A Mechanism for the Production of Ultrafine Particles from

Concrete Fracture

Authors: Nassib Jabboura, E. Rohan Jayaratnea, Graham R Johnsona, Joel Alroea, Erik Uhdez, Tunga

Salthammerz, Luke Cravigana, Ehsan Majd Faghihia, Prashant Kumarx, y, Lidia Morawskaa,*

aInternational Laboratory for Air Quality and Health, Queensland University of Technology, Brisbane,

Australia

xDepartment of Civil and Environmental Engineering, Faculty of Engineering and Physical Sciences,

University of Surrey, Guildford GU2 7XH, United Kingdom

yEnvironmental Flow (EnFlo) Research Centre, Faculty of Engineering and Physical Sciences,

University of Surrey, Guildford GU2 7XH, United Kingdom

zFraunhofer WKI, Department of Material Analysis and Indoor Chemistry, 38108 Braunschweig,

Germany

* Corresponding author contact details:

Tel: (617) 3138 2616; Fax: (617) 3138 9079

Email: l.morawska@qut.edu.au

1

Abstract

While the crushing of concrete gives rise to large quantities of coarse dust, it is not widely

recognized that this process also emits significant quantities of ultrafine particles. These particles

impact not just the environments within construction activities but those in entire urban areas. The

origin of these ultrafine particles is uncertain, as existing theories do not support their production by

mechanical processes. We propose a hypothesis for this observation based on the volatilisation of

materials at the concrete fracture interface. The results from this study confirm that mechanical

methods can produce ultrafine particles (UFP) from concrete, and that the particles are volatile. The

ultrafine mode was only observed during concrete fracture, producing particle size distributions with

average count median diameters of 27, 39 and 49 nm for the three tested concrete samples. Further

volatility measurements found that the particles were highly volatile, showing between 60 and 95 %

reduction in the volume fraction remaining by 125 °C. An analysis of the volatile fraction remaining

found that different volatile material is responsible for the production of particles between the

samples.

Keywords:

Ultrafine particles, concrete, secondary particles, airborne dust, urban pollution

Capsule:

For the first time, we investigate the nature and origin of ultrafine particles released during

the fracture of concrete.

2

1 Introduction

Airborne particles originate either as primary particles that are emitted directly from a source or

secondary particles that are produced in the air by the condensation of vapours emitted by the

source (Colbeck, 2014). Examples of primary particles are carbon soot particles emitted by motor

vehicles and industrial sources, windblown mineral dust and salt nuclei that originate from sea spray

during the breaking of ocean waves. Secondary particles are produced from precursor gases which

condense to form liquid droplets. A common example in the atmosphere is sulphur dioxide that is

oxidised to sulphuric acid gas which readily condenses to form volatile liquid droplets. Generally,

most primary particles are solid, while secondary particles are liquid. At the time of

emission/formation, primary particles are usually larger than secondary particles. Particles

originating from combustion processes consist of both primary particles in the form of soot as well

as secondary particles that are formed from the gaseous products (Kumar, Pirjola, Ketzel, &

Harrison, 2013). On the other hand, mechanical processes such as cutting, crushing and rubbing are

known to generate only primary particles directly from the source (Azarmi, Kumar, & Mulheron,

2014).

It is well known that coarse particles originate from mechanical activities such as fracturing and

crushing of building materials. These particles are generally larger than those classified as fine

particles (< 2.5 μm). The only known exception is the fracture of polymers containing nanocomposite

materials. For example, recent research by Sachse et al. (2012) has shown that the mechanical

drilling of silica-based polyamide 6 nanocomposites is accompanied by the emission of large

numbers of ultrafine particles. Van Broekhuizen, Van Broekhuizen, Cornelissen, and Reijnders (2011)

reviewed a number of mechanical processes involving engineered nanomaterials that resulted in the

emission of ultrafine particles. Ogura et al. (2013) studied the release of carbon nanotubes during

the grinding of polystyrene-based composites and listed a number of other studies that have

reported release of nanofibers from composite materials as a result of mechanical processes such as

cutting, drilling, sanding and grinding.

