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.
1
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.
2
Copyright Statement
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.
3
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.
4
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
5
Blank page
6
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).
7
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).
9
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.
10
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).
11
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).
12
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
13
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.
O3
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
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)
16
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)
17
�� ��� → �� + �� (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.
18
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.
21
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.
22
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).
23
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
24
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).
26
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).
30
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.
33
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.
34
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.
35
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 (-
36
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.
38
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.
40
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.
41
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.
43
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).
45
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.
47
Blank page
48
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: