046, 2014. Leather, K. E., Bacak, A., Wamsley, R., Archibald, A. T., Husk, A., Liu, X., Qu, H., Huey, L. G., Wang, Y., Sjostedt, S., Zeng, L., Lu, K., Shallcross, E., and Percival, C. J.: Temperature and pressure de- Wu, Y., Hu, M., Shao, M., Zhu, T., and Zhang, Y.: High Levels of pendence of the rate coefficient for the reaction between ClO and Daytime Molecular Chlorine and Nitryl Chloride at a Rural Site CH3O2 in the gas-phase, Phys. Chem. Chem. Phys., 14, 3425– on the North China Plain, Environ. Sci. Technol., 51, 9588–9595, 3434, https://doi.org/10.1039/c2cp22834c, 2012. https://doi.org/10.1021/acs.est.7b03039, 2017. Le Breton, M., McGillen, M. R., Muller, J. B. A., Bacak, A., Shall- Ma, Y., Brooks, S. D., Vidaurre, G., Khalizov, A. F., Wang, L., cross, D. E., Xiao, P., Huey, L. G., Tanner, D., Coe, H., and Per- and Zhang, R.: Rapid modification of cloud-nucleating ability of cival, C. J.: Airborne observations of formic acid using a chemi- aerosols by biogenic emissions, Geophys. Res. Lett., 40, 6293– 6297, https://doi.org/10.1002/2013GL057895, 2013.

Atmos. Chem. Phys., 18, 13481–13493, 2018 www.atmos-chem-phys.net/18/13481/2018/ M. Priestley et al.: Observations of organic and inorganic chlorinated compounds 13493

Mandronich, S.: Photodissociation in the Atmosphere Actinic Flux Sarwar, G., Simon, H., Xing, J., and Mathur, R.: Impor- and the Effects of Ground Reflections and Clouds, J. Geophys. tance of tropospheric ClNO2 chemistry across the North- Res., 92, 9740–9752, 1987. ern Hemisphere, Geophys. Res. Lett., 41, 4050–4058, Mielke, L. H., Furgeson, A., and Osthoff, H. D.: Observation of https://doi.org/10.1002/2014GL059962, 2014. ClNO2 in a mid-continental urban environment, Environ. Sci. Slaper, H., Blumthaler, M., Huber, M., and Kuik, E.: Comparing Technol., 45, 8889–8896, https://doi.org/10.1021/es201955u, ground-level spectrally resolved solar UV measurements using 2011. various instruments: A technique resolving effects of wavelength Mielke, L. H., Stutz, J., Tsai, C., Hurlock, S. C., Roberts, J. M., shift and slit width, Geophys. Res. Lett., 22, 2721–2724, 1995. Veres, P. R., Froyd, K. D., Hayes, P. L., Cubison, M. J., Jimenez, Smedley, A. R. D., Rimmer, J. S., Moore, D., Toumi, R., and J. L., Washenfelder, R. A., Young, C. J., Gilman, J. B., De Gouw, Webb, A. R.: Total ozone and surface UV trends in the J. A., Flynn, J. H., Grossberg, N., Lefer, B. L., Liu, J., Weber, R. United Kingdom: 1979-2008, Int. J. Climatol., 32, 338–346, J., and Osthoff, H. D.: Heterogeneous formation of nitryl chlo- https://doi.org/10.1002/joc.2275, 2012. ride and its role as a nocturnal NOx reservoir species during Spicer, C. W., Chapman, E. G., Finlayson-Pitts, B. J., Plastridge, R. CalNex-LA 2010, J. Geophys. Res.-Atmos., 118, 10638–10652, A., Hubbe, J. M., Fast, J. D., and Berkowitz, C. M.: Unexpectedly https://doi.org/10.1002/jgrd.50783, 2013. high concentrations of molecular chlorine in coastal air, Nature, Mielke, L. H., Furgeson, A., Odame-Ankrah, C. A., and Os- 394, 353–356, https://doi.org/10.1038/28584, 1998. thoff, H. D.: Ubiquity of ClNO2 in the urban boundary layer Thornton, J. A., Kercher, J. P., Riedel, T. P., Wagner, N. L., of Calgary, Alberta, Canada, Can. J. Chem., 94, 414–423, Cozic, J., Holloway, J. S., Dubé, W. P., Wolfe, G. M., https://doi.org/10.1139/cjc-2015-0426, 2016. Quinn, P. K., Middlebrook, A. M., Alexander, B., and Brown, Nguyen, T. B., Crounse, J. D., Teng, A. P., St. Clair, J. S. S.: A large atomic chlorine source inferred from mid- M., Paulot, F., Wolfe, G. M., and Wennberg, P. O.: Rapid continental reactive nitrogen chemistry, Nature, 464, 271–274, deposition of oxidized biogenic compounds to a temper- https://doi.org/10.1038/nature08905, 2010. ate forest, P. Natl. Acad. Sci. USA, 112, E392–E401, Vogt, R., Crutzen, P. J., and Sander, R.: A mechanism for halo- https://doi.org/10.1073/pnas.1418702112, 2015. gen release from sea-salt aerosol in the remote marine boundary Osthoff, H. D., Roberts, J. M., Ravishankara, A. R., Williams, E. J., layer, Nature, 383, 327–330, https://doi.org/10.1038/383327a0, Lerner, B. M., Sommariva, R., Bates, T. S., Coffman, D., Quinn, 1996. P. K., Dibb, J. E., Stark, H., Burkholder, J. B., Talukdar, R. K., Wang, D. S. and Ruiz, L. H.: Secondary organic aerosol from Meagher, J., Fehsenfeld, F. C., and Brown, S. S.: High levels chlorine-initiated oxidation of isoprene, Atmos. Chem. Phys., of nitryl chloride in the polluted subtropical marine boundary 17, 13491–13508, https://doi.org/10.5194/acp-17-13491-2017, layer, Nat. Geosci., 1, 324–328, https://doi.org/10.1038/ngeo177, 2017. 2008. Wang, W. H. and Finlayson-Pitts, B. J.: Unique markers of chlo- Oum, K. W.: Formation of Molecular Chlorine from the Photolysis rine atom chemistry in coastal urban areas: The reaction with of Ozone and Aqueous Sea-Salt Particles, Science, 279, 74–76, 1,3-butadiene in air at room temperature, J. Geophys. Res., 106, https://doi.org/10.1126/science.279.5347.74, 1998. 4939–4958, https://doi.org/10.1029/2000jd900683, 2001. Phillips, G. J., Tang, M. J., Thieser, J., Brickwedde, B., Schus- Wang, X., Wang, H., Xue, L., Wang, T., Wang, L., Gu, R., Wang, ter, G., Bohn, B., Lelieveld, J., and Crowley, J. N.: Signifi- W., Tham, Y. J., Wang, Z., Yang, L., Chen, J., and Wang, W.: cant concentrations of nitryl chloride observed in rural conti- Observations of N2O5 and ClNO2 at a polluted urban sur- nental Europe associated with the influence of sea salt chlo- face site in North China: High N2O5 uptake coefficients and ride and anthropogenic emissions, Geophys. Res. Lett., 39, 1–5, low ClNO2 product yields, Atmos. Environ., 156, 125–134, https://doi.org/10.1029/2012GL051912, 2012. https://doi.org/10.1016/j.atmosenv.2017.02.035, 2017. Reyes-Villegas, E., Priestley, M., Ting, Y.-C., Haslett, S., Bannan, Wong, J. P. S., Carslaw, N., Zhao, R., Zhou, S., and Ab- T., Le Breton, M., Williams, P. I., Bacak, A., Flynn, M. J., batt, J. P. D.: Observations and impacts of bleach wash- Coe, H., Percival, C., and Allan, J. D.: Simultaneous aerosol ing on indoor chlorine chemistry, Indoor Air, 27, 1082–1090, mass spectrometry and chemical ionisation mass spectrometry https://doi.org/10.1111/ina.12402, 2017. measurements during a biomass burning event in the UK: in- Yokouchi, Y., Noijiri, Y., Barrie, L. A., Toom-Sauntry, D., Machida, sights into nitrate chemistry, Atmos. Chem. Phys., 18, 4093– T., Inuzuka, Y., Akimoto, H., Li, H. J., Fujinuma, Y., and 4111, https://doi.org/10.5194/acp-18-4093-2018, 2018. Aoki, S.: A strong source of methyl chloride to the at- Riedel, T. P., Bertram, T. H., Crisp, T. A., Williams, E. J., Lerner, mosphere from tropical coastal land, Nature, 403, 295–298, B. M., Vlasenko, A., Li, S. M., Gilman, J., De Gouw, J., https://doi.org/10.1038/35002049, 2000. Bon, D. M., Wagner, N. L., Brown, S. S., and Thornton, J. A.: Nitryl chloride and molecular chlorine in the coastal marine boundary layer, Environ. Sci. Technol., 46, 10463–10470, https://doi.org/10.1021/es204632r, 2012. Roberts, J. M., Osthoff, H. D., Brown, S. S., and Ravishankara, A. R.: N2O5 oxidizes chloride to Cl2 in acidic atmospheric aerosol, Science, 321, 1059, https://doi.org/10.1126/science.1158777, 2008.

www.atmos-chem-phys.net/18/13481/2018/ Atmos. Chem. Phys., 18, 13481–13493, 2018 5. Paper 3. Detection of highly oxidised molecules from the

reaction of benzene + OH under different NOx conditions

Michael Priestley, Michael Le Breton, Thomas J. Bannan, Stephen D. Worrall, Sungah Kang, Thomas F. Mentel, Iida Pullinen, Sebastian Schmitt, Ralf Tillmann, Astrid Kiendler- Scharr, Jürgen Wildt, Olga Garmash, Archit Mehra, Asan Bacak, Mikael Ehn, Dudley E. Shallcross, Gordon McFiggans, Hugh Coe, Carl J. Percival

In preparation.

Research Highlights:

Iodide detected oxidation products are typically C6 and O8 or less whereas nitrate

observes up to C12 and O18 dimers. Cross correlation between the instruments for the same species ranges from good to poor where the latter is likely caused by inlet artefacts. Iodide detected N containing species from the chamber are identified in urban ambient air but only if the species has high O:C and high N:C.

Author Contributions:

Iodide Tof-CIMS data collection was performed by Michael le Breton and Thomas J. Bannan. The Experiment and Nitrate ToF-CIMS data were collected by Sungah Kang, Iida Pullinen and Thomas Mentel. Data analysis was performed by Michael Priestley. Data interpretation was aided by Dudley Shallcross, Gordon McFiggans, Stephen D. Worrall, Archit Mehra, and Asan Bacak. The article was written by Michael Priestley under the supervision of Hugh Coe and Carl J. Percival.

Blank page

Detection of highly oxygenated molecules from the reaction of

benzene + OH under different NOx conditions

Michael Priestley1, Michael Le Breton1†, Thomas J. Bannan1, Stephen D. Worrall1=, Sungah Kang2, Thomas F. Mentel2, Iida Pullinen2+, Sebastian Schmitt2, Ralf Tillmann2, Astrid Kiendler-Scharr2, Jürgen 5 Wildt2, Olga Garmash4, Archit Mehra1, Asan Bacak1, Mikael Ehn4, Dudley E. Shallcross3,5, Gordon McFiggans1, Hugh Coe1, Carl J. Percival1‖

1Centre for Atmospheric Science, School of Earth and Environmental Sciences, University of Manchester, Manchester, M13 9PL, UK 2Institut für Energie und Klimaforschung, IEK-8, Forschungszentrum Jülich, Jülich, Germany 10 3School of Chemistry, The University of Bristol, Cantock’s Close BS8 1TS, UK 4Institute for Atmospheric and Earth System Research/Physics, Faculty of Science, University of Helsinki, 00014 Helsinki, Finland 5Department of Chemistry, University of the Western Cape, Bellville, South Africa †Now at Department of Chemistry and Molecular Biology, University of Gothenburg, 412 96 15 Göteborg, Sweden ‖Now at Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109 +Now at: Department of Applied Physics, University of Eastern Finland, 7021 32 Kuopio, Finland = Now at School of Materials, University of Manchester, Manchester, M13 9PL, UK 20 Corresponding author: Carl Percival ([email protected])

Abstract

Benzene and other aromatics are a class of volatile organic compounds associated with urban emissions that contribute to the formation of secondary organic aerosol (SOA) and smog events. A nitrate time of flight 25 chemical ionisation mass spectrometer (ToF-CIMS) and an iodide ToF-CIMS were deployed at the Jülich plant

chamber as part of a series of experiments examining benzene oxidation by OH under high and low NOx conditions. Here we compare the characteristics of both systems. More larger species with 6 or more carbon

atoms and 1-4 oxygen atoms were detected under low NOx conditions whereas more smaller species with 6 or

less carbon atoms and 5-8 oxygen atoms were detected under high NOx conditions. In the iodide spectra,

30 approximately 80% and 35% of total signal were assigned formulae in the low and high NOx conditions

respectively. The nitrate scheme detects many oxidation products at high masses, including many C12 dimers

with an average O:C ratio of 1.20. In comparison, very few species with C≥10 and O≥8 were detected with the iodide scheme, and the average O:C ratio was lower at 0.98. This suggests a lower sensitivity to high carbon number, high oxygen content species when using the iodide reagent ion compared with nitrate under these 35 instrumental and experimental conditions. 29 common high mass formulae were detected by both instruments

mainly in the form of C6, O4-6 compounds with an average cross calibration factor of 4.11 ± 2.31 (1σ) Hz iodide -1 2 adduct ppt Nitrate adduct. The correlations between these species ranges from very good (R =0.88) to very poor (R2=0.10) in part due to a difference in sample inlet technique between the two instruments (Eisele type for the

nitrate and the ion molecule reaction region (IMR) for the iodode). A qualitative comparison of laboratory 40 observed benzene oxidation products was made with ambient urban air. 63 signals common to the ambient and chamber work were detected using iodide CIMS although fewer N containing compounds were detected in the

ambient than in the chamber data. Five of the ambient signals show a strong diurnal trend similar to NO2 with a time lag of 2-3 hours. Lower ambient benzene concentrations or increased complexity of the ambient urban air matrix are potential reasons for the lack of ambient observations of the oxidation products identified here with 45 more than 6 carbon atoms.

