Near-source hypochlorous acid emissions from indoor bleach cleaning †

a Cleaning surfaces with sodium hypochlorite (NaOCl) bleach can lead to high levels of gaseous chlorine (Cl 2 ) and hypochlorous acid (HOCl); these have high oxidative capacities and are linked to respiratory issues. We developed a novel spectral analysis procedure for a cavity ring-down spectroscopy (CRDS) hydrogen peroxide (H 2 O 2 ) analyzer to enable time-resolved (3 s) HOCl quanti ﬁ cation. We measured HOCl levels in a residential bathroom while disinfecting a bathtub and sink, with a focus on spatial and temporal trends to improve our understanding of exposure risks during bleach use. Very high (>10 ppmv) HOCl levels were detected near the bathtub, with lower levels detected further away. Hypochlorous acid concentrations plateaued in the room at a level that depended on distance from the bathtub. This steady-state concentration was maintained until the product was removed by rinsing. Mobile experiments with the analyzer inlet secured to the researcher's face were conducted to mimic potential human exposure to bleach emissions. The ﬁ ndings from mobile experiments were consistent with the spatial and temporal trends observed in the experiments with ﬁ xed inlet locations. This work provides insight on e ﬀ ective strategies to reduce exposure risk to emissions from bleach and other cleaning products.

The rst observation of emissions of HOCl from the use of bleach-based cleaners was reported in 2017 using chemical ionization mass spectrometry (CIMS), when a laboratory oor was mopped with an aqueous solution of a commercial bleach product; room-averaged peak HOCl levels on the order of 100 parts per billion by volume (ppbv) were reported. 1Similar levels (15-370 ppbv) of HOCl were reported following mopping the oor of a test house with bleach. 17Finally, cleaning surfaces in a university athletic center with a dichlor (dichloro-s-triazinetrione) spray was reported to form HOCl and other chlorinated species. 38These studies demonstrate that cleaning surfaces with bleach and similar cleaning products can emit high levels of HOCl.However, further study is required to relate observed emissions to exposure.
It is reasonable to suspect that concentrations of emitted species from bleach and other cleaning agents are highest close to the point of use, 39 and that spatial distributions within a room or building may be very heterogeneous due to slow mixing of air.In performing exposure risk assessments, it will be important to take total emissions and occupants' positions relative to the cleaned surfaces into account for the duration of elevated HOCl levels following bleach use.In this work, we quantify temporally-and spatially-resolved HOCl mixing ratios when cleaning surfaces in a residential bathroom with a commercial bathroom bleach product.This study introduces a novel cavity ring-down spectroscopy (CRDS) method to conduct time-resolved (∼3 s) HOCl quantication in a domestic setting.This work will improve our understanding of exposure risk to HOCl and secondary products during bleach use and may help establish mitigation strategies.

Sampling location
Measurements were performed in an apartment bathroom (7.4 m 3 ) in Saskatoon, Saskatchewan, Canada.The unit layout is shown in the ESI (Fig. S1).† Prior to and during cleaning, all apartment windows were closed, all fans in the apartment were turned off, and all interior doors were open unless otherwise noted.Air change rates (ACR) were not measured during these experiments.We assume that the ACR is similar to the reported average of ∼0.5 per hour (h −1 ) for residential buildings because exterior doors and windows were closed during most experiments. 40There was also no forced air circulation or mechanical ventilation in the unit during cleaning experiments, except for a communal duct from the building constantly venting all bathrooms.Air change rates would likely be driven by indooroutdoor temperature gradients and leaks in the building envelope and around window and door casements. 40athroom lighting was provided by three incandescent light bulbs that were covered by a white, plastic, top-open shade positioned approximately 117 cm above the sink.An air vent was present on the wall at the bathtub rear, 208 cm above the bottom of the bathtub.There were no windows in the bathroom; windows were located in the bedroom and living room (Fig. S1 †).

