Heard D.E. (editor) Analytical Techniques for Atmospheric Measurement

Подождите немного. Документ загружается.

288 Analytical Techniques for Atmospheric Measurement

To reliably be detected, a particle must accurately pass through the focal points of

all the lasers and absorb a sufficient amount of the desorption laser energy to generate

ions. This will not occur for all of the particles, as the particle beam is slightly divergent.

Allen et al. (2000) measured detection efficiencies ranging from 1 hit in 10

to 10

particles

when sampling ambient particles with the ATOFMS. The scattering and absorbance of the

respective laser photons are dependent upon the composition and structure of the particle

and the laser wavelengths used. These factors can lead to several size and composition

biases in the detection efficiency (Carson et al., 1997b; Kane & Johnston, 2000; Silva &

Prather, 2000). While present in all differentially pumped instrument designs, the particle

beam focusing biases can be more pronounced with LDI than with the thermal methods,

as the laser focal points are generally smaller than the thermal vaporiser surfaces. Again,

the use of aerodynamic lenses at the beginning of the twenty-first century has helped to

mitigate these over certain size ranges (Su et al., 2004).

Many laser designs have been used for the desorption, such as ArF (193 nm) (e.g. Kane

et al., 2001), XeCl (308 nm) (e.g. Yang et al., 1996), CO

106 m (e.g. Prather et al.,

1994), N

(337 nm) (e.g. Trimborn et al., 2000) and frequency tripled or quadrupled

Nd:YAG (355 or 266 nm) (e.g. Gard et al., 1997). The choice of laser used in a particular

implementation is influenced by several factors. These include the pulse power and

duration capabilities, power consumption, cost, portability and the ionisation threshold

(minimum power flux required to generate ions). However, an individual instrument

is not normally limited to a specific laser and many groups have experimented with

multiple types. The ionisation threshold is dictated by both the wavelength of the laser

light used and the composition of the particles under analysis, as the absorption bands

and ionisation energies vary greatly between particle constituents. Generally speaking,

less laser power is needed to achieve ionisation for shorter wavelengths, as fewer photons

per molecule are needed (Thomson & Murphy, 1993). As the amount of energy absorbed

by the particle increases, a greater fraction of the constituents become desorbed and/or

ionised, and a greater amount of molecular fragmentation occurs. These phenomena

continue to increase until a second threshold is reached, whereby a plasma is formed.

The key technical parameter is the total amount of energy delivered per unit area during

a laser pulse, known as the fluence.

A potential problem with using TOF mass spectrometers with LDI is that the high-

powered vaporisation and ionisation process produces ions with a wide distribution of

initial velocities, which if unchecked, would lead to loss of resolution. The PSPF method

has been used to remedy this (e.g. Hinz et al., 1996; Murphy & Thomson, 1995), but

the favoured method for many of the instrument designs since mid-1990s is the use of

reflectrons, and m/m resolutions of the order of hundreds are typically reported (Gard

et al., 1997; Reents & Ge, 2000). The ionisation process also routinely produces many

ions of both positive and negative charges, so another common technique is to use two

mass spectrometers of opposite polarities simultaneously, so two complete mass spectra

are delivered per particle detected. This technique was introduced by Hinz et al. (1996),

applied to the original LAMPAS, and has been used in other instruments since, such

as the LAMPAS 2, ATOFMS and RSMS III. This increases the amount of composition

information gained for individual particles, vastly improving the value of the data. Laser

desorption and ionisation is not limited to TOF mass spectrometry, although it is the

most common method. Yang et al. (1996) and Parker et al. (2000) presented laser-based

Mass Spectrometric Methods for Aerosol Composition Measurements 289

instruments that employ ion trap mass spectrometers. One advantage of this configuration

is that it permits MS-MS, see Section 6.4.1. For instance, detection and identification

of specific chemical species such as bis(2-ethylhexyl)-phosphoric acid and elements such

as uranium in particles have been demonstrated using this method (Gieray et al., 1998;

Reilly et al., 1997).