Particle pollution from construction activities impacts not just the surrounding areas but the greater

urban environment, particularly in megacities of countries with growing economies. For example,

according to the China Daily

(http://www.chinadaily.com.cn/china/2014-02/19/content_17292881.htm) there are presently over

5000 construction sites in Wuhan, China and the city is expected to invest 2 trillion yuan (US$329

billion) in urban construction over the next five years. The impact of this on the urban environment

will be quite significant (Faber, Drewnick, & Borrmann, 2015). In regards to the associated health

3

effects, it is well known that smaller particles have a greater impact than coarse particles, as they

can penetrate deeper into the lungs during inhalation. Recently, there has emerged evidence for the

release of ultrafine particles during mechanical processes involving the fracturing of non-composite

concrete. Asadi, M Hassan, and Dylla (2012) found large concentrations of nanoparticles during

asphalt and concrete preparation activities such as mixing, pouring and compaction. Approximately

49 % of the particles by number were in the smallest size category – between 10 and 30 nm. Kumar,

Mulheron, and Som (2012), using a fast-response differential mobility spectrometer, monitored

particle number concentration and size distribution in a confined room in the presence of simulated

building activities including crushing of concrete in a mixer and fracturing of concrete blocks with a

lump hammer. All processes were accompanied by elevated concentrations of ultrafine particles in a

tri-modal size distribution with peaks corresponding to fresh nuclei (<10 nm), nucleation mode (10 -

30 nm) and accumulation mode (30 - 300 nm). Later, Kumar and Morawska (2014) observed up to

93 % of particles by number in the ultrafine size range during the recycling of concrete and further

work exhibited the similar observations during cutting, drilling and mixing of concrete (Azarmi et al.,

2014). The observation of ultrafine particle emissions during the fracture of concrete is contrary to

our conventional knowledge of aerosol physics where, as shown above, ultrafine particles are not

usually produced by purely mechanical means. The study opened up a number of pertinent

questions regarding the mechanism of production of these ultrafine particles and their

physicochemical characteristics. The objective of the present study was to carry out further research

into this phenomenon in order to answer these two questions.

Here, we present a hypothesis that the observed airborne nanoparticles are secondary particles,

formed though nucleation. Conventional knowledge suggests that attractive cohesive forces

between particles are increasingly difficult to overcome at smaller particle sizes and that the simple

mechanical fracture of primary particles is effectively limited to sizes greater than approximately 500

nm (Hinds, 2012). Mechanical fracture would be more likely disperse any ultrafine material into

comparatively large clusters of primary particles producing an aerosol size distribution exhibiting a

peak in the low micrometer size range. Smaller clusters of primary particles below 500 – 700 nm in

diameter would also be present in the aerosol but their concentrations would decrease significantly

with decreasing cluster size (Corn, 1961a, 1961b). We hypothesize that during the fracture process,

the temperature of the concrete fissure rises high enough to vaporise semi-volatility species present

in the concrete. Once emitted into the atmosphere, these condensable gases would undergo

nucleation to form the observed ultrafine particles. This study aimed to derive evidence for such a

physical mechanism and determine the volatility and chemical nature of the nanoparticles emitted

during the fracture of concrete.

4

2 Methods

This study set out to explore particle production from concrete fracture in a laboratory setting. The

initial experiments were designed to determine whether particle formation could be observed from

concrete fracture conclusively, using a Scanning Mobility Particle Sizer (SMPS) system and an

experimental chamber. Subsequent experiments using a Volatility Tandem Differential Mobility

Analyser (VTDMA) characterised the volatility of the particles produced. An alternate particle

production method was attempted using an electric furnace. Gas Chromatography mass

spectrometry (GC/MS) was used for bulk concrete sample analysis, to determine the presence of

volatile material in the samples. A control study was also conducted to establish whether external

containments were responsible for the observed particle production.

Concrete Composition. The concrete samples chosen for the study were based on materials typically

used by industry and consumers for building and construction in Australia and the UK, rather than a

mix specifically developed for this project. Three different samples were used, termed RC (research

concrete), RSC (rapid set concrete) and CMLC (construction materials laboratory concrete). The RC

sample was freshly made high strength research concrete provided by the QUT Civil Engineering

Department. The RSC sample was purchased commercially from a local hardware store (Bastion™

Rapid Set Concrete, Bunnings Warehouse®), and also provided by the QUT Civil Engineering

Department. The CMLC sample was provided by the Construction Materials Laboratory (CML) at the

University of Surrey, conforming to the same standards as the previous study by Kumar et al. (2012).

The RC and RSC sample dimensions followed Australian concrete test cylinder standards at 100 mm

in diameter by 200 mm in length. The samples obtained from the CML were 100 × 100 × 40 mm.

Apparatus. All testing was conducted inside a 1 m3 experimental chamber, which was kept at a slight

overpressure to ensure that it was not contaminated by ambient air. This was achieved using

compressed air pumped into the chamber through HEPA filters. The input window on one side was

modified to accommodate a conductive bag to allow access to the chamber while maintaining the

seal. The sampling tubes were placed at the opposite end with their inlets covered with paper towel

serving as a filter to stop excessive numbers of coarse particles entering the measurement

equipment. A small fan was placed inside the chamber to facilitate mixing.