1. Introduction Benzene is an aromatic volatile organic compound (VOC) commonly used as a vehicular fuel additive (Verma & Des Tombe 2002) and as a chemical intermediate in the manufacture of a range of products e.g. detergents (Oyoroko & Ogamba 2017), lubricants (Rodriguez et al. 2018), dyes (Guo et al. 2018) and pesticides (Wang et 50 al. 2014). Whilst current global estimates of benzenoid emission to the atmosphere by vegetation is of a comparable order with that of anthropogenic activity (~ 5 times lower, Misztal et al. 2015), and background concentrations of benzene are enhanced by increased biomass burning (Archibald et al. 2015), emission of benzene to the urban atmosphere is dominated by vehicle exhausts (Gentner et al. 2012) and solvent evaporation (Hoyt & Raun 2015). In the UK, benzene has been identified as a pollutant whose ambient concentration is not 55 to exceed the national air quality objective of 5.0 µg m-3 measured as an annual average (Department for Environment Food and Rural Affairs 2017).

As well as a toxin (Snyder et al. 1975) and carcinogen (Thomas et al. 2014), benzene is photochemically active

and contributes to the formation of ozone (O3) and secondary organic aerosol (SOA) both of which act to modify the climate and contribute to poor air quality (Henze et al. 2007; Ng et al. 2007). SOA formation from 60 benzene has previously been quantified with a focus on the contribution from smaller mass, ring cleavage reaction products such as dicarbonyl aldehydes (Johnson et al. 2005). For example aromatic oxidation forms aldehydes such as glyoxal that contribute up to 37% of the SOA formed in the LA basin (Knote et al. 2014). Epoxides are another class of compounds formed from aromatic VOC oxidation that contribute readily to SOA formation due to their high reactivity in the condensed phase (Glowacki et al. 2009). These smaller ring 65 cleavage products typically make up a larger fraction of SOA mass than the ring retention products (Borrás & Tortajada-Genaro 2012) and so have traditionally been the main focus for SOA quantification. Other major, toxic benzene oxidation products such as catechol, nitrophenol and maleic anhydride are also known components of SOA (Borrás & Tortajada-Genaro 2012).

More recently the autoxidation mechanism has been demonstrated to occur in aromatic systems (Molteni et al. 70 2016; Wang et al. 2017) producing highly oxygenated molecules (HOM, defined as containing 6 or more oxygen atoms and products of the auto-oxidation mechanism) incorporating up to 11 oxygen atoms. The auto- oxidation mechanism of intra molecular H shifts from the carbon backbone to the peroxy radical centre forming peroxide groups is consistent with other VOC systems (Rissanen et al. 2014) initiated by a variety of different oxidants (e.g. Mentel et al. 2015; Berndt et al. 2016). This mechanism is known to form HOM and as a 75 consequence SOA in ambient rural environments (e.g. Ehn et al. 2014) and is believed to be important in urban

and suburban environments where reductions in NOx concentrations can make the auto-oxidation mechanism competitive (Praske et al. 2018).

The inclusion of an alkyl group to a benzene ring, e.g. ethyl benzene, facilitates HOM formation (Molteni et al. 2016) and is believed to be a consequence of the greater degrees of freedom afforded to the molecule. Whilst

80 benzene is not the most reactive of aromatic VOCs, τOH ≈ 9.5 days for benzene vs. τOH ≈ 6-10 hours for xylenes ( [OH] = 2.0 × 106 molecules cm-3)(Atkinson & Arey 2007), it is ubiquitous in the urban environment and is the

simplest C6 aromatic ring system to study. The demonstration of HOM formation from the benzene system suggests the autoxidation mechanism for aromatic systems could potentially contribute to SOA formation in urban environments.

85 Oxidation of benzene occurs nearly exclusively via hydroxyl radical (OH) addition to form the cyclohexadienyl

radical/benzene-OH adduct, which subsequently adds O2 to form the hydroxycyclohexadiene peroxy radical

(C6H6-OH-O2) (Volkamer et al. 2002) (Fig. 1, blue). Two subsequent reaction pathways are postulated for this

peroxy radical: either elimination of HO2 yields phenol and secondary OH attack must occur for further oxidation (Fig. 1, green); or an endocyclic di-oxygen bridged carbon centred radical intermediate is formed by 90 the addition of the peroxy group to (typically) a β-carbon (Glowacki et al. 2009)). This di-oxygen bridge carbon

centred radical may add another O2 and form a peroxy radical (named as BZBIPERO2 in the master chemical mechanism, MCM, Saunders et al. (2003)) (Fig. 1, red). Auto-oxidation may continue forming either: a second oxygen bridged radical, described as type II auto-oxidation (Molteni et al. 2016) (Fig 1, BZBIPERO2-diB, gold), or the more typical auto-oxidation of intra molecular hydrogen abstraction forming a hydroperoxide 95 carbon centred radical (type I auto-oxidation).

At each step, termination of the peroxy radical to hydroperoxide, carbonyl or hydroxyl or reduction to an alkoxy

radical is possible. Reduction of C6H6-O2-OH to the alkoxy equivalent by NO is not an active reaction channel as the rate is slow and requires unrealistically high NO concentrations (Glowacki et al. 2009). However, reaction

with other RO2 species can follow this reaction pathway, although the significance of this reaction channel is

100 uncertain in NOx dominated conditions.

Fig 1. Reaction scheme of benzene oxidation by OH with proposed dibridged species adapted and expanded from Molteni et al. (2016). Names are taken from the MCM where available.

Typically, the presence of NOx alters the product distribution of VOC oxidation and reduces SOA formation 105 (e.g. Stirnweis et al. 2017). With other atmospherically relevant VOC precursors of SOA e.g. isoprene or

monoterpenes, high NO conditions can supress SOA formation (Wildt et al. 2014) as reduction of RO2 to RO ultimately leads to fragmentation of the RO species (Surratt et al. 2006, Nguyen et al. 2015), but it can also form epoxides, aldehydes and hydroperoxides which readily partition to the aerosol phase and contribute to SOA formation (Surratt et al. 2010).

110 The further reaction of NOx with these oxygenated VOCs produces species including nitro organics, nitrates, peroxy nitrates and peroxyacyl nitrates (Atkinson 2000) that can also condense and contribute to SOA

formation. Under conditions where NOx is present, the HO2:OH ratio is low as HO2 is reduced by NO to form

OH. This allows more OH oxidation to occur and less termination of RO2 by HO2 to occur; whereas under low

NOx conditions VOC consumption is lower, as OH recycling from HO2 relies on the slower HO2 + O3 reaction 115 (Atkinson 2000). Here low and high NOx are relative terms and are defined by the available VOC. This

reduction of the HO2:OH ratio as a function of NO has been observed at Mace Head, Ireland where clean low

NOx marine influenced air can be contrasted with polluted NOx containing continental air (Creasey et al. 2002).

In that instance, an increase in NO concentration from 0.01 ppb to 1 ppb reduces the HO2:OH ratio from 200:1 to 10:1.

120 Condensation reactions are more prevalent between multifunctional oxygenated compounds such as multifunctional alcohols and peroxy radicals, the latter of which are known to self-react and cross-react, forming

higher mass oligomers (Surratt et al. 2006). SOA yields from aromatics are sensitive to NOx concentrations as

higher concentrations of NOx compete with these RO2/HO2 condensation reactions (Ng et al. 2007). In neutrally (non-acid) seeded experiments where a surface is made available onto which material with low vapour pressures

125 can condense, more SOA is formed under low NOx conditions and a delayed onset in SOA formation is

observed. This delay is a consequence of NO concentrations approaching zero, indicating that the RO2 + HO2

reaction is important for SOA growth as it becomes competitive with the RO2 + NO reaction at low NO (Ng et al. 2007).

Time of flight chemical ionisation mass spectrometry is a measurement technique frequently used to probe VOC 130 oxidation due to the ability to detect many low concentration compounds simultaneously in real time (e.g. Chhabra et al. 2015). As a result of the sensitivity of the nitrate ionisation scheme towards HOM, this reagent ion is typically used to study the auto-oxidation mechanism and HOM formation. However, to achieve carbon mass closure of the system, multiple ionisation schemes are required due to their complimentary yet differing sensitivities towards OVOCs with different oxidation states and functional groups (Isaacman-VanWertz et al. 135 2017).

In this paper the oxidation of benzene by OH under high and low NOx conditions is investigated in the Jülich plant atmosphere chamber (JPAC) with two time of flight chemical ionisation mass spectrometer (ToF-CIMS) using the iodide ionisation scheme. A nitrate ToF-CIMS is used to validate and compare measurements of detected benzene oxidation products. The compounds identified here are then used to assess the contribution of 140 benzene oxidation products to ambient urban air.

2. Experimental The experiments were performed in a 1450 L borosilicate continuous flow reactor of the Jülich Plant Atmospheric Chamber (JPAC) described elsewhere (Mentel et al. 2009).

OH / 107 Reacted -3 -1 -3 -1 - Exp NOx / ppb J(O1D) / x10 s JNO2 / x10 s molecules cm Benzene / ppb 3

0 (Low NOx) ≤ 0.3 65 2.6 0 12.0

65 2.6 0 5.8 1 (High NOx) 20.00 65 2.6 4.3 6.5

Table 1. Summary of selected experiments performed at JPAC used in this study. Two OH concentrations are reported for

145 the NOx containing studies. The first describes steady state concentrations with JNO2 and ultra violet (UV) light on, the

second with only the JNO2 light on.

The chamber is operated in a continuous flow mode in order to homogenise the mixture giving an average residency time of ~45 minutes. Benzene (Merck, ≥99.7%) was introduced into the chamber by flowing purified ambient air over a diffusion source to maintain a constant concentration and was monitored by recording the 150 inlet and outlet concentrations by quadrupole proton transfer chemical ionisation mass spectrometry (PTR- CIMS). The PTR-MS (IONICON) was calibrated using a benzene diffusion source. The difference between the outlet and inlet concentrations, measured at a 2 minute time resolution and switched every 25 minutes, describes the reacted benzene, from which the OH concentration is calculated.

OH is generated by photolysis of O3 by UV lamps (λ < 254 nm, Philips, TUV 40W herein termed TUV) and 1 155 subsequent O D reaction with H2O. The O3 is introduced to the chamber by a separate line to the benzene in order to prevent reactions occurring in the line and was monitored by UV absorption (Model 49, Thermo

Environmental instruments). NOx was measured by chemiluminescence (CLD 770 and PLC 760, Eco Physics).

For the high NOx experiment, NO (Linde, 99.5±5 ppm NO in N2) was injected into the chamber via a separate PFA line. Twelve discharge lamps termed UVA (HQI, 400 W/D; Osram, Munich, Germany) are selectively -3 -1 160 activated to photolyse NO2 (JNO2 4.3x10 s ) to increase the NO/NO2 ratio, although NO2 remains the dominant

NOx compound at > 70% of total NOx. Temperature and humidity were maintained at 288 ± 1 K and 67 ± 2 % through the experiments. The experimental conditions are summarised in Table 1.

2.1. ToF-CIMS Instrumentation Iodide

165 The University of Manchester ToF-CIMS has been described in detail by Priestley et al. (2018). The iodide ToF-CIMS ion molecule reaction region (IMR) was held at a constant pressure of 100 mbar by a scroll pump (Agilent SH-112) controlled with a servo control valve placed between the scroll pump and the IMR. The short segmented quadrupole (SSQ) region was held at 1.30 mbar by a second scroll pump (Triscroll 600). Chamber air was drawn into the IMR via a critical orifice at 2.2 standard litres per minute (slm) where it mixed 170 with I- ions. The reagent I- ions were created by flowing 10 standard cubic centimetres per minute (sccm) of o nitrogen (N2) over a methyl iodide (CH3I) permeation tube made of 1/8” PFA and held at 40 C. This flow met 2

210 slm N2 and was flowed through a Po 10 mCi alpha emitting reactive ion source (NRD Inc.). The mass ------calibration was performed for 7 known masses: NO3 , I , I .H2O, I .HCOOH, I .HNO3, I2 and I3 , covering a mass range of 62 to 381 m/z. The mass calibration was fitted to a 3rd order polynomial and was accurate to within 2 175 ppm, ensuring peak identification was accurate below 20 ppm. The resolution was 3231 at 127 m/z and 3720 at 381 m/z. The inlet to the I- ToF-CIMS sampled from the middle of the chamber and was comprised of 50 cm 1/4” O.D. PFA tubing giving residency time of <1.5 s.

Nitrate Chamber air was drawn into the instrument at 10 slm through a 30cm (ID 8 mm) stainless steel tube into the - 241 180 chemical ionisation source where NO3 reagent ions were generated by flowing HNO3 (Merck) through a Am source. The Nitrate ToF-CIMS is calibrated using the methodology described in Ehn et al. (2014). Briefly, the calibration factor for all HOM with carbon numbers equal to or greater than 5 are assumed to be the same as that - for H2SO4 as these are assumed to form NO3 adducts at the collision limit (Ehn et al. 2014). The H2SO4

calibration factor is determined by generating steady state concentrations of H2SO4 within the chamber. H2SO4

185 production is calculated by the reaction of OH with SO2, which is rate limiting. The concentration of OH is estimated by reaction with the tracer (E)-2-butene. In the absence of particles the first order wall loss is the

dominant sink of H2SO4.

3. Results We detect a range of benzene oxidation products, including many with the same formula as those listed in the 190 MCM, as well as highly oxidised species with both ToF-CIMS instruments. Well resolved peaks at a given unit mass were frequently observed in the I- spectrum. The separation between the I- cluster with its large negative mass defect and the positive defect of the high H containing deprotonated species allows for good peak - separation at the resolution of the ToF-CIMS (Fig 2). Beyond I and H2O, the largest adduct signals we

identified with the iodide CIMS were low mass species: formic acid, nitric acid, CH2O3, C3H6O3, HONO,

195 C4H4O3 and C3H8O3.

- Fig 2. Example of peak separation between an iodide adduct I.C2H2O3 and deprotonated signal C10H17O4 detected during the

low NOx experiment.

3.1. Product distributions under different NOx conditions measured by iodide CIMS

200 The most obvious difference between low and high NOx experiments is the number of nitrogen containing

compounds detected. 40 N containing VOC reaction products (CHON) were detected in low NOx conditions of

which 23 were adducts. As the chamber had a NOx background of ~ 300 ppt, some CHON products will still

form in the low NOx experiment. In the high NOx experiment 73 CHON compounds were detected of which 47

were adducts, many of which were C2-C6. These include IC3H3NO7, C6H3N3O11, IC4H3NO7, IC4H5NO7,

205 C6H3N3O12, IC6H4N2O5, the molecular formulas corresponding to the MCM products, NBZFUO, HCOCOHPAN, NDNPHENO, NBZFUOOH, NDNPHENO2, NDNPHENOOH, DNPHEN respectively (Fig 3).

-3 -1 -1 Fig 3. High resolution difference mass spectrum (between JO1D = 2.6x10 s and JO1D = 0 s ) from the high NOx experiment. An example of several N containing species are observed when OH oxidation is occurring.