Unlike electron or chemical ionisation methods, laser ionisation events are discrete,

intrinsically limited to individual particles and can produce complete mass spectra when

coupled to TOF-MS, so in this sense, these instruments can truly be described as single

particle mass spectrometers. When sampling ambient data, their real strength is their

ability to deliver composition and size information for specific particles, allowing the

explicit determination of mixing states with a very high time resolution. As the laser

pulses deliver much more energy to the particles than thermal methods, they are able to

vaporise all of the components of the particles and are not limited to the non-refractory

fraction. For this reason, they are also ideal for the study of trace components in particles

and are incredibly useful in assigning sources to particles based on these. On the other

hand, providing quantitative chemical information with LDI is inherently difficult. The

many complexities associated with the combined desorption and ionisation process result

in the limitation that for most LDI instruments, the chemical composition data derived

from sampling ambient particles are mainly qualitative in nature. If enough laser energy

to cross the plasma formation threshold can be delivered, the Particle Blaster can deliver

quantitative data on the elementary compositions of particles, as all the atoms present

are converted to positive ions (Reents & Ge, 2000; Reents & Schabel, 2001). A similar

quantitative method was also demonstrated by Mahadevan et al. (2002). However, as

the constituents are completely atomised in these processes, no molecular information

is retained.

At lower fluences, the constituent molecules remain more intact, allowing some

chemical information to be obtained, but at the same time, the precise ionisation

process can be highly variable in both the ions it produces and their abundances. A

significant component of this variability is due to uncertainties in the amount of laser

energy delivered to the particles and work has indicated that the reproducibility may

be significantly improved if the beam homogeneity and shot-to-shot power variability

of the desorption laser are better constrained (Wenzel & Prather, 2004). However,

individual chemical components can interact with each other during the combined

desorption and ionisation process, unevenly distributing energy and charge between the

chemical components (Neubauer et al., 1998; Reilly et al., 2000). These phenomena

are known as ‘matrix effects’ and the results are often difficult to predict, as they

are known to be non-linear and dependent on both the explicit compositions of

the particles and the precise amounts of energy absorbed. The study of complex

molecules is in turn hampered by the highly variable fragmentation behaviours that result

from this.

When looking at the contributions to the mass spectra from inorganic ions, it has

been found that their relative contributions, if not their absolute magnitudes, can be

quantified, providing that the chemical matrices are fairly consistent (Gross et al., 2000;

Middlebrook et al., 2003). In principle, quantitative inversions based on this behaviour

can be applied to ambient data, providing that an a priori particle matrix can be assumed

for a given population of particles and the mass spectra obtained can be linked with

290 Analytical Techniques for Atmospheric Measurement

either laboratory standards or composition data from other collocated instruments in the

field (Bhave et al., 2002; Dall’Osto et al., 2004; Fergenson et al., 2001). Work to produce

a reliable inversion method based on these techniques is ongoing.

The interactions between constituents can be used to the instrument’s advantage. For

instance, the technique of matrix-assisted LDI (MALDI) deliberately uses these matrix

effects to generate more predictable ionisation behaviours and can be applied to aerosol

mass spectrometers (Mansoori et al., 1996). It involves mixing or coating the material

under analysis with a specific substance designed to absorb the laser light and transfer

the energy during the desorption and ionisation in a more controlled manner. While

this is a powerful technique in the laboratory or for identifying very specific compo-

nents (e.g. Arnold & Reilly, 1998), it is not ideal for general ambient sampling as

it involves perturbing the aerosol composition. Another technique where interactions

between constituents is taken advantage of is in situ chemical ionisation, where the

desorption laser is operated at a lower fluence to generate specific ions (mainly Na

and K

) from the particle. These ions combine with the neutral desorbed molecules

and like other forms of chemical ionisation, do not result in molecular fragmen-

tation, making it useful for the identification of organics. Lazar et al. (2000) were able

to demonstrate the capability of this technique to study surface coatings of tributyl

phosphate on soil particles, but as this is a very specific technique, it is not suitable for

general-purpose work.

6.4.4.3 Two-step laser methods

The steps of desorption and ionisation do not necessarily need to be performed together

in laser-based instruments. Instead, a pulse from a lower energy CO

or eximer laser

can be used to vaporise the particle before a second ionisation laser ionises the resultant

vapours immediately afterwards (Cabalo et al., 2000; Morrical et al., 1998; Zelenyuk

et al., 1999). As the ionisation is performed in the gas phase, most of the problems

associated with matrix effects are avoided. Also, as less energy is required to perform the

ionisation, the power delivered by the ionisation laser can be reduced so the molecules are

subject to less fragmentation. However, the trade-off with this general technique is that

the instrument’s capability for studying inorganics is reduced, especially the refractory

fraction. By controlling the desorption laser’s power, it is also possible to selectively study