Particle size distribution (PSD) measurements were conducted with an SMPS, a system which

combines a condensation particle counter (CPC), a TSI 3071a Electrostatic Classifier (EC) and a Kr85

5

Neutraliser (TSI, Shoreview, MN). Two CPCs were used, a TSI 3010 butanol based CPC or a TSI 3787

water based CPC. Across all experiments, particle concentrations were measured with either 3010 or

3787 TSI CPCs, having size ranges of 10 – 300 nm and 5 – 300 nm, respectively.

Measurements of particle volatility were conducted with a VTDMA, an apparatus ensemble of two

TSI 3010 CPCs and two TSI 3071a ECs in conjunction with a TSI thermodenuder. The VTDMA

preselected particles of the required size, which were passed through a thermodenuder set at a

certain temperature. The effect of this heating process was observed through the resulting PSD,

acquired with the second EC.

Samples were heated in an electric furnace and the gases and particles produced were passed into

the chamber. The furnace was kept at slight overpressure and before each use it was heated to 300

°C to remove any residues present on its inner walls. This prevented the formation of FuGPs from

contaminent solvents introduced into the testing chamber. All experiments conducted with the

furnace occurred at 300 °C.

Study Design. Fracture generated particles (FrGPs) were created by striking the concrete with a

hammer. The sample was placed inside the chamber on a wooden board, and a one handed, 2 kg

lump hammer was used to fracture the concrete by gripping the handle through the conductive bag.

Prior to fracturing, the chamber was flushed with filtered air until particle concentrations were

below 50 particles cm-3. When striking the concrete, two methods were used. Firstly, the concrete

was struck without causing fractures. This was done by letting the hammer strikes land in the centre

of the sample with the strike direction perpendicular to the concrete. The second method involved

fracturing the concrete. This was achieved by striking at a 45° angle towards the edge of the sample.

During the SMPS down scan, the concrete was struck for the full 60 s, allowing the particle

concentration to build up. When the up scan started, striking ceased, with sampling beginning

immediately. These two methods allowed the contribution of FrGPs to be separated from the overall

particle distribution created by striking the concrete.

Furnace generated particles (FuGPs) were produced from concrete samples heated in a ceramic

electric furnace. The sample was placed inside the furnace outside the heating zone before the tube

was sealed, and a metal rod was placed at the inlet to allow the concrete samples to be moved. The

tube was sealed with Parafilm® wax tape. The furnace was flushed with air through a HEPA filter,

and then heated to the target temperature. Once the furnace temperature had settled, the metal

rod was used to move the concrete samples into the middle of the furnace. The furnace was then

connected to the chamber and a slight airflow through a HEPA filter at the inlet was used to direct

6

any particles generated, allowing them to mix inside. The PSD was measured using the same SMPS

system as the fracture experiments.

A VTDMA was used to measure the volatility of any observed ultra fine particles (UFPs). The PSDs

previously measured during fracture experiments were used to select four particle diameters for

pre-selection by the classifier, to characterise the UFP mode. The thermodenuder was set to

increase in 25 °C increments, between the ambient temperature (approximately 25 °C) and 125°C.

For FrGPs, the method involved four scans, one for each preselected size, with an up scan time of

150 s and a down scan time of 60 s. During the down scan of the last preselected size, the

temperature was increased by 25 °C and additional FrGPs were generated using theprevious

method.

Qualitative dynamic headspace GC/MS was applied to determine the volatile organic components

(VOCs) in the concrete. 8 - 15 mg of the material was placed in the centre of a standard thermal

desorption glass tube and fixed between glass wool plugs. The volatile components were thermally

extracted by purging the glass tube for 8 min with a controlled helium flow of 20 ml/min at 60 °C and

100 °C, respectively. The analytes were collected on an electrically cooled trap (-30 °C) filled with

quartz beads, Carbograph™ 2 and Carbograph™ 1 and then injected into the GC-system by quickly

heating the trap to 300 °C (6 min hold time) under continuous He flow (10 ml/min) and a 1:10 split

ratio. The GC (Agilent 7890 series) was equipped with a HP5MS column (60 m x 250 µm x 0.3 µm);

the oven temperature program was 32°C (1 min) with 8°C/min to 280°C (5 min). The mass selective

detector (Agilent 5975N) was operated in the full scan mode. The substances were identified by

Probability Based Matching (PBM) library search (McLafferty & Tureček, 1993) and on the basis of

original standards. The samples provided for GC/MS analysis were from the fine concrete residue

after fracture experiments were conducted.