210 The number of different detected molecules containing six carbon atoms or less is greater under high NOx

conditions whereas the number of different detected molecules with a carbon number greater than C6 is more

frequently observed in low NOx conditions, albeit at a much lower frequency. The greater frequency of

molecules with high carbon numbers detected under low NOx conditions is consistent with other VOC oxidation systems that indicate the increased propensity for condensation reactions and oligomerisation to occur (as 215 described in the introduction).

Fig 4. Summary of CHON statistics of detected oxidation products during high (20 ppb, purple) and low (0.3 ppb, grey)

NOx benzene oxidation with the iodide ionisation scheme. a, b, c, d. Frequency distributions of the atoms C, H, O and N for the detected oxidation products. e. The average number of atoms per detected oxidation product split into C, H, O and N. f. 220 The average element ratios of O:C, H:C and N:C for the detected oxidation products. g. The average mass of a detected

oxidation product for high and low NOx conditions. g. The parameterised average carbon oxidation state (OSc) of the

detected oxidation products for high and low NOx conditions.

The O:C ratio of the detected products is slightly larger in the high NOx case (Fig 4, f). Whilst the carbon

number of an ‘average’ detected molecule is marginally higher in the low NOx case (5.2 vs. 5.1; Fig 4, e), the

225 average oxygen number is higher in the high NOx case (4.0 vs. 4.2; Fig 4, e) with the overall effect of increasing

the observed O:C ratio (Fig 4, f). The frequency of oxygen containing products is dependent on the NOx

condition.There is a greater frequency of high O content (O5-O10) species detected in the high NOx case and a

greater frequency of low O content (O-O4) species detected in the low NOx case (Fig 4, c).

The high O content in the high NOx case is likely due to the inclusion of NOx to form nitrated organic

230 compounds e.g. dinitro-phenols and dinitro-catechols are observed that are not present under low NOx conditions. In both cases, compounds containing more than 8 oxygen atoms are detected at significantly lower levels than those containing less than 8 oxygen atoms.

Despite the higher oxygen content of molecules in the high NOx case, the derived average carbon oxidation state

(OSc, Kroll et al. 2011) is lower (0.13) than in the low NOx case (0.22) (Fig 4, h). This is due to the assumption 235 that all organic nitrogen is in the form of nitrates, which is not necessarily the case. As well as the higher O:C

ratio, the H:C ratio is also marginally higher in the high NOx case (Fig 4, f). This indicates that the loss of

aromaticity is more prevalent in the high NOx case, consistent with more open ring products.

Fig 5. Time series of PANs detected by iodide CIMS. The time during which the response is unchanged is greater than the 240 residence time suggesting off gassing from the walls of the chamber.

A series of nitrated species, C2H3NO5, C3H5NO5, C4H7NO5 and C6H11NO5 are tentatively identified as peroxy

acyl nitrates (PANS) of the form CxHx-1NO5, a class of compounds that have previously been detected by iodide CIMS (Phillips et al. 2013). PAN growth begins when the OH oxidation begins and stops once when photochemical OH production is stopped (Fig 5). The growth curves appear independent of the concentration of

245 NO2 as no change in temporal profile is observed once NO2 photolysis is stopped. The time scale of the unchanged growth curves is much greater than the chamber residence time (~45 mins) suggesting a dominant source of PANs is the chamber wall.

Fig 6. Frequency distributions of atoms (CHON) in detected molecules for the high NOx (20 ppb) experiment with J(O1D) = 250 2.6 x10-3 s-1.

Of the 73 CHON species identified in the high NOx case only 20 of these exhibits higher signal when JNO2 is

active, suggesting these species are more likely to be organic nitrates or nitrites from the reaction of RO/RO2 +

NO. When NO2 photolysis is stopped, the signals for the remaining 53 species grow, indicating that at least

some of their signal contribution comes from other formation by other mechanisms i.e. RO/RO2 + NO2.

255 A higher distribution of NO1-5 containing compounds is found when the JNO2 is inactive, except for NO4

compounds which are more commonly detected during NO2 photolysis. A higher incidence of NO4 compounds

when more NO and less NO2 is available may suggest a greater propensity to incorporate NO to material that has already been oxidised i.e. already contains 2 oxygen atoms such as a hydroxyl-alkoxy radical. Hydroxyl- alkoxy radicals should be more prevalent when NO concentrations are high as the abundance of NO reduces any

260 peroxy radicals to alkoxy radicals. The CHON atom frequency distributions for the JNO2 photolysis are summarized in Fig 6.

Fig 7. Mass spectra of identified species at key points during the high NOx experiment. a. nitrophenol. b.

C4H3NO6. Both signals enhance when OH oxidation begins when J(1OD) and JNO2 are active (green). When the JNO2

265 is deactivated, the signals grow again (red). Growth with J(1OD) and JNO2 active suggests RO + NO reaction is

dominant thus forming R-NO2 species. When JNO2 is deactivated, this channel must become less prevalent as some

NO can now remain oxidised as NO2; however, the signal still grows. This suggests either isomeric compounds of these species contribute further signal or degassing from the walls causes their increased concentration in the chamber. The presence of signal before the experiment suggests the chamber background is already quite high.

270 Nitrophenol (C6H5NO3) is assigned at m/z 266 and m/z 311 is assigned as dinitrophenol (C6H4N2O5). These

species are formed from the addition of NO2 to an alkoxy radical which is formed by the reaction of OH with an alcohol (Saunders et al. 2003). In the case of the former this is from the phenol + OH reaction to yield the phenoxy radical and for the latter from nitrophenol + OH reaction to form the nitrophenoxy radical. Both these

species are formed when NO2 photolysis has been stopped indicating the concentration of NO2 is controlling

275 their formation rather than the concentration of alkoxy radicals, whose concentration decreases once NO2 photolysis is stopped (e.g. Fig 7, a).

C4H3NO6 is identified at m/z 288. This forms immediately under high NO conditions, suggesting nitrates and or

nitrites constitute the majority of signal. When NO2 photolysis stops, the signal grows again. This suggests a

shift in the isomers contributing to the signal from RO/RO2 + NO products (e.g. RONO and RONO2) to

280 RO/RO2 + NO2 products (e.g. RNO2, ROONO2).

3.2. MCM Products The MCM lists 137 multigenerational oxidation products of benzene including important intermediate and

radical species. Using the iodide ToF-CIMS, we find 100 and 105 peaks under low and high NOx conditions corresponding to potentially 104 and 117 unique formulae respectively, accounting for isomers. These signals 285 are a mixture of adducts and non-adduct peaks that match the exact masses of the MCM oxidation products to within 20 ppm error. Whilst only ~100 of the species with a formula cited in the MCM are detected, signals

increase for 362 and 350 peaks in the mass spectrum, under low and high NOx conditions and when oxidation is occurring indicating many more products are made than are explicitly accounted for in the MCM. The MCM

identified signals contribute 80% of the total signal recorded in the low NOx experiment, but only 35% in the - - 290 high NOx experiment, controlling for signal from I and I.H2O .

3.3. Auto-oxidation Products To test the extent to which the iodide reagent ion is able to detect benzene auto-oxidation products, a number of theoretically suggested formulae based on the auto-oxidation mechanism with propagation and termination steps were devised and then searched for within the spectra. The applied mechanism builds upon the current MCM 295 benzene oxidation scheme. It consists of auto-oxidative intra molecular hydrogen shifts from the carbon

backbone to a peroxy radical group forming a peroxyl group, followed by O2 addition to form a new peroxy radical (Fig 8). The peroxy radical groups were terminated to carbonyl, hydroxyl, nitrate or peroxyl groups or reduced to form the alkoxy equivalent.

300 Fig 8. Auto-oxidation and termination steps in the absence and presence of NO.

The peroxy radicals chosen are the initial benzene-OH-O2 adduct and its phenol equivalent as OH attack has been observed up to three times in a benzene OH flow tube system (Molteni et al. 2016). These two species may form an endocyclic dioxygen bridge where instead of the peroxy group abstracting an H atom, it adds to the carbon backbone. From benzene, and phenol, BZBIPERO2 and PHENO2 form, both of which feature in the 305 MCM. Also included here is the catechol equivalent labelled CATECO2-b. Each of these peroxy radicals may terminate as described previously (Fig 8.). BZBIPERO2 and PHENO2 may also form a second endocyclic di oxygen bridge as suggested by Molteni et al. (2016). These peroxy radicals may also terminate. This gives a total of 47 species, 33 of which are not in the MCM. These are summarised in Fig 9.

Of the 47 formulae predicted by following the mechanistic routes, signals pertaining to 35 and 13 formulae were

310 identified in the low and high NOx iodide spectra respectively. Of these, 4 and 1 have no known isomers or

isobars (fig 9, see species labelled with +) in the low and high NOx iodide spectra respectively. This suggests

these signals refer to the molecules predicted here. For the single signal in the high NOx case, this is C6H7O8 (PHENO2-diB), the second endocyclic auto-oxidation product of the Phenol-OH-O2 adduct. This molecule is

also detected in the low NOx case, along with C6H8O8 (PHENO2-diB-hydroperoxy), C6H7NO5 (phenol-oh-

315 nitrate) and C6H7NO4 (benzene-oh-nitrate):

Fig 9. Reaction scheme of the auto-oxidation mechanism considered for the iodide spectra. Blue highlights peaks detected in

low NOx conditions and red in high NOx, pink describes peaks detected in both. An asterisk (*) indicates the formula is not present in the MCM. A plus (+) indicates the formula is unique i.e. no isomers are listed in the MCM or the devised

320 theoretical products. More nitrated peaks are observed in high NOx conditions and more peroxide peaks are observed in low

NOx conditions. In general, more peaks identified as highly oxidised species are detected in low NOx conditions, despite

nearly double the number of peaks being detected in high NOx conditions (444 vs. 726). No oxidation products of a di- oxygen bridged species from catechol + OH oxidation are observed.

3.4. Comparison of Nitrate and Iodide CIMS 325 The nitrate ionisation scheme detects a greater number of higher mass species with a higher oxygen content (Fig 10, c, e) and higher average carbon oxidation state (Fig 10, h), whereas the iodide ionisation scheme detects lower mass compounds with fewer oxygens and a lower OSc, as reported elsewhere (Isaacman-VanWertz et al. 2017). For example, the average mass of a nitrate adduct is ~240 g mol-1 with ~30% of that attributable to oxygen, whereas the average mass of an iodide adduct is 140 g mol-1 with ~20% attributable to oxygen (Fig 10, g). These 330 calculations do not include the reagent ion.

Fig 10. Summary of CHON statistics of detected oxidation products during the high (20 ppb) NOx benzene oxidation for iodide (purple) and nitrate (green) ionisation schemes. a, b, c, d. Frequency distributions of the atoms C, H, O and N for the detected oxidation products. e. the average number of atoms per detected oxidation product split into C, H, O and N. f. The 335 element ratios of O:C, H:C and N:C for the detected oxidation products. g. The average mass of a detected oxidation product detected by iodide and nitrate ionisation. g. The parameterised average carbon oxidation state (OSc) of the detected oxidation products detected by iodide and nitrate ionisation.

The nitrate CIMS is able to detect a greater frequency of high mass species greater than C6 including a large

number of dimers (C12) whereas the iodide scheme is unable to detect many species greater than C6 (Fig 10, a).

340 The reduced frequency of C>6 compounds detected by the iodide CIMS compared with the nitrate CIMS suggests its sensitivity to those higher mass compounds is low.

Few species greater than C12 are detected with either reagent ion. The iodide scheme detects a maximum number

of molecules containing O3 and O4 and very few with more than O7 (Fig 10, c). The nitrate scheme observes a

broad range of oxygen numbers peaking between O9 and O11 but up to O18. At higher hydrogen numbers (H>8), 345 fewer molecules containing an odd number of Hs relative to molecules containing even Hs are observed in the nitrate spectrum. This indicates the detectability of non-radical species is more common than radicals, whereas the iodide scheme does not discriminate between the two (Fig 10, b).

Fig 11. Comparison of H:C vs O:C, On vs Cn and average carbon oxidation state (OSc, described below) vs Cn detected by

350 nitrate and iodide CIMS. intensity of color represents number of identified compounds (arbitrary units). Green shows NO3 ionisation, purple shows iodide.

Iodide detects a H:C/O:C of about 1.5 whereas nitrates detects H:C/O:C of about 1.2 due to the higher O:C ratio of the detected species (Fig 11). With both nitrate and iodide, the maximum variation in oxygen number per

molecule is detected for C6 compounds although for nitrate this is closely followed by C12 i.e. C6 dimers.

355 For iodide, the variation in oxygen number per molecule decreases symmetrically as the number of carbons

moves away from C6. However for nitrate, the mean number of oxygen atoms per molecule increases linearly as the carbon number increases. This defines the oxygen content limit for the iodide CIMS in this system. .

The OSc detected by the ionisation schemes is different for the same carbon number. For C6 compounds, nitrate observes an OSc ~ 2, whereas iodide observes an OSc < 1. This difference indicates that over the same mass 360 range, the nitrate is more sensitive to more oxidised carbon containing compounds. Over the entire carbon range, for the iodide scheme, the OSc decreases from ~3 at low carbon numbers to ~ -1 at high carbon numbers due to the relatively increasing number of C-H bonds to C-O bonds, thus reflecting its selectivity for compounds with a mid-ranged oxygen content. For the nitrate scheme, as carbon number increases, the OSc plateaus at ~ 1, indicating the selectivity for more C-O bonded species. Below reagent ion masses, a range of carbon oxidation 365 states at low carbon numbers are recorded for the iodide spectrum. These are attributable to fragments found at lower masses than the reagent ion m/z 127.

Twenty nine formulae are found to be common between the two instruments representing the overlap of oxidation product signals with 16 of those being iodide and nitrate adduct matches (Fig 12). The behaviour of the signals at the start of oxidation is highly variable and is specific to each instrument and ion measured by that 370 instrument. Typically, the level of highly oxidised species increases significantly from the outset but relaxes to

lower levels rapidly where they grow again to reach steady state. This early spiking behaviour has been observed in other studies investigating HOM formation using the nitrate ToF-CIMS (e.g. Ehn et al. 2014) and is thought to be a consequence of a latent aerosol sink shifting the equilibrium of HOM species from the gas to the aerosol phase. The iodide ionisation scheme is between 15 - 60 times more sensitive than the nitrate scheme, but 375 suffers from a slower response time in terms of growth to steady state and relaxation to background levels. This is most likely due to semi-volatile products that have deposited on the inlet and instrument walls/IMR taking longer to move through the iodide CIMS whereas the nitrate CIMS with the Eisele type inlet does not suffer from this problem (Eisele & Tanner 1993). The correlations between the signals of the same formula between both instruments vary from 0.10 to 0.88; the variation is likely due to the differing mechanics of sampling 380 introduction into the two instruments (Table 2). Where correlations are very poor, it is likely that different isomers are detected by the different ionisation schemes.