the outer layers of the particle (e.g. Lazar et al., 1999; Woods et al., 2002). Molecular

fragmentation associated with the single-step processes is reduced, but not eliminated

entirely, and because the energy of individual photons is too small to cause ionisation, the

molecules must be bombarded with laser light intense enough to induce multiple photon

absorption, leading to uncertainties in the amount of energy delivered to individual

molecules. To address this, the Chapel Hill instrument uses the 9

harmonic of an

Nd:YAG laser (118 nm, in the vacuum UV) to perform the ionisation in a two laser

based system. As only single photons of this energy (10.5 eV) are required to ionise

most molecules, the uncertainty in the energy delivered and the amount of molecular

fragmentation that occurs are both reduced substantially. Sykes et al. (2002) presented a

modification to the Chapel Hill instrument whereby the laser vaporisation stage could be

replaced with a heated surface in a similar manner to the thermal desorption techniques,

but the ionisation was still performed with the VUV laser. This allows the temperature

Mass Spectrometric Methods for Aerosol Composition Measurements 291

of the vaporisation to be controlled, allowing study into the volatility of the chemical

constituents. The two CW detection lasers were still used to detect the presence of the

particle, as these controlled the triggering of the ionisation laser. The instrument has

been shown to be very powerful in both the identification and quantification of organic

species within particles but its relatively low sensitivity makes the implementations best

suited for laboratory studies, at least so far.

An alternative ionisation method that has been demonstrated recently is photoelectron

resonance capture ionization (PERCI) (LaFranchi & Petrucci, 2004; LaFranchi et al., 2004),

which uses the low-energy electrons produced from a metallic surface when illuminated

with a pulsed laser. These electrons transfer onto the gas-phase vapour products of the

particles after laser or thermal desorption, forming negative ions. As the energy of the

electrons is small and controllable, this too results in reduced fragmentation of the organic

molecules. It is also possible to use laser desorption coupled with continuous ionisation

techniques. Reilly et al. (2004) presented a method where particles are collected in bulk

on a surface before being vaporised either with a pulsed laser or alternatively, heating the

surface in a similar manner to the thermal methods. The ionisation can be performed by

electron impact, chemical ionisation or glow discharge, and an ion trap is used for analysis.

This technique could potentially benefit from the quantitative capability that decoupled

desorption/ionisation offers with the sensitivity needed for ambient sampling. However,

as the particles are handled in bulk, there is no sizing or single particle capability.

6.5 Analysis of ambient data

Many of these instruments described above have been used to sample ambient aerosols

for a variety of different purposes at a multitude of locations. These include ground-based

sites in urban, rural and remote environments, and mobile platforms such as on-road

vans, ships and aircraft. These have presented extra challenges at the analysis stage,

as the aerosol climatology can be very complex and often highly unpredictable, given

that aerosol mass spectrometry delivers data on a much more detailed level than was

previously possible. As each different instrument design produces fundamentally different

data, many novel approaches towards data analysis have been developed.

As stated before, the real strength of the thermal methods is the ability to deliver

quantitative composition data. The AMS in particular is capable of delivering simulta-

neous mass concentrations, in gm

−3

, for multiple chemical components by applying

a matrix-based inversion to the mass spectra and calibrations of ammonium nitrate

standards (Allan et al., 2004a). Typical data products include chemical loadings of nitrate,

sulphate, ammonium and organic matter. When combined with the sizing data described

previously, size-resolved dM/dlogD distributions (also in gm

−3

) can also be generated

for the individual species (Allan et al., 2003a). Processed data of these types have the

advantage of being directly comparable to those produced by other aerosol chemistry

instrumentation such as bulk samplers and in situ analysers (e.g. Allan et al., 2004b;

Drewnick et al., 2003; Topping et al., 2004) and the products of numerical modelling

tools (e.g. Tang et al., 2004), facilitating overall data synthesis and interpretation.

While the APCI-MS and TD-CIMS can use chemical ionisation and tandem mass

spectrometry to deliver rich data on the organic chemicals present in a sample, EI

292 Analytical Techniques for Atmospheric Measurement

and single quadrupole analysis is a limited technique when trying to identify specific

organic species, as fragmentation produces many different peaks in the mass spectrum.