To quantify the level of VOCs which may have deposited on the concrete samples from the ambient

air, sample pieces of concrete and aluminium foil were left to accumulate volatile material, then

placed in the furnace and heated while the particle concentration was measured with a CPC. These

pieces were initially cleaned in the furnace at 300 °C for several hours to remove any surface volatile

compounds, after which they were left in ambient air inside the laboratory for 6 weeks. Before

testing the concrete and aluminium foil, the furnace was raised to a high temperature to remove any

contaminants. For concrete, several pieces were placed inside the cooled furnace, in direct contact

with the heating element at the centre. The concrete samples were handled with gloves or forceps

to avoid contamination and the furnace was sealed and flushed with filtered air. Two methods were

employed to determine the level of deposition. Firstly, the furnace was heated slowly while

7

recording the temperature and measuring concentration with a CPC at regular intervals. Secondly,

the sample was initially positioned inside the furnace but outside its heating zone while the furnace

was heated to the target temperature. When the target temperature was reached, the sample was

moved into the centre of the furnace using a metal pipe at the inlet. The inlet of the pipe was sealed.

This allowed for rapid heating of the sample, quickly achieving a supersaturated environment for

condensation to occur. The first method was also conducted on the aluminium foil. For both

methods, a HEPA filter was connected to the inlet of the furnace and the CPC was connected to the

outlet. During measurements, the CPC drew 1 L min-1 of filtered air through the furnace. When the

CPC was not measuring, no flow was present inside the furnace.

Data Analysis. Determining the count median diameter (CMD) of the UF mode in a multimodal

lognormal distribution requires the separation of the modes into single lognormal functions. This

was conducted with the Multipeak Fitting Package in the software IGOR Pro (Wavemetrics Inc., Lake

Oswego, OR). By manually fitting the data to a lognormal distribution, the package determined the

parameters defining individual lognormal modes according to the equation

y ( x )=A exp(−log2( xx0 )d2 ) , (1 )

where x0 is the peak location, d is the width and A is the amplitude. Hence, the different modes

could be separated. For this study, only the ultrafine particle (UFP) mode is relevant.

To determine the CMD for a particular PSD, we used the following formula:

CMD = (∏i=1Dini)

1N ,(2)

where ni is the number of particles for the ith diameter Di, and N is the total number of particles.

When dealing with multiple PSDs of the same sample, the arithmetic mean of the CMDs is taken as

the mean CMD. While three data points are considered to be the lower bound for accurate

arithmetic means, the availability of concrete was not always sufficient to produce three PSDs of

each sample.

Volume fraction remaining (VFR) is a measure of volatility, defined as a ratio of a particles CMD at a

specific temperature and particle size such that

8

VFR=(CMDi

CMD0)

3

.(3)

The VFR of the ith scan was calculated with respect to the initial scan (CMD0), which was conducted at

room temperature. When more than one preselected particle size was measured, an arithmetic

mean of all the VFR values at that particular temperature was also calculated. The uncertainty in VFR

measurements is 6%, taken as 1% for each CMD measurement (Modini, Harris, & Ristovski, 2010).

VFR plots are used to identify chemical species by their volatile properties. Several methods of curve

comparison can be used, and for this study the curves were characterised by their numerical

derivative, as established in other studies (G. R. Johnson, Z. Ristovski, & L. Morawska, 2004),

(Johnson, Ristovski, D'Anna, & Morawska, 2005). The characteristic volatilisation temperature used

to assess similarity between species is taken to be the location of the maximum slope in the

volatilisation curve. To determine this, the average VFR values were fitted to a cubic polynomial

function. From there, numerical derivatives were computed to identify the largest rate of change in

VFR with respect to temperature, providing the characteristic volatilisation temperature for the

specific curve.

3. Results

Particles originating from concrete fracturing. Figures 1 (a), (c) and (e) show that when the concrete

is struck with no fracture event, no particle production was seen below 100 nm. When a mechanical

process in the form of a fracture event was applied to the concrete however, UFPs were seen as a

distinct mode. In each case the PSD exhibited two distinct features; a sub-100 nm mode, fully visible

in the SMPS measurement range (termed the ultrafine mode) and an additional feature at greater

than 100 nm (termed the coarse mode). The latter was so designated because it appeared to be the

tail end of a larger diameter coarse mode with a CMD outside of the instrument’s range. The

ultrafine modes for the three tested samples, RC, RSC and CMLC, were found to have unimodal

peaks with average CMDs of 27 nm, 39 nm and 49 nm respectively.

The VFR analysis shows that the particles comprising the ultrafine mode were volatile, as presented

in Figures 1 (b), (d) and (f). The average VFR for both RC and RSC is similar over the range of 0 to 100

°C, with divergence between 100 and 125 °C. However, the CMLC average shows a much slower

decay over the tested temperature range. From figures 1b,d,f we see that for RC, FrGPs have

evaporated almost completely leaving only a 10% refractory residue by 125 °C, while the RSC VFR

9

left a much larger refractory core comprising about 40% of the original particle volume. The CMLC

VFR is still decreasing rapidly by 125 °C, and as such no comment on a refractory core can be made.