Fig 12. Cross calibration of species with common formulae detected by nitrate and iodide ionisation during the high NOx

(20ppb) experiment (20 minute time resolution). Green shows NO3 ionisation, purple shows iodide.

385 Where nitrate and iodide adducts for the same highly oxidised species exist, cross calibration factors were calculated. This was done by dividing the iodide adduct time series by the nitrate adduct time series at a period where both signals had reached steady state. 16 highly oxidised species adduct formulae overlap with an average cross calibration factor of 6.23 ± 8.39 (1σ) Hz ppt-1 (Fig 13). The standard deviation is larger than the stated

value due to the inclusion of an outlier. This outlier is C6H6O4 to which the iodide CIMS is much more sensitive

390 when compared to the other highly oxidised species. The cross calibration factor for C6H6O4 is 36.70 ± 2.19 (1σ), which is high for this instrument. Removing this signal results in an average cross calibration factor of 4.11 ± 2.31 (1σ) Hz ppt-1.

Adduct R2

C6H6O4 0.88

C6H4O4 0.87

C6H6O6 0.83

C6H5O6 0.81

C6H9O6 0.79

C6H6O7 0.78

C6H12O6 0.73

C6H11O6 0.72

C6H7O6 0.72

C6H8O5 0.70

C6H8O4 0.67

C6H8O6 0.60

C6H6O5 0.58

C6H9O5 0.57

C7H8O4 0.22

C6H10O4 0.10 2 Table 2. Summary of R between the highly oxidised species (20 minute time resolution) at high NOx.

- Whilst maximum concentrations of highly oxidised species measured by the NO3 CIMS are all 3 - 4 ppt 395 (varying by a factor of 1.25), the corresponding maximum signals on the I- CIMS vary from 5 - 20 Hz (factor of 4). Unlike nitrate, iodide is not expected to form adducts with highly oxidised species at the collision limit. That is demonstrated here by the spread in magnitude of iodide signals for identified highly oxidised species.

Fig 13. Individual and average cross calibration factor(s) derived for the common highly oxidised species detected by the 400 nitrate and iodide CIMS.

- The outlier signal of C6H6O4 suggest the I CIMS is more sensitive to this compound than the rest of the iodide highly oxidised species signals, which all reach similar peak ion counts of 5 - 10 Hz (Fig. 12). The similarity in peak ion counts suggests these compounds have a similar sensitivity with respect to I-. The I- CIMS is known to

be highly sensitive to carboxylic acids and di-carboxylic acids. Whilst other O4 compounds have a similar cross

405 calibration factor to other non O4 containing compounds, the higher signal of the C6H6O4 suggests it is the dicarboxylic acid, muconic acid. Another explanation is the contribution from different isomers of the same

exact mass to a single signal increases the number of counts. For instance C6H6O4 has 3 isomers listed in the MCM (BZEMUCCO2H, BZIBIPEROH, PBZQOH) and 1 suggested formula, the acyl termination product of

the catechol-OH-O2 adduct from the predicted oxidation products from section 3.3.

410 3.5. Detection of highly oxidised species in the field

The signals here identified as benzene highly oxidised species from the iodide spectra in both high and low NOx cases are applied to an ambient dataset to assess the capability of the iodide ToF-CIMS to detect highly oxidised species in an urban environment. The instrument was deployed at the University of Manchester between 2014- 10-29 and 2014-11-11. Further experimental details can be found in Priestley et al. (2018). Of the highly

415 oxidised species identified in the low and high NOx experiments by the iodide CIMS, 63 deprotonated and adduct signals were identified in the ambient data set.

Fig 14. Summary of CHON statistics of detected oxidation products during the high NOx (20ppb) chamber sampling (JPAC, blue) and ambient urban sampling (MAN, red) both using the iodide ToF-CIMS. a, b, c, d. Frequency distributions of the 420 atoms C, H, O and N for the detected oxidation products. e. the average number of atoms per detected oxidation product split into C, H, O and N. f. The element ratios of O:C, H:C and N:C for the detected oxidation products. g. The average mass of a detected oxidation product detected in the chamber and in ambient data. g. The parameterised average carbon oxidation state (OSc) of the detected in the chamber and in ambient data.

A greater number of C6 compounds are detected in the chamber than in ambient data but slightly higher mass C 425 compounds are detected in ambient (Fig 14, a). This likely reflects the mixed VOC sources in ambient contributing to higher carbon number oxidation products that are isomeric rather than in the chamber where

benzene was the only VOC source. More oxygen containing compounds are detected in the chamber where O<8 but above this, threshold detection is much rarer in both environments (Fig 14, c). This again appears to highlight the limitation of the ionisation scheme to detect higher mass and more oxygenated compounds. The 430 number of nitrogen containing compounds is much greater in the chamber data than in the ambient most likely

due to the sustained high concentrations of NOx. Less detectable nitrated highly oxidised species exists in the ambient conditions (Fig 14, d). The higher N content in the chamber causes the calculated OSc for the chamber data to be lower than ambient (Fig 14, h). The average mass of an oxygenated VOC molecule in the chamber

data is lower than in the ambient (Fig 14, g). This is a consequence of the high concentration of the C6 benzene 435 precursor in the chamber compared with mixed VOC sources in the ambient.

Fig 15. Diurnal profiles of highly oxidised species and NOx which show bimodal distributions normalised to the

first measurement point. Cross correlations between highly oxidised species is best with NO2. Only C6H9O5 shows the best correlation with a lag = 0 hours; the rest show the best correlation with an offset of 2-3 hours..

440 Many of the highly oxidised species signals show a diurnal profile consistent with photochemical pathways with a maximum at solar noon and minimum during the night. A selection of the detected species show temporal

behaviour that mimics the diurnal profile of NOx with a time lag of 2-3 hours, suggesting their formation is

reliant on the presence of NOx (Fig 15). This could either be because the starting material is co-emitted with NO

and ages to produce oxidised material as NO2 is formed, or its production is directly related to the presence of

445 NO2. It is not possible to unequivocally prove the signals in the ambient data set originate from the oxidation of benzene (or any other aromatic species) but this data at least indicates the potential for that occurrence. Many of the signals have higher hydrogen content than would be expected for benzene oxidation products without loss of aromaticity or condensation possibly suggesting an alkane source.

450 Fig 16. Comparison of O:C vs N:C for C6 to C9 adducts detected during high and low NOx conditions in the ambient and chamber datasets. The size of the point represents the number of species detected. The colour represents the number of carbon atoms. Circles represent detection in the chamber data whereas crosses represent detection in the ambient data. A

cluster of C6 compounds with high O:C and high N:C ratio are detected in the ambient dataset. Few low O:C low N:C compounds or compounds with more than 6 carbon atoms are detected in the ambient dataset.

455 Of the highly oxidised species observed in the ambient data, a higher N:C and higher O:C is observed than for

that in the chamber (Fig 16) under either condition of NOx. Typically a high O:C ratio of a compound detected with the iodide CIMS would suggest a molecule with a low carbon number, however the analysis here only

considers C≥6 compounds, therefore the low O:C and N:C species observed in the chamber that are not observed

in the ambient must have a higher carbon number, which can be seen. The lack of C>6 compounds in the ambient

460 data suggests the formation of C>6 compounds in the chamber are not representative of ambient processing. The

NOx concentrations of ~20 ppb under the high NOx conditions in the chamber are more representative of the

ambient NOx (average 17 ppb) and it is under these conditions that condensation reactions are less prominent. Further, ambient air is more complex than a single VOC system and the benzene concentrations in the chamber

are higher than typical ambient concentrations. It is likely that the formation of C>6 compounds, presumably 465 from condensation reactions, under urban ambient conditions is much less likely for the benzene system than the chamber data indicates. The tuning of the instrument was similar between data collections and so is not thought to be affecting the selectivity of the instrument hence the observations are not considered an instrumental artefact.

4. Conclusion 470 An iodide time of flight chemical ionisation mass spectrometer was deployed at the Jülich plant chamber as part

of a series of experiments examining benzene oxidation by OH under high and low NOx conditions. A range of deprotonated and adduct signals were identified including high mass (AMU) nitrogen containing organic

signals. The prevalence of C≥6 and O-O4 detected species increases under low NOx conditions whilst the C≥6 and

O5-O7 signal is higher under high NOx conditions. The detection of C≥10 and O≥8 containing compounds is much

475 less frequent in both cases, indicating a sensitivity limit. The higher O:C ratio under high NOx conditions is

symptomatic of detected NOx containing organic compounds containing more oxygen, although as the oxygen is

assumed to be part of nitrated compounds, this results in a lower OSc than the low NOx case. In the iodide spectrum 104 and 117 signals match the formulae of 137 MCM products detected. These 137 formulae

correspond to ~80% and ~35% of identified signal for the low and high NOx systems respectively. This suggests

480 many oxidation products in the high NOx experiment are unidentified.

The nitrate scheme detects many more oxidation products at higher masses including many C12 dimers. These

are more oxidised with a majority of O9-O11 species and species up to O18 observed, giving an O:C rato of ~1.0 which is much greater than detected by iodide. Accordingly, the parameterised OSc differs depending on the ionisation scheme used.

485 Overlap in detectability of the same formula compounds exists for the two instruments and is best captured for

the C6 compounds with O4-6. The cross correlated time series’ of these highly oxidised species appear quite different in temporal profile, most likely due to the method of introduction of the sample into the instrument. The faster response time of the nitrate CIMS gives a higher resolution profile from which mechanistic processes can be inferred, but the iodide is more sensitive in this mass range. Cross calibrations of the iodide species by 490 the calibrated nitrate CIMS indicates an average calibration factor of 4.11 ± 2.31 (1σ) Hz ppt-1. The high signal

detected for some C6H6O4 relative to the concentration of highly oxidised species detected by the nitrate CIMS suggests its identity is muconic acid as the iodide scheme is especially sensitive to carboxylic and di-carboxylic acids.

Formulae identified here were compared with an ambient urban iodide data set where 63 common signals 495 (adducts and deprotonated) were identified. Many less N containing compounds were identified in the ambient

data although the O:C and N:C ratios were higher than in the chamber. The ambient diurnal profiles of five

identified highly oxidised species (two of which contain N) showed a similar bimodal trend to that of NO2 with the best correlation generally occurring with a time lag of two to three hours, indicating that their production is

dependent on the presence of NOx.

500 No oxidation products with more than 6 carbon atoms that were first identified in chamber were found in ambient conditions. This suggests the propensity for dimerization of benzene oxidation products, as was observed in the chamber, is not high under ambient conditions. This may be due to the increased complexity of ambient air reducing the likelihood of the product distribution observed in the chamber from forming. Also, ambient concentrations of benzene are lower than those studied in the chamber here. These, combined with 505 different oxidant regimes between the chamber and ambient data could potentially explain the lack of ambient observation.

Acknowledgements

This work was conducted during a PhD study supported by the Natural Environment Research Council (NERC) EAO Doctoral Training Partnership and is fully-funded by NERC whose support is gratefully acknowledged 510 (Grant ref no. NE/L002469/1). Part of the research in this study was performed at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

References Archibald, A.T. et al., 2015. Long-term high frequency measurements of ethane, benzene and methyl chloride at Ragged Point, Barbados: Identification of long-range transport events. Elementa: Science of the 515 Anthropocene, 3, p.000068.

Atkinson, R., 2000. Atmospheric chemistry of VOCs and NOx. Atmospheric Environment, 34(12–14), pp.2063– 2101.

Atkinson, R. & Arey, J., 2007. Mechanisms of the gas-phase reactions of aromatic hydrocarbons and PAHs with OH and NO3 Radicals. Polycyclic Aromatic Compounds, 27(1), pp.15–40.

520 Berndt, T. et al., 2016. Hydroxyl radical-induced formation of highly oxidized organic compounds. Nature Communications, 7(May), p.13677.

Borrás, E. & Tortajada-Genaro, L.A., 2012. Secondary organic aerosol formation from the photo-oxidation of benzene. Atmospheric Environment, 47(x), pp.154–163.

Chhabra, P.S. et al., 2015. Application of high-resolution time-of-flight chemical ionization mass spectrometry 525 measurements to estimate volatility distributions of α-pinene and naphthalene oxidation products. Atmospheric Measurement Techniques, 8(1), pp.1–18.

Creasey, D.J., Heard, D.E. & Lee, J.D., 2002. Eastern Atlantic Spring Experiment 1997 (EASE97) 1. Measurements of OH and HO2 concentrations at Mace Head, Ireland. Journal of Geophysical Research: Atmospheres, 107(D10), p.ACH 3-1-ACH 3-15.

530 Department for Environment Food and Rural Affairs, 2017. National air quality objectives and European

Directive limit and target values for the protection of human health, London.

Ehn, M. et al., 2014. A large source of low-volatility secondary organic aerosol. Nature, 506(7489), pp.476– 479.

Eisele, F.L. & Tanner, D.J., 1993. Measurement of the gas phase concentration of H2SO4 and methane sulfonic 535 acid and estimates of H2SO4 production and loss in the atmosphere. Journal of Geophysical Research, 98(D5), p.9001.

Gentner, D.R. et al., 2012. Elucidating secondary organic aerosol from diesel and gasoline vehicles through detailed characterization of organic carbon emissions. Proceedings of the National Academy of Sciences, 109(45), pp.18318–18323.

540 Glowacki, D.R., Wang, L. & Pilling, M.J., 2009. Evidence of formation of bicyclic species in the early stages of atmospheric benzene oxidation. Journal of Physical Chemistry A, 113(18), pp.5385–5396.

Guo, Y. et al., 2018. Treatment of real benzene dye intermediates wastewater by the Fenton method: characteristics and multi-response optimization. RSC Advances, 8(1), pp.80–90.

Henze, D.K. et al., 2007. Global modeling of secondary organic aerosol formation from aromatic hydrocarbons: 545 high- vs low-yield pathways. Atmospheric Chemistry and Physics Discussions, 7(5), pp.14569–14601.

Hoyt, D. & Raun, L.H., 2015. Measured and estimated benzene and volatile organic carbon (VOC) emissions at a major U.S. refinery/chemical plant: Comparison and prioritization. Journal of the Air and Waste Management Association, 65(8), pp.1020–1031.