Cavity ring-down spectroscopy
Gas-phase HOCl mixing ratios were measured in real-time using cavity ring-down spectroscopy (CRDS; Picarro PI2114 Hydrogen Peroxide (H 2 O 2 ) Analyzer; Picarro Inc.).This instrument was designed to selectively quantify gas-phase H 2 O 2 mixing ratios using the 2n OH overtone stretch at ∼1400 nm.This overtone absorption is observed when the OH stretch is excited from the ground vibrational state to the 2nd vibrational excited state (n = 0 to n = 2).A detailed discussion of cavity-enhanced spectroscopy methods, including CRDS, is provided in Young et al. 41 We have previously demonstrated the application of this instrument for indoor air quality research by measuring time-resolved H 2 O 2 mixing ratios during surface cleaning with a commercial H 2 O 2 -based cleaning solution. 39The 2n OH stretch of HOCl is also at ∼1400 nm. 42The soware of the PI2114 analyzer was modied to include the HOCl absorption in the least-squares analysis of the absorption spectra acquired by the spectrometer in its normal operating mode.
We are not aware of any quantitative spectroscopy of HOCl in the near infrared.Cavazza et al. 42 have analyzed the frequency spectrum of the 2n 1 band, presenting spectroscopic parameters and lists of assigned transitions, but they do not report any line intensities.We therefore acquired our own spectra of HOCl in the frequency region probed by the Picarro PI2114.The work proceeded in two steps.In the rst step, a PI2114 instrument was used to acquire high resolution spectra of a vapour sample from the headspace over a solution of sodium hypochlorite in water, as well as spectra of pure water in the same frequency region.Removing the known empty-cavity base line and the absorption due to water vapour yields the HOCl spectrum shown in Fig. S2 in the ESI.† The H 2 O 2 spectrum is also shown in Fig. S2 † for comparison.All the strong absorption features of the HOCl spectrum align with frequencies reported by Cavazza et al. 42 One sees that the HOCl and H 2 O 2 spectra have some overlapping absorption, but also very distinct features, so it is straightforward to distinguish the absorption of these two compounds by least-squares curve tting.
The second calibration step enabled HOCl quantitation.Zero air was bubbled through a sodium hypochlorite buffer solution at 50 standard cubic centimetres per minute (sccm) and the ow was diluted with dry zero air to vary the HOCl concentration.Detailed calibration information is provided in the ESI.† The method limit of detection (LOD), calculated as 3s above the mean signal when well-mixed ambient apartment air was not inuenced by cleaning activities and HOCl was assumed to be absent, was determined to be 20 ppbv for unaveraged 3 s resolution data and 10 ppbv for 16 s averages.
This CRDS instrument is, to our knowledge, the only instrument other than CIMS that can quantify HOCl at the ppbv level with temporal resolution of seconds to minutes.The instrument response time measured during calibrations was on the order of 4 s during increases in HOCl mixing ratios on the order of 10 ppbv and ∼10 s during decreases.When adding our PFA inlet in this work, partitioning of HOCl was expected to potentially impact the response time based on the high Henry's Law constant (K H ) for HOCl (∼660 M atm −1 ), 43 which would theoretically result in a tubing-driven delay of ∼5-10 s in response time.Our observations of ∼18 s response to the removal of >1 ppmv HOCl from ambient air (see Table 2) are consistent with this, and are similar to predicted CIMS response times for species with similar K H . 44 While CIMS has the advantage of being able to speciate numerous chlorinecontaining species associated with bleach use, its applicability to indoor measurements is limited by several factors.These include safety concerns due to radioactive ion sources, high expense, and the need for specialized training for operation, maintenance, and data analysis.The CRDS instrument does not suffer from any of these limitations.Further, it is portable, quiet, and can operate with infrequent calibration and maintenance over weeks or months at a time.This modied CRDS instrument is ideal for indoor measurements where minimizing disruption to occupants is desirable.This manuscript demonstrates its applicability to indoor air quality measurements.