While the VTDMA can measure volatility at higher temperatures, the particle concentration

decreases rapidly causing the measurements to have high uncertainty. The RC and RSC samples

appear to be composed of internally mixed combinations of similar volatile and refractory materials

but in different volume ratios, based on the large fraction of material remaining by the observed VFR

plateau.

The observed behaviour described above suggests several possible formation mechanisms. One

possibility is that the entire particle formed through nucleation of a single volatile species and that

the refractory core formed through the partial oxidisation of this volatile component to a less

volatile oxidation product during thermodenuder treatment. An example would be the partial

charring of an organic species to leave a carbon core. A second possibility is that the particles grew

to their final measured size through the condensation of a condensable but somewhat volatile

species onto a pre-existing core of lower vapour pressure material, in which case this core was

simply smaller in the case of the RC sample than for the RSC sample case.

The former process appears more likely as the refractory core has been observed previously with

VTDMA measurements involving thermo-denuder treatment of particles composed of condensable

organic species (Philippin, Wiedensohler, & Stratmann, 2004). An explanation for the difference in

the refractory residue fraction in the two concrete samples may be due to differences in

thermodenuder residence times. The thermodenuder flowrate for the RC VFR measurements was

half that of the RSC and CMLC VFR measurements, and the RC aerosol residence time in the

thermodenuder was therefore twice as long. This effect would see the RC sample evaporate further

than the others, indicating that at similar flow rates the VFR curves of the two experiments could be

similar.

Concrete Composition Analysis. A chemical analysis of the concrete samples was conducted to

detect volatile substances. For the RC sample, only the compound 2,2,4-trimethylpentanediol

diisobutyrate (TMPD-DIB) was found in substantial amounts. TMPD-DIB is a commonly used

plasticiser for PVC, has a boiling point of 281.5 °C (Cain, Wijk, Jalowayski, Pilla Caminha, & Schmidt,

2005), and has been reported as an emittent of floor constructions (Järnström, Saarela, Kalliokoski, &

Pasanen, 2008). For the RSC sample, several volatile solvents were found present, most prominently

acetone, methyl isobutyl ketone and dibutylphthalate. Chemical analysis conducted on the CMLC

sample showed it to be bereft of large quantities of volatile substances but found trace quantities of

acetic acid, siloxanes and 2-ethyl-1-hexanol. The presence of siloxanes (here,

10

octamethylcyclotetrasiloxane and decamethylcyclopentasiloxane could be identified) is not unusual

for an aerosol measurement laboratory, as the conductive silicon tubing that is used in aerosol

sampling is known to emit organic compounds such as siloxanes (Timko et al., 2009).

Particles Generated in the Furnace. Furnace experiments were carried out for RC concrete samples.

It was found that heating it up to the limits of the equipment (300 °C) produced no FuGPs at all.

Considering the produced FrGPs were volatile below 100 °C, some particle formation was expected.

However, it is of note that the furnace heating method was unlikely to have occurred on the

hypothesised time scale of a fracture, as discussed in section 4.4.

Characterising the Volatility Curves. From Figures 1 (b), (d) and (f), the maximum slopes of the

volatilisation curves for RC, RSC and CMLC samplesoccured at 79.7, 76.6 and 125.0 °C respectively

with R2 coefficients of 0.99, 0.98 and 1.00. Here we see that the RC and RSC characteristic

temperatures are relatively similar. The characteristic volatilisation temperatures of around 80°C is

not necessarily representative of the boiling point of the volatile compound, in fact the material

observed in the thermodenuder may not be the original condensable volatile compound. It may be a

by-product of the manufacturing process or even a new material produced during the thermal

treatment in the thermodenuder. Additionally, for sufficiently small particles, the Kelvin effect would

decrease the measured evaporation temperature of the VOC compared to a bulk sample.

It should be noted that the CMLC VFR curve is incomplete due to a slow decay leaving its end out of

the measuring range. As such, the curve has no point of inflection, and the numerical derivative is

the end point. While this means the total curve cannot be characterised with a volatilisation

temperature, it is still possible to differentiate the curve from the RC and RSC samples, suggesting

different chemical species. Additionally, that the cubic fit of the VFR curves is greater than 1 for small

temperature is an artefact of the polynomial fitting applied and has no physical meaning, and can be

ignored.

Controlling for VOC Adsorption. To ensure that any FrGPs seen were from the sample and not

ambient VOCs that deposited on the surface of the concrete, a control experiment was run using

both RC samples and aluminium foil. RC samples and aluminium foil were left in ambient laboratory

conditions over a 6 week period, to accumulate condensed VOCs. During subsequent heating in the

furnace, the RC samples did not produce any particles, while the aluminium foil produced large

numbers of particles, with concentrations in excess of 80 000 cm-3 by 205 °C. This result was found to

be true for both instantaneous and gradual heating methods. Due to the volatilisation temperature,

it is hypothesised that the observed particles were ammonium sulphate, which has a known

11

volatilisation temperature between 161 and 210 °C (G. Johnson, Z. Ristovski, & L. Morawska, 2004).