Isaacman-VanWertz, G. et al., 2017. Using advanced mass spectrometry techniques to fully characterize 550 atmospheric organic carbon: current capabilities and remaining gaps. Faraday Discuss., 200, pp.579–598.

Johnson, D. et al., 2005. Simulating the Formation of Secondary Organic Aerosol from the Photooxidation of Aromatic Hydrocarbons. Environmental Chemistry, 2(1), p.35.

Knote, C. et al., 2014. Simulation of semi-explicit mechanisms of SOA formation from glyoxal in aerosol in a 3-D model. Atmospheric Chemistry and Physics, 14(12), pp.6213–6239.

555 Kroll, J.H. et al., 2011. Carbon oxidation state as a metric for describing the chemistry of atmospheric organic aerosol. Nature Chemistry, 3(2), pp.133–139.

Mentel, T.F. et al., 2015. Formation of highly oxidized multifunctional compounds: autoxidation of peroxy radicals formed in the ozonolysis of alkenes – deduced from structure–product relationships. Atmospheric Chemistry and Physics, 15(12), pp.6745–6765.

560 Mentel, T.F. et al., 2009. Photochemical production of aerosols from real plant emissions. Atmospheric Chemistry and Physics, 9(13), pp.4387–4406.

Misztal, P.K. et al., 2015. Atmospheric benzenoid emissions from plants rival those from fossil fuels. Scientific Reports, 5(June), pp.1–10.

Molteni, U. et al., 2016. Formation of highly oxygenated organic molecules from aromatic compounds. 565 Atmospheric Chemistry and Physics Discussions, (December), pp.1–39.

Ng, N.L. et al., 2007. Secondary organic aerosol formation from m-xylene, toluene, and benzene. Atmospheric Chemistry and Physics, 7(14), pp.3909–3922.

Oyoroko, E. & Ogamba, E.N., 2017. Effects of detergent containing linear alkyl benzene sulphonate on behavioural response of Heterobranchus bidorsalis , Clarias gariepinus and Heteroclarias. 570 Biotechnological Research, 3(3), pp.59–64.

Phillips, G.J. et al., 2013. Peroxyacetyl nitrate (PAN) and peroxyacetic acid (PAA) measurements by iodide chemical ionisation mass spectrometry: first analysis of results in the boreal forest and implications for the measurement of PAN fluxes. Atmos. Chem. Phys., 13(3), pp.1129–1139.

Praske, E. et al., 2018. Atmospheric autoxidation is increasingly important in urban and suburban North 575 America. Proceedings of the National Academy of Sciences, 115(1), pp.64–69.

Priestley, M. et al., 2018. Observations of Isocyanate, Amide, Nitrate, and Nitro Compounds From an Anthropogenic Biomass Burning Event Using a ToF-CIMS. Journal of Geophysical Research: Atmospheres, (2), pp.1–26.

Rissanen, M.P. et al., 2014. The formation of highly oxidized multifunctional products in the ozonolysis of 580 cyclohexene. Journal of the American Chemical Society, 136(44), pp.15596–15606.

Rodriguez, P.G. et al., 2018. Thermochemical stability and friction properties of soft organosilica networks for solid lubrication. Materials, 11(2), pp.1–12.

Saunders, S.M. et al., 2003. Protocol for the development of the Master Chemical Mechanism, MCM v3 (Part A): tropospheric degradation of non-aromatic volatile organic compounds. Atmos. Chem. Phys., 3(1), 585 pp.161–180.

Snyder, R., Kocsis, J.J. & Drew, R., 1975. Current concepts of chronic benzene toxicity. Critical Reviews in Toxicology, 3(3), pp.265–288.

Stirnweis, L. et al., 2017. Assessing the influence of NOx concentrations and relative humidity on secondary organic aerosol yields from α-pinene photo-oxidation through smog chamber experiments and modelling 590 calculations. Atmospheric Chemistry and Physics, 17(8), pp.5035–5061.

Surratt, J.D. et al., 2006. Chemical Composition of Secondary Organic Aerosol Formed from the Photooxidation of Isoprene. The Journal of Physical Chemistry A, 110(31), pp.9665–9690.

Thomas, R. et al., 2014. Characterization of Changes in Gene Expression and Biochemical Pathways at Low Levels of Benzene Exposure S. D. Peddada, ed. PLoS ONE, 9(5), p.e91828.

595 Verma, D.K. & Des Tombe, K., 2002. Benzene in gasoline and crude oil: Occupational and environmental implications. American Industrial Hygiene Association Journal, 63(2), pp.225–230.

Volkamer, R. et al., 2002. OH-initiated oxidation of benzene. Physical Chemistry Chemical Physics, 4(9), pp.1598–1610.

Wang, S. et al., 2017. Formation of Highly Oxidized Radicals and Multifunctional Products from the 600 Atmospheric Oxidation of Alkylbenzenes. Environmental Science and Technology, 51(15), pp.8442–

8449.

Wang, T. et al., 2014. Benzene homologues in environmental matrixes from a pesticide chemical region in China: Occurrence, health risk and management. Ecotoxicology and Environmental Safety, 104(1), pp.357–364.

605 Wildt, J. et al., 2014. Suppression of new particle formation from monoterpene oxidation by NOx. Atmospheric Chemistry and Physics, 14(6), pp.2789–2804.

6. Conclusions

The iodide ToF-CIMS demonstrates capability to detect a range of air pollutants and atmospheric species of importance to air quality studies in urban environments. These include trace gases relevant to biomass burning, oxidation and SOA formation.

Measurements were made in Manchester UK over a period where typical winter conditions were sampled. This period also included a large scale anthropogenic burning event (bonfire night / Guy Fawkes Night). This ambient data set provided a good opportunity to explore the capabilities of the iodide reagent ion and the ToF mass analyser. Particular attention was given to the detection of low concentration trace gases that are found in the UK urban environment and specific to a unique biomass burning event. Further chamber studies at the Jülich plant chamber (JPAC) were used to assess the instrument’s ability to detect secondary products of an anthropogenic VOC precursor (benzene) oxidation. This chamber study was complemented by a secondary ToF-CIMS using the nitrate ionisation scheme. The ionisation scheme comparison is useful to assess the overlap in detected species within the benzene oxidation system and so contextualise the strengths and limitations of iodide ionisation regarding the detection and identification of secondary products.

Calibration Factor Calibrant Identified as m/z -1 Method (Hz ppt ) Formic acid I.HCOOH 173 6.20 Gas mixture

Acetic acid I.CH3COOH 187 2.01 Gas mixture

Propanoic acid I.CH3CH2COOH 201 1.28 Gas mixture

Butanoic acid I.CH3(CH2)2COOH 215 1.37 Gas mixture

Pentanoic acid I.CH3(CH2)3COOH 229 0.99 Gas mixture

Chlorine I.Cl2 197 4.52 Gas mixture Hypochlorous acid I.HOCl 179 0.56 Online synthesis -6 Dichloromethane I.CH2Cl2 221 7.20 x10 Gas mixture -5 Chloroform I.CHCl3 245 9.36 x10 Gas mixture

3-Chloropropionic acid I.C2H3O2Cl 221 10.42 Filter desorption

Nitryl chloride I.ClNO2 208 4.60 Online synthesis

Nitrogen pentoxide I.N2O5 235 2.70 Online synthesis Hydrogen cyanide I.HCN 154 1.93 Cross Calibration Isocyanic acid I.HNCO 170 2.65 Cross Calibration Table 4. Summary of calibration factors found for this work.

As a result of wealth of data produced by the ToF-CIMS, it was necessary to develop new peak assignment tools to speed up the identification process. This led to the development of a novel Kendrick mass defect solver to match unknown peaks within the

55 mass spectrum with those identified a posteriori. The sensitivity of the iodide ionisation scheme to inorganic chlorine trace gases is well documented, however when coupled with the ToF instrument, it was possible to fully characterise the capability of the instrument to detect both organic and inorganic chlorinated compounds under ambient UK winter conditions.

The range of different atmospheric species detected by the instrument led to the employment of many different calibration methods (table 4.) including; making gas mixtures, e.g. formic acid on a laboratory manifold; online synthesis, where the material is not readily available (e.g. HOCl); and thermal desorption, where the readily available substance is not present in the gas phase (e.g. 3-chloro propionic acid).

A series of novel CHON (carbon, hydrogen, oxygen and nitrogen containing) molecules and ClOVOCs (chlorine and oxygen containing volatile organic compounds) were identified in the ambient urban winter data set. Mixing ratios of the low molecular weight CHON compounds were found to enhance by 2-13 times when sampling the anthropogenic biomass burning plume formed on bonfire night. This represents the first online measurements of trace gases from bonfire night using a research instrument, i.e. not operational monitoring as found in the UK networks e.g. AURN. Building on previous work (Le Breton et al. 2013), biomass burning markers HCN and HNCO were both detected by the ToF-CIMS and were used to define the plume sampling period. Using the HNCO/CO ratio it was possible to distinguish flaming and burning phases of the event (Roberts et al. 2011) and so calculate burning phase specific normalised excess mixing ratios (NEMRS) for the identified amides, HCN and HNCO. Analysis of the routinely monitored air pollutants at a co-located monitoring station showed that within a year and a half period, bonfire night 2014 saw the greatest enhancements in CO mixing ratios, whereas enhanced NO2 concentrations had been greater at other times.

As well as the ClOVOCs, inorganic chlorine compounds ClNO2, Cl2 and HOCl were also detected. This represents the only measurements of urban Cl2 and HOCl in the UK, and contributes to the small number of UK ClNO2 observations. The diurnal behaviour of

ClNO2 was typical in that night time maxima were observed, but concentrations were consistently lower than their peak values by the early morning when photolysis begins.

Cl2 concentrations were highest at mid-day when photolysis was at a maximum which is consistent with other studies (e.g. Liu et al. 2017). This suggests Cl2 is formed via a photochemical mechanism rather than directly emitted. The diurnal behaviour of the

ClOVOCs varies; some exhibit photochemical behaviour consistent with Cl2 but others correlate well with NO2 suggesting the presence of NOx influences their formation.

The impact of the newly identified organic species on chlorine radical production was compared with the inorganic chlorine measurements on days with high and low incoming short wave radiation flux. In order to produce a more accurate production rate, idealised photolysis rates taken from a TUV model were scaled by the observed UV-A transmittance, measured at the University of Manchester by a Brewer photo- spectrometer, to replicate attenuation by clouds and aerosol. The inorganic Cl species were found to contribute the vast majority (97%) to steady state Cl concentrations, but varied depending on environmental conditions.

The production of the Cl2 intermediate and its photolysis to form Cl are highly dependent on the short wave radiant flux contributing 74% of Cl on a high flux day but only 14% on a low flux day. The change in steady state Cl production from ClNO2 is relatively insensitive to the change in short wave radiation flux as the formation of ClNO2 is not dependent on a photochemical mechanism. The concentrations of ClNO2 remains relatively unchanged between the high and low flux days and so the ClNO2 source term remained fairly similar. However, the sensitivity of steady state Cl production from Cl2 photolysis is highly variable as enhanced concentrations of Cl2 (up to a maximum of 16 ppt) are only observed on high flux days. The behaviour of ClOVOCs and HOCl was more similar to that of Cl2 in terms of production; however their lower source concentrations and lower photolysis rates mean they contribute ~3% each to the budget on either day.

The increased mass range and broad sensitivity of the iodide ToF-CIMS make it an ideal instrument to probe the matrix of oxygenated organic gas phase species produced from VOC oxidation. Benzene is one such anthropogenic VOC whose oxidation is relatively well understood. The iodide ToF-CIMS was deployed alongside a nitrate ToF-CIMS in a chamber study in order to assess the ability of the iodide ionisation scheme to detect benzene oxidation products. The nitrate ionisation scheme is typically used in studies to investigate the formation of highly oxidised molecules (HOM) and dimers as it is sensitive to molecules of a high mass range (e.g. m/z 300-500) and high O:C ratio (e.g. O:C > 1.0). This was found to be the case for the benzene system, although overlap between the two instruments does exist, typically with compounds containing 6 carbon atoms and with 4 to 6 oxygen atoms. The detection of molecules that contained 8 oxygen atoms and greater is significantly reduced for the iodide ToF-CIMS. For the same number of carbon atoms per molecule, the nitrate scheme observes a parameterised average organic carbon oxidation state of 2 whereas iodide observes only 1. The selectivity biases of the ionisation schemes mean neither one can be used in isolation to describe bulk properties of all VOC oxidation products. The nitrate scheme is ideal to

57 investigate highly oxidised species, whereas the iodide scheme is best used to explore lower oxygen content products, such as those encountered early in the oxidation process.

The influence of NOx on the product distribution of the benzene OH system was investigated. The low NOx conditions favour the formation of high carbon number molecules (6 or more atoms) that contain 1 to 4 oxygen atoms, consistent with condensation reactions expected under these conditions. The presence of NOx increased the number of fragmented oxidation products with less than 6 carbon atoms, but higher nitrogen and oxygen content (5-7 atoms). High oxygen content is likely associated with the nitrogen in the form of nitrates (e.g. NO3, NO5). Whilst it is not possible to unequivocally identify the exact species detected here, the variation of

NO/NOx fraction produced different product distributions, suggesting the formation of a variety of different N containing CHON compounds. 63 species observed in the chamber were identified in the Manchester winter ambient data, 11 of which contained nitrogen. These 11 CHON compounds all contained six carbon atoms and had high O:C (1.5 - 2.0) and high N:C (0.3 - 0.5) ratios. CHON compounds identified in the chamber that do not meet these criteria were not observed in the ambient data. This suggests dimerization of potential benzene oxidation products and other compounds with: higher carbon numbers (i.e. more than 6 carbon atoms), low O:C (0.5 - 1.0) and low N:C (0.1 - 0.2) are not likely to form under ambient conditions. This may be due to the lower ambient benzene concentration, more complex VOC mixture competing with reaction products or a different oxidant regime controlling oxidation.

6.1. Future work There are few laboratory studies of anthropogenic waste burning, which is a substantial yet poorly understood emission source, especially in developing countries. This is in part due to the complex nature of waste as a fuel source, which is likely to vary not only around the world, but potentially between cities within the same country and even smaller spatial scales. When natural biomass is burned, the nitrogen content of the emitted VOCs e.g. HCN, is dependent on the nitrogen content of the fuel (Coggon et al. 2016). The nitrogen content of man-made material such as plastics can vary dramatically e.g. poly-ethylene (PET), poly-propylene (PP) and polyvinylchloride (PVC) are all common domestic plastics devoid of nitrogen, however nylons and other polyamides have high nitrogen contents. A better constraint on anthropogenic waste composition would indicate if it is likely a lower NEMR for e.g. HCN:CO is more typical for anthropogenic waste burning. Additionally, it was not possible to de-convolve the influence of emissions from fireworks from the emissions from bonfires. Future work

58 should aim to definitively separate the effects of these two phenomena on the atmospheric composition. This may give a relative importance of each activity or the mixing of two sources for health impact studies e.g. do the metals known to be emitted from fireworks mix with the enhanced primary and secondary aerosols? Does that facilitate the formation of novel products via aqueous chemistry? Does that increase the aerosol toxicity? Could this be relevant where burned waste contains high metal content?