Atmospheric particles contain a myriad of different organic compounds (Jacobson et al.,

2000; Saxena & Hildemann, 1996), so in most real-world environments, the ensemble

organic data are impossible to completely deconvolve. The analysis of ambient organic

data in the AMS is often limited to the study of generalised classes of organic species such

as hydrocarbon like and more oxidised types, based on markers within the mass spectra

(Allan et al., 2003b; Canagaratna et al., 2004; Zhang et al., 2005). The temperature-

programmed mode of operation of the TDPMS is useful in addressing this, as the mass

spectra are delivered as a function of volatility, separating the components.

Instruments using LDI tend to produce very large data volumes during ambient

sampling, as individual mass spectra for as many as hundreds of thousands of particles

can be captured during a field campaign. The need to reduce this data to a form

that is both usable and scientifically valid has prompted the utilisation of some

innovative new techniques and their ongoing development is seen by many as one of

the main challenges currently facing the single-particle mass spectrometry community

(Sipin et al., 2003).

Most methods are based around grouping particles according to common features

in their mass spectra and linking these groups to particle compositions and sources.

Providing that the particle sampling and detection biases can accurately be taken into

account (Wenzel et al., 2003), these analyses can be used to deliver quantitative data

on the size and number concentrations (instrument capability permitting) of speciated

particle populations with a high time resolution. Due to the inherent amount of variability

in the spectra captured, the criteria for grouping the particles must be carefully chosen.

Many methods have been used over the years, each with their own advantages and

disadvantages, and these are very much in a state of continual development.

A common generalised method is to compare all the particle spectra after data collection

has been completed, grouping them according to their numerical similarity and identi-

fying the common features within the groups to gain information about their compo-

sitions. Examples of such approaches include principle component analysis (Hinz et al.,

1996), fuzzy cluster analysis (Hinz et al., 1999), multivariate patch analysis (Parker et al.,

2000) and hierarchical cluster analysis (Murphy et al., 2003). The grouping produced by

these methods can be further inspected when making scientific conclusions. While these

techniques are very powerful, they are computationally intensive and can only be applied

to complete datasets, making near-real-time analysis very difficult.

An alternative approach used by some groups is to analyse spectra sequentially using

pattern matching algorithms such as the ART-2A (Adaptive Resonance Theory) neural

network (Bhave et al., 2001; Carpenter et al., 1991; Phares et al., 2001; Song et al.,

1999) and ADAMS (Algorithm for Discriminant Analysis of Mass Spectra) (Tan et al.,

2002). These can group particle spectra according to similarities with previously seen

spectra and can adapt to unexpected results and outliers by creating new groups on

the fly or progressively altering the reference spectra. While useful for offline analysis,

these techniques have the added advantage of permitting real-time analysis. The scientific

validity of the clustering produced during an analysis run is highly dependent on the

matching and learning parameters of the algorithm and any previous training it has had,

so these must be controlled carefully.

Mass Spectrometric Methods for Aerosol Composition Measurements 293

6.6 Applications of aerosol mass spectrometric

techniques to atmospheric aerosol sampling

The purpose of this section is not to provide a thorough review of every field application

of aerosol mass spectrometers since mid-1990s, but to give a series of examples of different

applications to demonstrate the wide applicability of the techniques.

Until the mid-1990s aerosol mass spectrometers were laboratory-based developmental

instruments. However, since that time their use in sampling the ambient atmosphere has

increased substantially and many current large field campaigns contain one or more types

of such instruments. Many of the early field applications not only demonstrated proof

of concept of aerosol mass spectrometric techniques as valuable tools, but supplied new

insights into aerosol composition and mixing state. During the first Aerosol Characteri-

sation Experiment (ACE-1) in the southern ocean, Murphy and co-workers used PALMS

to investigate the marine boundary layer aerosol under pristine conditions (Murphy et al.,

1998a,b). Sample data are shown in Figure 6.6. The study showed that almost all of

the submicron aerosols detected contained sea salt, nearly all the observed sulphate was

mixed with the sea-salt particles, the chloride-to-bromide ratio was approximately that of

seawater, and iodine was enhanced to approximately the concentration of bromine. The

iodine appeared to be correlated with organic particles and the chloride and bromide were

anticorrelated with sulphate (Murphy et al., 1997). Silva et al., (1999) used both positive

and negative ion spectra to differentiate particles formed during biomass burning from

those of other combustion sources. This is possible, since potassium produces strong

positive ion signals and these provide a key marker for biomass burning as a source.