As the concrete sample did not appear to retain nor emit any VOCs during this time, it is unlikely that

any observed FrGPs have an airborne source.

4 Discussion

Our results show that the fracture of concrete will indeed give rise to particles in the ultrafine mode.

This is not expected or supported by the current theoretical framework and confirms previous work

of Kumar et al (2012). We must now consider what mechanisms are responsible for the observed

particles.

Proposed Mechanism. As is the case in other processes which form volatile secondary particles, the

likely conditions for the observed ultrafine mode are high temperature and supersaturation.

Concrete fracture is a highly complicated event with many processes occurring throughout a bulk

sample simultaneously. The energy absorbed to create a fracture surface is termed the ‘fracture

energy’ (Wittmann, Mihashi, & Nomura, 1990). Concrete is a quasibrittle material (Bazant & Planas,

1997), and as such the fracture energy is low enough that hammer strikes are capable of creating

fracture fronts. As the fracturing event is extremely quick and the fracture energy relatively small,

the rest of the energy imparted into the concrete is bound to dissipate quickly as heat. When no

fracture occurs, this energy can be dissipated throughout the bulk concrete sample. If, however, a

fracture event does occur the energy is localised at the crack interface. This is hypothesised to create

a sudden localised temperature increase which will evaporate volatile components available near

the crack region. As this event occurs on a fast time scale, the evaporated volatile compounds have

yet to disperse and condense from the supersaturated gas to form secondary volatile particles.

A possible particle formation pathway may involve the creation of an intermediary substance, which

undergoes a chemical change to the produced FrGPs. However, such chemical reactions are typical

of processes in the upper atmosphere which have access to sunlight, and are unlikely to occur in an

isolated system. As such, the likely explanation is that the particle formation seen is either

homogeneous or heterogeneous nucleation. As homogeneous nucleation requires highly

supersaturated conditions, heterogeneous nucleation is usually considered more likely. However, as

the chamber was clean of seed particles when sampling took place, they could only have been

introduced into the system during the fracturing event, a purely mechanical event ignoring the

12

aforementioned UFP production. This implies enough volatile material was vaporised during the

fracture event, high enough for homogeneous nucleation to occur.

Given that the observed particles were volatile and below 100 nm in size, and that there is no known

method for a mechanical process to create volatile secondary particles, the proposed mechanism for

particle production is nucleation, specifically homogeneous nucleation occurring during the instant

of the fracturing event.

Possible Fracture Generated Particle Candidates. The results of the chemical analysis showed a

large variety in trace chemicals in the concrete samples. This includes 2,2,4-trimethylpentanediol

diisobutyrate (TMPD-DIB, BP 281.5 °C) in the RC sample, Methyl isobutyl ketone (BP 117 to 118 °C)

and Dibutylphthalate (BP 340 °C) for the RSC sample and 2-ethyl-1-hexanol (BP 180 to 186 °C) for the

CMLC sample. As the RSC VFR curve plateaues rather than gradually decreasing, it suggestes a two

stage, internally mixed aerosol was likely formed from a volatile component and a second compound

with a higher boiling point. The fact that two substances with vastly different boiling points were

found in the RSC chemical analysis further supports internal mixing of the produced FrGPs in this

instance.

The question of how such solvents were found in concrete samples is relevant, particularly since

they are not required for the production of concrete. Concrete does require chemical admixtures

that aid in smoothing the concrete before setting. Such admixtures usually contain a variety of

superplasticisers such as sulfonated melamine formaldehyde (Neville & Aitcin, 1998) among other

compounds, of which none were detected during the GC/MS analysis of the samples used. In fact,

the CMLC sample was specifically prepared without any admixtures. While pre-packaged cement

manufacturers do not usually list the specific ingredients in an admixture, making it impossible to

know beforehand the exact composition of the final concrete without specific analysis (as was the

case with the RC and RSC samples), volatile chemicals added to the concrete during manufacture

have to be ruled out as the source as particle production was observed for concrete with no

specifically added volatile admixtures (CMLC).

The other source is contamination of the materials used in concrete manufacture. Concrete

production involves the use of bulk materials such as cement, sand and aggregate, which could

easily be contaminated along the supply chain. While it was established earlier that contamination

from organics found in the air is not the source of FrGPs, and then subsequently confirmed by the

chemical analysis, the materials used to make the concrete may have come into contact with a

variety of volatile solvents, particularly as the concrete was created in a regular research laboratory.