Unfortunately without further investigation, it was not possible to unequivocally state the origin of the urban Cl sources identified in the urban winter dataset. ClNO2 is known to have both marine and anthropogenic origins having been measured inland where marine air mass influence is minimal. It is uncertain in this case whether urban sources of particulate Cl e.g. from road salt, as has been recently suggested, had any influence here. It is also unclear from where the night time source of Cl2 originated. More relevant for the chlorine radical budget, more work is required to understand the source of day time Cl2, including its formation mechanism, the identity of its precursors and their origins. Within the urban environment it may be possible that Cl2 photolysis varies as a function of shading from buildings and other urban structures and so the relative importance of the Cl radical to the oxidant budget may vary on a scale of 10s of metres. The calculation of steady state chlorine used the available Cl precursors measured by the ToF-CIMS however this did not include HCl. Measurements of this key Cl species alongside those made here would further constrain the steady state Cl radical concentration, against which its relative importance as an oxidising agent could be properly compared with OH. Also, as this study took place during the winter, a comparison with summer time would be prudent. This would indicate how much greater the production and photolysis of Cl2 influences the Cl radical budget, under conditions when photolysis is faster and sustained over a longer period.

The importance of atmospheric oxidants is ultimately due to how they control the oxidation of VOCs. Relevant oxidants beyond OH, such as O3, NO3 and Cl should be investigated for their impacts on anthropogenic VOC oxidation, e.g. benzene. This is particularly relevant in an urban setting where concentrations of these oxidants and VOCs are likely to be high. Other aromatic compounds such as ethyl benzene, toluene and xylenes, are more reactive than benzene and so similar analysis of those species, should also take place.

Regarding the oxidation of benzene by OH, further work should investigate why higher mass, low O:C ratio, low N:C ratio compounds were not detected in the ambient data. Potential explanations to explore include the ambient benzene concentrations being

59 lower than chamber concentrations, or the increased instances of competing reactions of benzene oxidation products with other VOC oxidation products. Repeated chamber experiments with the introduction of a second VOC, e.g. 1-3 butadiene, another common anthropogenic UK VOC, may form condensation products that are detectable in the ambient data.

The utility of the CIMS is well demonstrated for the study of a range of relevant chemical species such as biomass burning markers, halogenated species and oxidised material. How concentrations of these vary spatially is one such question that can only be explored by deploying the instrument on a moving platform such as an aircraft. Quadrupole CIMS have been deployed on various aircraft such as the UK BAe146 (Le Breton et al. 2012), the NASA DC8 (Crounse et al. 2006) and NOAA P3 (Zheng et al. 2011) and more recently the ToF-CIMS has been deployed on US aircraft (NOAA WP- 3D, Lee et al. 2014). The deployment of the ToF-CIMS on the UK BAe146 would provide the ability to study the effects of urban and anthropogenic activity. This could include the composition of urban and industrial outflows and the oxidative transformations of air masses at altitude. The multiplex advantage and increased mass range of the ToF-CIMS in comparison to quadrupole based systems allows for greater overlap with instruments already on board the BAe146 such as the GC-MS (gas chromatography mass spectrometer); and with the FIGAERO inlet, the AMS (aerosol mass spectrometer).

The utility of the ToF-CIMS is further enhanced by the addition of the Filter Inlet for Gas and AEROsol (FIGAERO) (Lopez-Hilfiker et al. 2014). This multi-channel inlet collects aerosol on a filter whist the instrument is sampling in gas mode. Periodically, the gas sampling is stopped and the filter is heated to ~200oC to thermally desorb the collected aerosol. The constituents of the aerosol are then measured as gas phase species if the instrument is sensitive to them. This development allows for the near simultaneous detection of organic species in the gas and particle phases, allowing for the study of phenomena such as phase partitioning. The major downfall of the FIGAERO is the thermal desorption can decompose organic molecules so their fragments are detected rather than the original structure. This restricts the detection of high molecular mass species in the particle phase. This draw back led to the new development of the Extractive Electrospray Ionisation (EESI) source (Lopez-Hilfiker et al. 2017) which can use positive and negative ion modes to selectively detect organic aerosol constituents. The deployment of the FIGAERO and EESI inlets would provide yet more chemical detail of the constituents of organic aerosol and allow for the investigation of gas/particle partitioning and the ambient conditions that lead to differences in the composition of

60 each phase. The detailed chemical composition of the aerosol phase is useful to inform atmospheric aerosol model development and validate model outputs.

Issues surrounding the comprehensive calibration of species the ToF-CIMS can detect will be one of the biggest challenges for its future use. A large and comprehensive set of work to understand the effects of different functional groups (e.g. halogenated, ONO, OOH), carbon number and isomerism on organic sensitivities would be invaluable in order to better understand and predict instrument response to different molecular parameters. This work would require stronger collaboration with synthetic organic chemists who can provide the material necessary that cannot be purchased from an industrial supplier.

As previously discussed, the inability to separate different species with the same unit mass e.g. CH2O2 and NO2 is a disadvantage of the H-ToF-CIMS. However, with increased resolution, the separation of masses within the spectrum becomes possible. The development of the Aerodyne L-ToF-CIMS extends the capabilities of the ToF-CIMS with a ToF drift tube of increased length and hence increased resolution to m/dm ≤ 8,000 (Aerodyne Research Inc., 2018). The deployment of such an instrument alongside the ToF-CIMS may provide greater insight into its limitations due to interfering signals in the mass spectrum.

One such analysis that has yet to be fully realised with iodide ToF-CIMS is the study of urban fluxes. Quadrupole iodide CIMS has been used to quantify NOy fluxes from biogenic sources in forests (Turnipseed et al. 2006) and the PTR-ToF-MS has been used to study biogenic emission of methanol in a rural environment (Müller et al. 2010), but studies of urban fluxes using the iodide ToF-CIMS have yet to be made. Flux measurements could be performed for trace gases such as; HONO, whose flux is important for the quantification of oxidant budgets; oxygenated VOCs, such as organic acids; and biomass burning emissions from urban burning. This analysis is potentially useful for the quantification of the constituents of the nocturnal residual layer that would better define its contribution to the next day’s chemistry.

The ToF-CIMS is demonstrably capable of measuring many different species that are important to the study of air quality. The comparison of different urban centres around the world, particularly those with high populations that are known to suffer from very poor air quality such as Beijing would be a clear next step. Using the chemical markers identified here would be useful to contextualise the air quality problem compared with cities in the UK and Europe and better quantify the differences in contributions from different VOCs and oxidant conditions.

As poor urban air quality is one of the greatest public health burdens on populations worldwide, the large scale and continuous monitoring of air pollutants is recognised as necessary to inform and audit mitigation strategies. The concept of the smart city blends urban planning and networked device integration (e.g. the Internet Of Things, IOT) including air quality sensors, that can be deployed at a much finer resolution than current air quality monitoring stations. Currently, much research is focused on the accuracy, precision and reproducibility of low cost sensors that could effectively be distributed in such a network (Lewis et al. 2016). Whilst environmental variables such as temperature and relative humidity can interfere with the resultant measurements, the fact that low cost sensors often rely on non-discriminatory detection methods, such as the electro chemical cell, means other non-targeted air pollutants can be the source of interference (Mead et al. 2013). The identity and origin of these interference species need to be investigated in order to improve the efficacy of these low cost sensors e.g. improve their limit of detection. These sensors also require benchmarking against established techniques such as the ToF-CIMS e.g. to improve reproducibility between instruments.

Bibliography Aerodyne Research, inc., 2011. Time-of-Flight Chemical-Ionization Mass Spectrometer, BILLERICA, MA. Ahlberg, E. et al., 2017. Secondary organic aerosol from VOC mixtures in an oxidation flow reactor. Atmospheric Environment, 161, pp.210–220. Aljawhary, D., Lee, A.K.Y. & Abbatt, J.P.D., 2013. High-resolution chemical ionization mass spectrometry (ToF-CIMS): application to study SOA composition and processing. Atmos. Meas. Tech., 6(11), pp.3211–3224. Aouizerats, B. et al., 2015. Importance of transboundary transport of biomass burning emissions to regional air quality in Southeast Asia during a high fire event. Atmospheric Chemistry and Physics, 15(1), pp.363–373. Atkinson, R., 2000. Atmospheric chemistry of VOCs and NOx. Atmospheric Environment, 34(12–14), pp.2063–2101. Bannan, T.J., Bacak, A., et al., 2017. Ground and Airborne U.K. Measurements of Nitryl Chloride: An Investigation of the Role of Cl Atom Oxidation at Weybourne Atmospheric Observatory. Journal of Geophysical Research: Atmospheres, 122(20), p.11,154-11,165. Bannan, T.J. et al., 2014. Importance of direct anthropogenic emissions of formic acid measured by a chemical ionisation mass spectrometer (CIMS) during the Winter ClearfLo Campaign in London, January 2012. Atmospheric Environment, 83, pp.301–310. Bannan, T.J., Murray Booth, A., et al., 2017. Seasonality of Formic Acid (HCOOH) in London during the ClearfLo Campaign. Journal of Geophysical Research: Atmospheres, pp.488–498. Bannan, T.J. et al., 2015. The first UK measurements of nitryl chloride using a chemical ionization mass spectrometer in central London in the summer of 2012, and an investigation of the role of Cl atom oxidation. Journal of Geophysical Research, 120(11), pp.5638–5657. Bertram, T.H. et al., 2011. A field-deployable, chemical ionization time-of-flight mass spectrometer. Atmospheric Measurement Techniques, 4(7), pp.1471–1479. Bianchi, F. et al., 2016. New particle formation in the free troposphere: A question of chemistry and timing. Science, 352(6289), pp.1109–1112. Bianchi, F. et al., 2017. The role of highly oxygenated molecules (HOMs) in determining the composition of ambient ions in the boreal forest. Atmospheric Chemistry and Physics, 17(22), pp.13819–13831. Borbon, A. et al., 2013. Emission ratios of anthropogenic volatile organic compounds in northern mid-latitude megacities: Observations versus emission inventories in Los Angeles and Paris. Journal of Geophysical Research: Atmospheres, 118(4), pp.2041–2057. Le Breton, M., Bacak, A., Muller, J.B.A., et al., 2013. Airborne hydrogen cyanide measurements using a chemical ionisation mass spectrometer for the plume identification of biomass burning forest fires. Atmospheric Chemistry and Physics, 13(18), pp.9217–9232. Le Breton, M., Bacak, A., Muller, J.B.A., et al., 2013. Airborne hydrogen cyanide measurements using a chemical ionisation mass spectrometer for the plume identification of biomass burning forest fires corrigendum. Atmospheric Chemistry and Physics, 13(18), pp.9217–9232.

Le Breton, M. et al., 2012. Airborne observations of formic acid using a chemical ionization mass spectrometer. Atmospheric Measurement Techniques, 5(12), pp.3029–3039. Le Breton, M. et al., 2018. Chlorine oxidation of VOCs at a semi-rural site in Beijing: Significant chlorine liberation from ClNO2 and subsequent gas and particle phase Cl-VOC production. Atmospheric Chemistry and Physics Discussions, (January), pp.1–25. Le Breton, M. et al., 2017. Enhanced ozone loss by active inorganic bromine chemistry in the tropical troposphere. Atmospheric Environment, 155, pp.21–28. Le Breton, M. et al., 2014. Simultaneous airborne nitric acid and formic acid measurements using a chemical ionization mass spectrometer around the UK: Analysis of primary and secondary production pathways. Atmospheric Environment, 83(3), pp.166–175. Le Breton, M. et al., 2014. The first airborne comparison of N2O5 measurements over the UK using a CIMS and BBCEAS during the RONOCO campaign. Anal. Methods, 6(24), pp.9731–9743. Brilli, F. et al., 2014. Proton Transfer Reaction Time-of-Flight Mass Spectrometric (PTR- TOF-MS) determination of volatile organic compounds (VOCs) emitted from a biomass fire developed under stable nocturnal conditions. Atmospheric Environment, 97, pp.54–67. Brophy, P. & Farmer, D.K., 2015. A switchable reagent ion high resolution time-of-flight chemical ionization mass spectrometer for real-time measurement of gas phase oxidized species: characterization from the 2013 southern oxidant and aerosol study. Atmospheric Measurement Techniques, 8(7), pp.2945–2959. Buffaloe, G.M. et al., 2014. Black carbon emissions from in-use ships: A California regional assessment. Atmospheric Chemistry and Physics, 14(4), pp.1881–1896. Carslaw, D.C. et al., 2016. Have vehicle emissions of primary NO2 peaked? Faraday Discuss., 189(0), pp.439–454. Christian, T.J. et al., 2009. Trace gas and particle emissions from domestic and industrial biofuel use and garbage burning in central Mexico. Atmospheric Chemistry and Physics, 10(2), pp.565–584. Coggon, M.M. et al., 2016. Emissions of nitrogen-containing organic compounds from the burning of herbaceous and arboraceous biomass: Fuel composition dependence and the variability of commonly used nitrile tracers. Geophysical Research Letters, 43(18), pp.9903–9912. Council of the European Union, 1996. Council Directive 96/62/EC on ambient air quality assessment and management, Brussels. Cox, M., Sturrock, G. & Fraser, P., 2003. Regional sources of methyl chloride, chloroform and dichloromethane identified from AGAGE observations at Cape Grim, Tasmania, 1998–2000. Journal of Atmospheric Chemistry, 45, pp.79–99. Creasey, D.J. et al., 1997. Implementation and initial deployment of a field instrument for measurement of OH and HO2 in the troposphere by laser-induced fluorescence. J. Chem. Soc, 93(16), pp.2907–2913. Criegee, R., 1975. Mechanism of Ozonolysis. Angewandte Chemie International Edition in English, 14(11), pp.745–752. Crounse, J.D. et al., 2006. Measurement of gas-phase hydroperoxides by chemical ionization mass spectrometry. Analytical Chemistry, 78(19), pp.6726–6732.