Since these earlier studies, aerosol mass spectrometric techniques have rapidly

developed and are now used widely in field experiments. There are several studies of

long-range transport of aerosols that link their mixed composition to source and particle

history. For example, Guazzotti et al. (2003) measured single particle mass spectra with

the ATOFMS instrument during INDOEX (INDian Ocean EXperiment), from a research

ship in the Indian Ocean, and compared sources of particulate emanating from the

Indian subcontinent and from the Arabian Peninsula. Clear differences in the mass

spectra were discernable between these two source regions, with the Indian subcontinent

being heavily influenced by biomass burning from household fires whereas fossil fuel

combustion dominated in the Arabian Sea area. In the former case 74% of the observed

aerosols were shown to be of biomass burning origin, whereas in the latter case 68% of

the measured aerosols were of fossil fuel combustion origin. Source identification and

processing was also studied by Dall’Osto et al. (2004), who studied single particle mass

spectra at Mace Head, a coastal Atlantic Ocean site on the west coast of Ireland, during the

2002 NAMBLEX (North Atlantic Marine Boundary Layer Experiment) field campaign,

using a commercial ATOFMS. Particles of several different sources were measured during

the course of a two-week measuring period (Figure 6.7) and Saharan dust influences

were identified with particles displaying strong aluminium and silicon signatures, whereas

more locally produced dust was rich in calcium. Sea salt was often measured but it was

shown to be processed to varying degrees, displaying a spectrum similar to pure sea

salt when locally produced but demonstrated displacement of the chloride by nitrate in

more aged particles, as nitric acid uptake has displaced the less volatile chloride from the

particles.

294 Analytical Techniques for Atmospheric Measurement

0.001

0.01

Detector signal (mA)

12010080604020

–

NaCl

–

NaSO

–

0.001

0.01

Detector signal (mA)

(a) Mass/charge ratio

12010080604020

(b) Mass/charge ratio

–

HSO

–

HSO

–

0.001

0.01

Detector signal (mA)

12010080604020

–

HSO

–

HSO

–

? & SO

–

Figure 6.6 Sample mass spectra from the PALMS instrument taken during ACE-1 at Cape Grim. The

three negative ion spectra show that (a) sodium sulphate particles, which were the most abundant type of

sulphate-containing mass spectra, can be distinguished from (b) either ammonium sulphate or sulphuric

acid particles, and (c) a probable mixture of methane sulphonic acid (MSA) and ammonium sulphate

or sulphuric acid. The latter particles were very rare. The assignments were made by comparison with

laboratory data (Murphy et al., 1998a).

Urban measurements have also been made extensively using aerosol mass spectrometry.

Motor vehicle and diesel truck emissions were studied on a single particle basis by Gross

et al. (2000) during a tunnel study using the ATOFMS. Three clear classes of compounds

were observed: those that were composed of polycyclic aromatic hydrocarbons (PAHs);

Mass Spectrometric Methods for Aerosol Composition Measurements 295

100

200

300

400

500

600

(a)

Aged sea-salt

Pure sea-salt

Mixed sea-salt

Unscaled ATOFMS counts

03/08/2002 13:00

01/08/2002 23:00

05/08/2002 10:00

06/08/2002 00:00

08/08/2002 20:00

11/08/2002 18:00

12/08/2002 20:00

16/08/2002 19:00

18/08/2002 14:00

19/08/2002 11:00

21/08/2002 11:00

Figure 6.7 Demonstration that the ATOFMS can be used to assess the extent of ageing of sea salt

particles in the marine boundary layer. Mass spectra from sea salt were further grouped according to the

ratio of the nitrate-to-chloride ion ratios, indicating the level of displacement of chloride by the nitrate

ion (Dall’Osto et al., 2004).

carbon particles, or soot; and particles containing inorganic compounds that produce

strong ion signals of aluminium, calcium, iron and barium. The study was able to separate

the number fraction of these groups and showed that the PAH number fraction is

approximately 60% with the other two groups approximately 10% each. The work shown

in Figure 6.8, measured during a study in Baltimore, by Lake et al. (2003) demonstrated

that the RSMS-III instrument was capable of obtaining submicron particle mass spectra

as a function particle size for particles greater than 45 nm aerodynamic diameter. In

the urban environment, this ability allows the instrument to probe the highly dynamic

ultra-fine particles produced in large numbers by motor vehicle emissions. The authors

demonstrated that carbon-containing particles were mostly present in the smallest sizes

measured and correlated with the rush hour traffic flow peak.