13

It is possible that concrete directly from a work site would exhibit different volatile materials in a

GC/MS analysis, and subsequently produce FrGPs with a chemical makeup.

The amount of material needed in the concrete to create the supersaturated conditions for

nucleation is not necessarily high. The cumulative mass concentration for the UFP peak from the

CMLC sample (figure 1e) was approximately 100 ng m-3, which would require only small amounts of

volatile material to be present in the concrete.

Literature Comparison. The results of this study are in agreement with Kumar et al. (2012) and Asadi

et al. (2012), with variation in measured CMDs but consistent UFP production for mechanical

methods. The measured CMD for the CMLC sample of this study was 48.7 nm, which is comparable

to the results for dry concrete at 40 nm by Kumar et al. (2012) (recall that the CMLC sample from

this study was provided by the author of (Kumar et al., 2012)). However it should be noted that the

coarse modes measured in this study were often larger or appeared to extend to higher

concentrations outside of the measurement range than the UFP modes. This differs from (Kumar et

al., 2012), but is similar to the results of (Asadi et al., 2012). The only other studies to have

investigated similar phenomenon are those researching the fracture of nanocomposites, for

example, Sachse et al. (2012). However, this study also found that their reference materials without

nanofillers also produced UFPs in lower concentrations. While this was conducted on the drilling of

polymers, it is possible that a similar mechanism to that proposed in this study is responsible for the

UFP production in this case. The study by Ogura et al. (2013) found that UFPs produced from

grinding polystyrene were volatile, which supports this hypothesis.

Limitations and Future Research Goals. It is of note that attempts to directly measure the fracture

temperature were unsuccessful due to several limiting factors, including fractures occurring over

extremely fast timescales, over small surface areas and under regions of intense pressure. Due to

this, measuring the temperature of the fracture is impossible with currently available technology.

Additionally, collecting the produced FrGPs during fracture for a chemical analysis was not possible

due to the presence of the much larger modes, and low UFP concentrations. Without the fracture

temperature, it is not possible to determine whether the concrete reaches the temperatures

required to vaporise the volatile materials off the concrete.

Using a furnace to heat the concrete to the required temperature was designed to circumvent the

issue, however no particles were detected when heating RC samples to 300 °C, whether done slowly

(several minutes) or quickly (several seconds). As the fracturing event occurs on a much smaller time

14

frame, less than one second, a more rigorous heating method with a high heat conductance may be

required to determine whether directly heating concrete can produce UFPs.

A remaining issue is determining directly what the chemical composition of the particles are, without

having to rely on inference from chemical analysis of the bulk material coupled with volatilisation

temperatures taken from a VFR curve. Future research incorporating hygroscopicity or mass

spectrometry measurements in tandem with volatilisation measurements could aid in identifying the

volatile components.

A noticeable difference between the observed PSDs from the different concrete samples is the

variation in maximum particle number concentration. The RC samples in particular produced PSDs of

varying particle concentrations unlike the other samples, even though the methodology for all the

samples was identical, as outlined in section 2.3.1. This was most likely due to the large number of

variables associated with striking concrete with a hammer. As the only control on the fracturing was

the time spent striking, the number of fractures, their size, location and the force imparted by the

hammer were not controlled. However, as the fundamental particle creation event is hypothesised

to only require a fracture and impact energy over a small period of time, the lack of these controls is

only likely to affect the concentration of the mode not its CMD. As such any conclusions drawn from

the data will not be affected.

As was mentioned in the previous section, applying mechanical methods to the polymers used to

hold nanocomposite materials appears to lead to similar UFP production. If this is to be confirmed, a

study similar to this one needs to be conducted for polymers in rigorously controlled conditions.

Pollution from construction activities can affect entire urban environments. The health effects of

particles are well-documented. Therefore, understanding the origin of these particles is important

for their control and mitigation.

Acknowledgements

The authors would like to thank Dr Xuemei Liu of the QUT Civil Engineering department, as well as

the University of Surrey Construction Materials Laboratory for providing the concrete for this study.

Additionally, the authors would like to thank Mr Casey Roff and Mr Chunlei Li for their help during

laboratory experiments.

15

References

Asadi, S., M Hassan, M., & Dylla, H. (2012). Characterization of Nano Particles Released During Asphalt and Concrete Laboratory Activities. Paper presented at the Construction Research Congress

Azarmi, F., Kumar, P., & Mulheron, M. (2014). The exposure to coarse, fine and ultrafine particle emissions from concrete mixing, drilling and cutting activities. Journal of hazardous materials, 279, 268-279.

Bazant, Z. P., & Planas, J. (1997). Fracture and size effect in concrete and other quasibrittle materials (Vol. 16): CRC press.