Crounse, J.D. et al., 2011. Peroxy radical isomerization in the oxidation of isoprene. Physical Chemistry Chemical Physics, 13(30), p.13607. D’Silva, T.D. et al., 1986. Studies of Methyl Isocyanate Chemistry in the Bhopal incident. Journal of Organic Chemistry, 51(20), pp.3781–3788. Department for Energy and Climate Change, 2016. Summary results of the domestic wood use survey, London. Department for Environment Food & Rural Affairs, 2018. National Statistics Release: Air quality statistics in the UK 1987 to 2017, London. Department for Environment Food & Rural Affairs, 2014. National Statistics Release: Emissions of Air Pollutants in the UK, 1970 to 2013, London. Department for Environment Food & Rural Affairs, 2016. National Statistics Release: Emissions of Air Pollutants in the UK, 1970 to 2016, London. Department for Environment Food & Rural Affairs, 2002. The National Emission Ceilings Regulations 2002, London: The Stationery Office. Department for Environment Food & Rural Affairs, 2013. UK Hydrocarbon Monitoring Network. Department for environment food and rural affairs, Didcot. Desai, M.A., Mehta, S. & Smith, K.R., 2004. Indoor smoke from solid fuels: Assessing the environmental burden of disease at national and local levels. Environmental Burden of Disease Series, (4), p.82. Dunmore, R.E. et al., 2015. Diesel-related hydrocarbons can dominate gas phase reactive carbon in megacities. Atmospheric Chemistry and Physics, 15(17), pp.9983–9996. Dusek, U. et al., 2006. Size Matters More Than Chemistry for Cloud-Nucleating Ability of Aerosol Particles. Science, 312(5778), pp.1375–1378. Dyke, P., Coleman, P. & James, R., 1997. Dioxins in ambient air, bonfire night 1994. Chemosphere, 34(5–7), pp.1191–1201. Ehn, M. et al., 2014. A large source of low-volatility secondary organic aerosol. Nature, 506(7489), pp.476–479. Eisele, F.L. & Tanner, D.J., 1993. Measurement of the gas phase concentration of H2SO4 and methane sulfonic acid and estimates of H2SO4 production and loss in the atmosphere. Journal of Geophysical Research, 98(D5), p.9001. European Commission, 1970. Directive 70/220/EEC on the Approximation of the Laws of the Member States Relating to Measures to be Taken Against Air Pollution by Gases from Positive-Ignition Engines of Motor Vehicles, Brussels. European Parliament & Council of the European Union, 2008. Directive 2008/50/EC of the European Parliament and of the Council of 21 May 2008 on ambient air quality and cleaner air for Europe. Official Journal of the European Communities, pp.1–43. Falasca, S., Moroni, M. & Cenedese, A., 2013. Laboratory simulations of an urban heat island in a stratified atmospheric boundary layer. Journal of Visualization, 16(1), pp.39–45. Fenn, J. et al., 1989. Electrospray ionization for mass spectrometry of large biomolecules. Science, 246(4926), pp.64–71. Fiore, A.M. et al., 2012. Global air quality and climate. Chemical Society Reviews, 41(19), pp.6663–6683. Font, A. et al., 2015. Using metal ratios to detect emissions from municipal waste incinerators in ambient air pollution data. Atmospheric Environment, 113, pp.177–

186. France, O.B. et al., 2013. Clouds and aerosols. Climate Change 2013 the Physical Science Basis: Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, 9781107057, pp.571–658. Gehring, U. et al., 2013. Air pollution exposure and lung function in children: The ESCAPE project. Environmental Health Perspectives, 121(11–12), pp.1357–1364. Ghernaout, D. & Ghernaout, B., 2010. From chemical disinfection to electrodisinfection: The obligatory itinerary? Desalination and Water Treatment, 16(1–3), pp.156–175. Godschalk, R.W.L. et al., 2000. Exposure-route-dependent DNA adduct formation by polycyclic aromatic hydrocarbons. Carcinogenesis, 21(1), pp.87–92. de Gouw, J. & Warneke, C., 2007. Measurements of volatile organic compounds in the earth’s atmosphere using proton-transfer-reaction mass spectrometry. Mass spectrometry reviews, 26(2), pp.223–257. Guenther, A. et al., 1995. A global model of natural volatile organic compound emissions. Journal of Geophysical Research: Atmospheres, 100(D5), pp.8873– 8892. Guenther, A.B. et al., 2012. The Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGAN2.1): an extended and updated framework for modeling biogenic emissions. Geosci. Model Dev., 5(6), pp.1471–1492. Guttikunda, S.K. et al., 2005. Impacts of Asian megacity emissions on regional air quality during spring 2001. Journal of Geophysical Research D: Atmospheres, 110(20), pp.1–18. Haeger-Eugensson, M. & Holmer, B., 1999. Advection caused by the urban heat island circulation as a regulating factor on the nocturnal urban heat island. International Journal of Climatology, 19(9), pp.975–988. Hansel, A. et al., 1999. Improved Detection Limit of the Proton-transfer-reaction mass spectrometer: On-line monitoring of volatile organic compounds at volume mixing ratios of a few pptv. Plasma Sources Science and Technology, 8, pp.332–336. Harrison, T. & Shallcross, D.E., 2011. Smoke is in the air : how fireworks affect air quality Did you realise that fireworks cause measurable air pollution? Science in School, (21), pp.47–51. Health and Safety Executive, 2011. EH4O / 2005 Workplace exposure limits EH4O / 2005 Workplace exposure limits, London. Health Effects Institute, 2018. State of Global Air 2018, Henderson, S.B. & Johnston, F.H., 2012. Measures of forest fire smoke exposure and their associations with respiratory health outcomes. Current Opinion in Allergy and Clinical Immunology, 12(3), pp.221–227. Hewson, W. et al., 2013. Characterisation of GOME-2 formaldehyde retrieval sensitivity. Atmospheric Measurement Techniques, 6(2), pp.371–386. Hidalgo, J. et al., 2008. Advances in urban climate modeling. Annals of the New York Academy of Sciences, 1146(2008), pp.354–374. Hilboll, A., Richter, A. & Burrows, J.P., 2013. Long-term changes of tropospheric NO2 over megacities derived from multiple satellite instruments. Atmospheric Chemistry and Physics, 13(8), pp.4145–4169. Hinkel, K.M. & Nelson, F.E., 2007. Anthropogenic heat island at Barrow, Alaska, during winter: 2001–2005. Journal of Geophysical Research: Atmospheres, 112(D6),

p.D06118. Hirokawa, J., Kato, T. & Mafune, F., 2009. In Situ Measurements of Atmospheric Nitrous Acid by Chemical Ionization Mass Spectrometry Using Chloride Ion Transfer Reactions. Analytical Chemistry, 81(20), pp.8380–8386. Hoffman, R.C. et al., 2003. Knudsen cell studies of the reactions of N2O5 and ClONO2 with NaCl: development and application of a model for estimating available surface areas and corrected uptake coefficients. Physical Chemistry Chemical Physics, 5(9), pp.1780–1789. Hoker, J. et al., 2015. Comparison of GC/time-of-flight MS with GC/quadrupole MS for halocarbon trace gas analysis. Atmos. Meas. Tech., 8(5), pp.2195–2206. Hu, J. et al., 2017. Modeling biogenic and anthropogenic secondary organic aerosol in China. Atmospheric Chemistry and Physics, 17(1), pp.77–92. Huang, G. et al., 2017. Speciation of anthropogenic emissions of non-methane volatile organic compounds: A global gridded data set for 1970-2012. Atmospheric Chemistry and Physics, 17(12), pp.7683–7701. Hughey, C. a. et al., 2001. Kendrick mass defect spectrum: A compact visual analysis for ultrahigh-resolution broadband mass spectra. Analytical Chemistry, 73(19), pp.4676–4681. Ilyinskaya, E. et al., 2017. Understanding the environmental impacts of large fissure eruptions: Aerosol and gas emissions from the 2014–2015 Holuhraun eruption (Iceland). Earth and Planetary Science Letters, 472, pp.309–322. Iyer, S., Lopez-Hilfiker, F., Lee, B.H., et al., 2016. Modeling the Detection of Organic and Inorganic Compounds Using Iodide-Based Chemical Ionization. The Journal of Physical Chemistry A. Iyer, S., Lopez-Hilfiker, F., Lee, B.H., et al., 2016. Modeling the Detection of Organic and Inorganic Compounds Using Iodide-Based Chemical Ionization. The Journal of Physical Chemistry A, p.acs.jpca.5b09837. Jones, B.T. et al., 2014. Airborne measurements of HC(O)OH in the European Arctic: A winter – summer comparison. Atmospheric Environment, 99, pp.556–567. Jung, C.-R., Lin, Y.-T. & Hwang, B.-F., 2015. Ozone, particulate matter, and newly diagnosed Alzheimer’s disease: a population-based cohort study in Taiwan. Journal of Alzheimer’s disease : JAD, 44(2), pp.573–84. Junninen, H. et al., 2010. A high-resolution mass spectrometer to measure atmospheric ion composition. Atmos. Meas. Tech., 3(4), pp.1039–1053. Kebabian, P.L. et al., 2008. A Practical Alternative to Chemiluminescence-Based Detection of Nitrogen Dioxide: Cavity Attenuated Phase Shift Spectroscopy. Environmental Science & Technology, 42(16), pp.6040–6045. Kercher, J.P., Riedel, T.P. & Thornton, J.A., 2009. Chlorine activation by N2O5: simultaneous, in situ detection of ClNO2 and N2O5 by chemical ionization mass spectrometry. Atmos. Meas. Tech., 2(1), pp.193–204. Kiendler-Scharr, A. et al., 2009. Aerosol mass spectrometric features of biogenic SOA: Observations from a plant chamber and in rural atmospheric environments. Environmental Science and Technology, 43(21), pp.8166–8172. Kim, P.S. et al., 2015. Sources, seasonality, and trends of southeast US aerosol: an integrated analysis of surface, aircraft, and satellite observations with the GEOS- Chem chemical transport model. Atmospheric Chemistry and Physics, 15(18), pp.10411–10433.

Kim, S. et al., 2007. Measurement of HO2 NO2 in the free troposphere during the Intercontinental Chemical Transport Experiment–North America 2004. Journal of Geophysical Research, 112(D12), p.D12S01. Kita, K. et al., 2006. A Chemical Ionization Mass Spectrometer for Ground-Based Measurements of Nitric Acid. Journal of Atmospheric and Oceanic Technology, 23(8), pp.1104–1113. Koss, A.R. et al., 2017. Non-methane organic gas emissions from biomass burning: identification, quantification, and emission factors from PTR-ToF during the FIREX 2016 laboratory experiment. Atmospheric Chemistry and Physics Discussions, (October), pp.1–44. Kroll, J.H. & Seinfeld, J.H., 2008. Chemistry of secondary organic aerosol: Formation and evolution of low-volatility organics in the atmosphere. Atmospheric Environment, 42(16), pp.3593–3624. Kunii, O. et al., 2002. The 1997 haze disaster in indonesia: Its air quality and health effects. Archives of Environmental Health, 57(1), pp.16–22. Kürten, A. et al., 2012. Calibration of a chemical ionization mass spectrometer for the measurement of gaseous sulfuric acid. Journal of Physical Chemistry A, 116(24), pp.6375–6386. Leather, K.E. et al., 2012. Temperature and pressure dependence of the rate coefficient for the reaction between ClO and CH3O2 in the gas-phase. Physical Chemistry Chemical Physics, 14, pp.3425–3434. Lee, B.H., Lopez-Hilfiker, F.D., Mohr, C., Kurten, T., Worsnop, D.R. & Thornton, J.A., 2014. An iodide-adduct high-resolution time-of-flight chemical-ionization mass spectrometer: Application to atmospheric inorganic and organic compounds. Environmental Science & Technology, 48(11), pp.6309–6317. Lee, B.H., Lopez-Hilfiker, F.D., Mohr, C., Kurten, T., Worsnop, D.R., Thornton, J.A., et al., 2014. An Iodide-Adduct High-Resolution Time-of-Flight Chemical-Ionization Mass Spectrometer: Application to Atmospheric Inorganic and Organic Compounds. Environmental Science & Technology, 48(11), pp.6309–6317. Lee, J.D. et al., 2016. Detailed budget analysis of HONO in central London reveals a missing daytime source. Atmospheric Chemistry and Physics, 16(5), pp.2747–2764. Leggett, S., 1996. Forecast distributions of species and their atmospheric reactivities for the U.K. VOC emission inventory. Atmospheric Environment, 30(2), pp.215–226. Lemieux, P.M., Lutes, C.C. & Santoianni, D.A., 2004. Emissions of organic air toxics from open burning: a comprehensive review. Progress in Energy and Combustion Science, 30(1), pp.1–32. Lewis, A.C. et al., 2016. Evaluating the performance of low cost chemical sensors for air pollution research. Faraday Discussions, 189, pp.85–103. Lin, G. et al., 2014. Radiative forcing of organic aerosol in the atmosphere and on snow: Effects of SOA and brown carbon. Journal of Geophysical Research: Atmospheres RESEARCH, pp.9567–9577. Lin, Y.C. et al., 2012. Summertime formaldehyde observations in New York City: Ambient levels, sources and its contribution to HOx radicals. Journal of Geophysical Research Atmospheres, 117(8), pp.1–14. Liu, X. et al., 2017. High Levels of Daytime Molecular Chlorine and Nitryl Chloride at a Rural Site on the North China Plain. Environmental Science and Technology, 51(17), pp.9588–9595.