The aerodyne aerosol mass spectrometer (AMS) is ideally suited to measuring organic

aerosols in the urban environment and their interactions with inorganic species. Sample

data are shown in Figures 6.9–6.11. This is because it can sample small sized aerosol,

it targets non-refractory aerosol components and it produces quantitative mass loading

information. The field application of the instrument was first demonstrated by Jimenez

et al. (2003a) and the behaviour of urban aerosol has been investigated by Allan et al.

(2003b). Comparative studies in an urban environment, as displayed in Figure 6.9,

have shown that the instrument is indeed quantitative for common, submicron aerosol

components provided that several efficiencies are determined (Drewnick et al., 2003).

The modification of the particles as the urban pollution is advected from a city to the

surrounding region can also be observed (Alfarra et al., 2004; Boudries et al., 2004).

The observations show a separate small mode of particles in urban areas composed almost

entirely of organic species whose mass spectra are similar to those observed from engine

296 Analytical Techniques for Atmospheric Measurement

(a)

(b)

Aerodynamic

diameter, nm

Aerodynamic

diameter, nm

0.8–1

1500–2000

1000–1500

500–1000

0–500

dN/dlog(d

Particles/cm

0.6–0.8

0.4–0.6

0.2–0.4

0–0.2

Hlt fraction

5/21 1AM

5/21 1PM

5/22 1AM

5/22 1PM

5/22 11PM

5/21 1AM

5/21 1PM

5/22 1AM

5/22 1PM

5/22 11PM

114

143

228

436

114

143

228

436

Time

Figure 6.8 The top panel shows a contour plot of the fraction of carbon containing particles in the

measured particle population (identiﬁed by large signals at m/z 28) as a function of their aerodynamic

diameter. The bottom ﬁgure shows the number distribution of the same particle population as a function

of aerodynamic diameter and time. It is clear that the carbon containing particles are mostly below

100 nm and peak in the early morning rush hour, due to a combination of increased trafﬁc ﬂow and the

suppression of mixing caused by the presence of a nocturnal inversion (Lake et al., 2003).

Sulphate mass concentration / µg/m

07/03/01 07/07/01 07/11/01 07/15/01 07/19/01 07/23/01 07/27/01 07/31/01 08/04/01

Date

R&P 8400S

AMS

HSPH

PILS

Figure 6.9 A comparison of submicron sulphate mass concentration between the AMS, a Particle Into

Liquid Sampler (PILS) (Weber et al., 2001), a Rupprecht and Patashnick 8400S and a continuous sulphate

monitor developed at the Harvard School of Public Health (HSPH).

Mass Spectrometric Methods for Aerosol Composition Measurements 297

dM/dlogD

(µg m

–3

)

Slocan Park

dM/dlogDva (µg m

–3

)

234567

100

234567

1000

234567

100

234567

1000

Vacuum aerodynamic diameter (nm)

Organics

Sulphate

Nitrate

Langley

Figure 6.10 The AMS produces mass loadings as a function of particle vacuum aerodynamic diameter

for organics, sulphate and nitrate fractions. The upper panel shows an average distribution for the city

of Vancouver, and the lower panel shows a similar period at Langley, a rural location impacted by aged

air from the Seattle area. The data were collected during the PACIFIC 2001 air quality study in Canada

during 2001 (Alfarra et al., 2004).

oil particles. The accumulation mode appears to be a mixture of ammonium sulphate

and organic material with a variable amount of nitrate dependent on the source profile

of the region. The organic fraction in the larger mode has a very different mass spectral

signature and shows large peaks at m/z of 18 and 44 (H

and CO

), which have only

been observed to be dominant in test particles when sampling di- and polycarboxylic

acids. The AMS has also been used in several US air quality studies (e.g. Middlebrook

et al., 2003). In background air, the particle composition has been modified. For example,

during ACE-ASIA it was found that the submicron particulate mass was dominated

by accumulation mode acidified sulphate and organic material, whose mass spectral

fingerprint was again dominated by the CO

fragment. Nitrate is very much reduced

and is possibly organic in nature and the components appear to be internally mixed

(Topping et al., 2004). Comparisons between the AMS and offline techniques or bulk

online submicron species mass loading have shown good agreement between nitrate,

sulphate and organic fractions in many situations (e.g. Drewnick et al., 2003).

Ice nucleation is an extremely important phenomenon in the atmosphere, as liquid

droplets will not freeze without the presence of a seed until temperatures as cold as