Cain, W., Wijk, d. R., Jalowayski, A., Pilla Caminha, G., & Schmidt, R. (2005). Odor and chemesthesis from brief exposures to TXIB. Indoor Air-International Journal of Indoor Air Quality and Climate, 15(6), 445-457.

Colbeck, I. (2014). Aerosol Science: Technology and Applications: John Wiley & Sons.Corn, M. (1961a). The adhesion of solid particles to solid surfaces II. Journal of the Air Pollution

Control Association, 11(12), 566-584. Corn, M. (1961b). The adhesion of solid particles to solid surfaces, I. A review. Journal of the Air

Pollution Control Association, 11(11), 523-528. Faber, P., Drewnick, F., & Borrmann, S. (2015). Aerosol particle and trace gas emissions from

earthworks, road construction, and asphalt paving in Germany: Emission factors and influence on local air quality. Atmospheric Environment, 122, 662-671.

Hinds, W. C. (2012). Aerosol technology: properties, behavior, and measurement of airborne particles: John Wiley & Sons.

Järnström, H., Saarela, K., Kalliokoski, P., & Pasanen, A.-L. (2008). Comparison of VOC and ammonia emissions from individual PVC materials, adhesives and from complete structures. Environment International, 34(3), 420-427.

Johnson, G., Ristovski, Z., & Morawska, L. (2004). Application of the VH TDMA technique to coastal ‐ambient aerosols. Geophysical research letters, 31(16).

Johnson, G. R., Ristovski, Z., D'Anna, B., & Morawska, L. (2005). Hygroscopic behavior of partially volatilized coastal marine aerosols using the volatilization and humidification tandem differential mobility analyzer technique. Journal of Geophysical Research: Atmospheres (1984–2012), 110(D20).

Johnson, G. R., Ristovski, Z., & Morawska, L. (2004). Method for measuring the hygroscopic behaviour of lower volatility fractions in an internally mixed aerosol. Journal of Aerosol Science, 35(4), 443-455.

Kumar, P., & Morawska, L. (2014). Recycling concrete: an undiscovered source of ultrafine particles. Atmospheric Environment, 90, 51-58.

Kumar, P., Mulheron, M., & Som, C. (2012). Release of ultrafine particles from three simulated building processes. Journal of Nanoparticle Research, 14(4), 1-14.

Kumar, P., Pirjola, L., Ketzel, M., & Harrison, R. M. (2013). Nanoparticle emissions from 11 non-vehicle exhaust sources–A review. Atmospheric Environment, 67, 252-277.

McLafferty, F. W., & Tureček, F. (1993). Interpretation of mass spectra: University Science Books.Modini, R. L., Harris, B., & Ristovski, Z. (2010). The organic fraction of bubble-generated,

accumulation mode Sea Spray Aerosol (SSA). Atmospheric Chemistry and Physics, 10(6), 2867-2877.

Neville, A., & Aitcin, P.-C. (1998). High performance concrete—an overview. Materials and structures, 31(2), 111-117.

Ogura, I., Kotake, M., Shigeta, M., Uejima, M., Saito, K., Hashimoto, N., & Kishimoto, A. (2013). Potential release of carbon nanotubes from their composites during grinding. Paper presented at the Journal of Physics: Conference Series.

16

Philippin, S., Wiedensohler, A., & Stratmann, F. (2004). Measurements of non-volatile fractions of pollution aerosols with an eight-tube volatility tandem differential mobility analyzer (VTDMA-8). Journal of Aerosol Science, 35(2), 185-203.

Sachse, S., Silva, F., Irfan, A., Zhu, H., Pielichowski, K., Leszczynska, A., . . . Njuguna, J. (2012). Physical characteristics of nanoparticles emitted during drilling of silica based polyamide 6 nanocomposites. Paper presented at the IOP Conference Series: Materials Science and Engineering.

Timko, M. T., Yu, Z., Kroll, J., Jayne, J. T., Worsnop, D. R., Miake-Lye, R. C., . . . Destaillats, H. (2009). Sampling artifacts from conductive silicone tubing. Aerosol Science and Technology, 43(9), 855-865.

Van Broekhuizen, P., Van Broekhuizen, F., Cornelissen, R., & Reijnders, L. (2011). Use of nanomaterials in the European construction industry and some occupational health aspects thereof. Journal of Nanoparticle Research, 13(2), 447-462.

Wittmann, F., Mihashi, H., & Nomura, N. (1990). Size effect on fracture energy of concrete. Engineering Fracture Mechanics, 35(1), 107-115.

17

18

Figure 1 (a), (c), (e) The particle size distributions for the concrete samples tested and the

associated UF mode lognormal fit. If multiple experiments were conducted, they are

represented here as different colours with the chronological order of experiments being

blue, red, green. (b), (d), (f) The volume fraction remaining as measured by the VTDMA.