Lobert, J.M. et al., 1999. Global chlorine emissions from biomass burning: Reactive Chlorine Emissions Inventory. Journal of Geophysical Research Atmospheres, 104(D7), pp.8373–8389. Lohmann, R., Northcott, G.L. & Jones, K.C., 2000. Assessing the Contribution of Diffuse Domestic Burning as a Source of PCDD/Fs, PCBs, and PAHs to the U.K. Atmosphere. Environ. Sci. Technol., 34(14), pp.2892–2899. Lopez-Hilfiker, F. et al., 2017. Charaterization of a New Extractive Electrospray Ionization (EESI) Source for Online Particle Analysis and its Application to Atmospheric Aersosol. In Deutsche Gesellschaft für Massenspektrometrie (DGMS). Lopez-Hilfiker, F.D. et al., 2014. A novel method for online analysis of gas and particle composition: description and evaluation of a Filter Inlet for Gases and AEROsols (FIGAERO). Atmospheric Measurement Techniques, 7(4), pp.983–1001. Lu, J. et al., 2015. Improved Peak Detection and Deconvolution of Native Electrospray Mass Spectra from Large Protein Complexes. Journal of the American Society for Mass Spectrometry, 26(12), pp.2141–2151. Marshall, A.G. & Rodgers, R.P., 2004. Petroleomics : The Next Grand Challenge for Chemical Analysis Petroleum and Mass Spectrometry : Divergent. Accounts of Chemical Research, 37, pp.53–59. Mead, M.I. et al., 2013. The use of electrochemical sensors for monitoring urban air quality in low-cost , high-density networks. Atmospheric Environment, 70, pp.186– 203. Mentel, T.F. et al., 2015. Formation of highly oxidized multifunctional compounds: autoxidation of peroxy radicals formed in the ozonolysis of alkenes – deduced from structure–product relationships. Atmospheric Chemistry and Physics, 15(12), pp.6745–6765. Mielke, L.H., Furgeson, A. & Osthoff, H.D., 2011. Observation of ClNO2 in a mid- continental urban environment. Environmental science & technology, 45, pp.8889– 96. Molteni, U. et al., 2016. Formation of highly oxygenated organic molecules from aromatic compounds. Atmospheric Chemistry and Physics Discussions, (December), pp.1– 39. Monks, P.S. et al., 2015. Tropospheric ozone and its precursors from the urban to the global scale from air quality to short-lived climate forcer. Atmospheric Chemistry and Physics, 15(15), pp.8889–8973. Muller, J.B.A. et al., 2018. A novel semi-direct method to measure OH reactivity by chemical ionisation mass spectrometry (CIMS). Atmospheric Measurement Techniques Discussions, (April), pp.1–41. Müller, M. et al., 2010. First eddy covariance flux measurements by PTR-TOF. Atmospheric Measurement Techniques, 3(2), pp.387–395. Myhre, G. et al., 2013. Anthropogenic and Natural Radiative Forcing. Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, pp.659– 740. National Atmospheric Emissions Inventory, 2015. Emission Map Data for Non Methane VOC in 2015. Available at: http://naei.beis.gov.uk/data/map-uk- das?pollutant_id=9&emiss_maps_submit=naei-20180609142506 [Accessed June 1, 2018]. National Atmospheric Emissions Inventory, 2016. Pollutant Information: Non Methane

VOC. Available at: http://naei.beis.gov.uk/overview/pollutants?pollutant_id=9 [Accessed June 1, 2018]. New Scientist, 2018. London has already reached air pollution limits for 2018. New Scientist. Available at: https://www.newscientist.com/article/2159875-london-has- already-reached-air-pollution-limits-for-2018/ [Accessed May 4, 2018]. NikNaks, 2012. The Atmospheric Boundary Layer. Wikimedia Commons. Available at: https://commons.wikimedia.org/wiki/File:Atmospheric_boundary_layer.svg [Accessed June 10, 2018]. Nowak, J.B. et al., 2007. A chemical ionization mass spectrometry technique for airborne measurements of ammonia. Journal of Geophysical Research: Atmospheres (1984–2012), 112(D10). Oke, T.R., 1976. The distinction between canopy and boundary‐layer urban heat islands. Atmosphere, 14(4), pp.268–277. Orlando, J.J., Tyndall, G.S. & Wallington, T.J., 2003. The Atmospheric Chemistry of Alkoxy Radicals. Chemical Reviews, 103(12), pp.4657–4689. Ornes, S., 2016. Every breath we take. Physics World, 29(9), pp.39–41. Osthoff, H.D. et al., 2008. High levels of nitryl chloride in the polluted subtropical marine boundary layer. Nature Geoscience, 1(5), pp.324–328. Oudin, A. et al., 2016. Traffic-related air pollution and dementia incidence in Northern Sweden: A longitudinal study. Environmental Health Perspectives, 124(3), pp.306– 312. Park, K. et al., 2013. Large interannual variations in nonmethane volatile organic compound emissions based on measurements of carbon monoxide. Geophysical Research Letters, 40(1), pp.221–226. Phillips, G.J. et al., 2012. Significant concentrations of nitryl chloride observed in rural continental Europe associated with the influence of sea salt chloride and anthropogenic emissions. Geophysical Research Letters, 39(10), pp.1–5. Rao, X. et al., 2017. Effect of Particulate Matter Air Pollution on Cardiovascular Oxidative Stress Pathways. Antioxidants & Redox Signaling, 28(9), p.ars.2017.7394. Reed, C. et al., 2016. Interferences in photolytic NO2measurements: Explanation for an apparent missing oxidant? Atmospheric Chemistry and Physics, 16(7), pp.4707– 4724. Research, A., 2018. ToF-CIMS Technical Sheet. Reyes-Villegas, E. et al., 2018. Online Chemical Characterization of Food-Cooking Organic Aerosols: Implications for Source Apportionment. Environmental Science & Technology, 52(9), pp.5308–5318. Reyes-Villegas, E. et al., 2016. Organic aerosol source apportionment in London 2013 with ME-2: Exploring the solution space with annual and seasonal analysis. Atmospheric Chemistry and Physics, 16(24), pp.15545–15559. Rissanen, M.P. et al., 2014. The formation of highly oxidized multifunctional products in the ozonolysis of cyclohexene. Journal of the American Chemical Society, 136(44), pp.15596–15606. Roberts, J.M. et al., 2011. Isocyanic acid in the atmosphere and its possible link to smoke-related health effects. Proceedings of the National Academy of Sciences, 108(22), pp.8966–8971. Roberts, J.M. et al., 2008. N2O5 oxidizes chloride to Cl2 in acidic atmospheric aerosol.

Science, 321(5892), p.1059. Royal Society of Chemistry, 2015. ChemSpider. Available at: http://www.chemspider.com/ [Accessed June 19, 2018]. Sanchez, B. et al., 2016. CFD modeling of reactive pollutant dispersion in simplified urban configurations with different chemical mechanisms. Atmospheric Chemistry and Physics, 16(18), pp.12143–12157. Scott, C.E. et al., 2014. The direct and indirect radiative effects of biogenic secondary organic aerosol. Atmospheric Chemistry and Physics, 14(1), pp.447–470. Seinfeld, J.H. & Pandis, S.N., 1998. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change. Physics Today, 51(10), pp.88–90. Sekimoto, K. et al., 2017. Calculation of the sensitivity of proton-transfer-reaction mass spectrometry (PTR-MS) for organic trace gases using molecular properties. International Journal of Mass Spectrometry, 421, pp.71–94. Sekler, M.S. et al., 2004. Monitoring genotoxicity during the photocatalytic degradation of p-nitrophenol. Journal of Applied Toxicology, 24(5), pp.395–400. Simon, M. et al., 2016. Detection of dimethylamine in the low pptv range using nitrate chemical ionization atmospheric pressure interface time-of-flight (CI-APi-TOF) mass spectrometry. Atmospheric Measurement Techniques, 9(5), pp.2135–2145. Simpson, D. et al., 1999. Inventorying emissions from nature in Europe. Journal of Geophysical Research: Atmospheres, 104(D7), pp.8113–8152. Sitch, S. et al., 2007. Indirect radiative forcing of climate change through ozone effects on the land-carbon sink. Nature, 448(7155), pp.791–794. Slade, J.H. et al., 2017. Cloud droplet activation through oxidation of organic aerosol influenced by temperature and particle phase state. Geophysical Research Letters, 44(3), pp.1583–1591. Soldatova, L.N. et al., 2011. Selected papers from the 13th Annual Bio-Ontologies Special Interest Group Meeting. Journal of Biomedical Semantics, 2(Suppl 2), p.I1. Speidel, M. et al., 2007. Sulfur dioxide measurements in the lower, middle and upper troposphere: Deployment of an aircraft-based chemical ionization mass spectrometer with permanent in-flight calibration. Atmospheric Environment, 41(11), pp.2427–2437. Stirnweis, L. et al., 2017. Assessing the influence of NOx concentrations and relative humidity on secondary organic aerosol yields from α-pinene photo-oxidation through smog chamber experiments and modelling calculations. Atmospheric Chemistry and Physics, 17(8), pp.5035–5061. Stutz, J. et al., 2010. Nocturnal NO3 radical chemistry in Houston, TX. Atmospheric Environment, 44(33), pp.4099–4106. Suda, S.R. et al., 2014. Influence of Functional Groups on Organic Aerosol Cloud Condensation Nucleus Activity. Environmental Science & Technology, 48(17), pp.10182–10190. Surratt, J.D. et al., 2010. Reactive intermediates revealed in secondary organic aerosol formation from isoprene. Proceedings of the National Academy of Sciences, 107(15), pp.6640–6645. Terzano, C. et al., 2010. Air pollution ultrafine particles: Toxicity beyond the lung. European Review for Medical and Pharmacological Sciences, 14(10), pp.809–821. Turnipseed, A.A. et al., 2006. Eddy covariance fluxes of peroxyacetyl nitrates (PANs)

and NOy coniferous forest. Journal of Geophysical Research Atmospheres, 111(9), pp.1–17. UK Goverment, 1875. Public Health Act, London: Eyre and Spottiswoode. United Kingdom Parliament, 1956. Clean Air Act, 1956, London: The Stationery Office. United Kingdom Parliament, 1974. Control of Pollution Act 1974, London: The Stationery Office. United Kingdom Parliament, 1990. Environmental Protection Act, London: The Stationery Office. United Kingdom Parliament, 1989. The Air Quality Standards Regulations 1989, London: The Stationery Office. United Kingdom Parliament, 1981. The Motor Fuel (Lead Content of Petrol) Regulations 1981, London: Statutory Instruments. United Kingdom Parliament, 1991. The Road Vehicles Regulations 1991, London: The Stationery Office. Valorso, R. et al., 2011. Explicit modelling of SOA formation from α-pinene photooxidation: Sensitivity to vapour pressure estimation. Atmospheric Chemistry and Physics, 11(14), pp.6895–6910. Villanueva-Fierro, I., Popp, C.J. & Martin, R.S., 2004. Biogenic emissions and ambient concentrations of hydrocarbons, carbonyl compounds and organic acids from ponderosa pine and cottonwood trees at rural and forested sites in Central New Mexico. Atmospheric Environment, 38(2), pp.249–260. Wang, B. et al., 2016. Forests and ozone: Productivity, carbon storage, and feedbacks. Scientific Reports, 6(October 2015), pp.1–7. Wang, S. et al., 2017. Formation of Highly Oxidized Radicals and Multifunctional Products from the Atmospheric Oxidation of Alkylbenzenes. Environmental Science and Technology, 51(15), pp.8442–8449. Wang, X. et al., 2017. Observations of N2O5 and ClNO2 at a polluted urban surface site in North China: High N2O5 uptake coefficients and low ClNO2 product yields. Atmospheric Environment, 156(3), pp.125–134. WHO, 2014. Burden of disease from the join effects of Household and Ambient Air Pollution. World Health Organization, Global Health Risks. Available at: http://www.who.int/phe/health_topics/outdoorair/databases/FINAL_HAP_AAP_BoD _24March2014.pdf?ua=1 [Accessed June 25, 2018]. Wiedinmyer, C., Yokelson, R.J. & Gullett, B.K., 2014. Global emissions of trace gases, particulate matter, and hazardous air pollutants from open burning of domestic waste. Environmental Science and Technology, 48(16), pp.9523–9530. Williams, J., 2004. Organic trace gases in the atmosphere: An overview. Environmental Chemistry, 1(3), pp.125–136. Wong, J.P.S. et al., 2017. Observations and impacts of bleach washing on indoor chlorine chemistry. Indoor Air, 27(6), pp.1082–1090. Yuan, B. et al., 2016. A high-resolution time-of-flight chemical ionization mass + spectrometer utilizing hydronium ions (H3O ToF-CIMS) for measurements of volatile organic compounds in the atmosphere. Atmospheric Measurement Techniques, 9(6), pp.2735–2752. Yuan, B. et al., 2015. Secondary formation of nitrated phenols: insights from observations during the Uintah Basin Winter Ozone Study (UBWOS) 2014.

Atmospheric Chemistry and Physics Discussions, 15(20), pp.28659–28697. Zheng, W. et al., 2011. Characterization of a thermal decomposition chemical ionization mass spectrometer for the measurement of peroxy acyl nitrates (PANs) in the atmosphere. Atmospheric Chemistry and Physics, 11(13), pp.6529–6547.

Appendix A. Co-authorship in peer reviewed publications

1. E. Reyes-Villegas, M. Priestley, Y. C. Ting, S. Haslett, T. Bannan, M. Le Breton, P. I. Williams, A. Bacak, M. J. Flynn, H. Coe, C. Percival, and J. D. Allan, “Simultaneous aerosol mass spectrometry and chemical ionisation mass spectrometry measurements during a biomass burning event in the UK: Insights into nitrate chemistry,” Atmos. Chem. Phys., vol. 18, no. 6, pp. 4093–4111, 2018. 2. E. Reyes-Villegas, T. Bannan, M. Le Breton, A. Mehra, M. Priestley, C. Percival, H. Coe, and J. D. Allan, “Online Chemical Characterization of Food-Cooking Organic Aerosols: Implications for Source Apportionment,” Environ. Sci. Technol., vol. 52, no. 9, pp. 5308–5318, May 2018. 3. W. Zhou, J. Zhao, B. Ouyang, A. Mehra, W. Xu, Y. Wang, T. J. Bannan, S. D. Worrall, M. Priestley, A. Bacak, Q. Chen, C. Xie, Q. Wang, J. Wang, W. Du, Y. Zhang, X. Ge, P. Ye, J. D. Lee, P. Fu, Z. Wang, D. Worsnop, R. Jones, C. J. Percival, H. Coe, and Y. Sun, “Production of N2O5 and ClNO2 in summer in urban Beijing, China,” Atmos. Chem. Phys., vol. 18, no. 16, pp. 11581–11597, 2018. 4. M. Le Breton, Å. M. Hallquist, R. K. Pathak, D. Simpson, Y. Wang, J. Johansson, J. Zheng, Y. Yang, D. Shang, H. Wang, Q. Liu, C. Chan, T. Wang, T. J. Bannan, M. Priestley, C. J. Percival, D. E. Shallcross, K. Lu, S. Guo, M. Hu, and M. Hallquist, “Chlorine oxidation of VOCs at a semi-rural site in Beijing: significant chlorine liberation from ClNO2 and subsequent gas- and particle-phase Cl–VOC production,” Atmos. Chem. Phys., vol. 18, no. 17, pp. 13013–13030, Sep. 2018. 5. J. Matthews, A. Bacak, M. Khan, M. Wright, M. Priestley, D. Martin, C. Percival, and D. Shallcross, “Urban Pollutant Transport and Infiltration into Buildings Using Perfluorocarbon Tracers,” Int. J. Environ. Res. Public Health, vol. 14, no. 2, p. 214, 2017.