Sampling conditions for hypochlorous acid measurements
The analyzer was located outside of the bathroom, approximately 2.1 m from the open bathroom door (Fig. S1 †).Bathroom air was sampled through a 4.7 m PFA tubing inlet with 3/16 ′′ ID at a ow rate of 1 standard litre per minute (slpm).The sampling inlet was positioned at three different locations in the bathroom for stationary measurements (Table 1) and secured to a pair of glasses worn by the researcher for mobile measurements.When secured to the glasses, the sampling inlet was positioned ∼6 cm in front of the face of the researcher.During mobile measurements, the sampling inlet was secured to the face of the researcher at all times while they were in the bathroom.Reported HOCl mixing ratios during mobile measurements are assumed to be representative of those in inhaled air.When the researcher exited the bathroom during mobile cleaning events, the sampling inlet was placed on the analyzer outside the bathroom.

Cleaning activities
A commercial bleach cleaner containing 2.0% w/w sodium hypochlorite (NaOCl) was used to clean bathroom surfaces, with the main focus on the surface of a porcelain bathtub (∼2 m 2 ) in this work.A sufficient volume of cleaning solution was sprayed into the bathtub to fully coat the surface, and additional cleaning solution was applied to areas with visible soiling.The volume applied ranged from 40-85 mL, achieved with 80-170 pumps of the spray bottle.The product was le on the bathtub surface for a minimum of ten minutes, as per the product manufacturer's instructions.The bathtub was then scrubbed with a water-dampened brush which was occasionally rewetted.Scrubbing lasted 4-7 min; the duration was determined by the level of visible soiling of the bathtub surface.Aer scrubbing, the bathtub was rinsed with tap water (∼2 min) and the resulting solution entered the drain.
Bleach cleaner was also used to clean the sink during stationary and mobile measurements.The sink was cleaned similarly to the bathtub except that smaller volumes of bleach product were used.The entire sink was sprayed with approximately 25-35 mL achieved by 50-70 pumps of bleach product.The solution was le to sit for a minimum of ten minutes.Then the sink was scrubbed with a water-dampened brush and rinsed with tap water.Scrubbing and rinsing lasted ∼4 min, with the resulting solution entered the drain.

Data processing
The data collected from the CRDS was processed and graphed using IGOR Pro 8.04 (WaveMetrics, Inc.; Portland, USA).All time series presented in this work depict data collected at a time resolution of 3 s.The HOCl decay rate constant (k) was determined by tting a single exponential decay to the period of decreasing HOCl mixing ratios following their peak for each cleaning event (eqn (1)), where [HOCl] min is the measured background mixing ratio of HOCl, [HOCl] max is the peak mixing ratio, t is time, and t max is the time at which the peak mixing ratio was observed.

Stationary measurements during bathtub cleaning
Hypochlorous acid mixing ratios increased during bathtub cleaning events regardless of inlet location.Fig. 1 shows representative time series for cleaning events with the inlet at near-ceiling, mid-room, and near-tub locations (Table 1).Irrespective of inlet location, HOCl levels increased and then plateaued following bleach application to the bathtub surface.The liquid coating on the bathtub surface dried over the course of ∼5-10 min, leaving a visible residue.Elevated HOCl levels persisted until the bleach product was removed from the bathtub by scrubbing and rinsing.A sharp increase in HOCl View Article Online mixing ratios was consistently observed during scrubbing for the near-ceiling and mid-room locations, while no signicant increase was observed at the near-tub level.These peak HOCl levels began decreasing within 2 min via a rst order exponential decay to below the LOD.These temporal proles are quite different from those reported in the literature for bleach cleaning; other studies uniformly report a rapid increase of HOCl following bleach application followed by a rapid decrease to below the LOD, with no plateau. 1,17The different temporal proles may be due to several factors.Previous studies mopped the oor with aqueous sodium hypochlorite solutions and researchers had no further physical interactions with the cleaned surface, whereas we sprayed the solution onto the surface.The composition of the cleaning solutions was also somewhat different.While NaOCl was the active ingredient, additives may affect the partitioning of HOCl.For example, the bleach cleaner used in this work was deposited as a foam that evaporated to leave a thin residue.We hypothesize that the bleach product used in this work le a long-lasting HOCl reservoir on the bathtub surface, and that the observed plateau reects the steady-state HOCl mixing ratio (with loss terms including dilution, surface partitioning, and removal via air change).The spike in HOCl mixing ratios during scrubbing is likely due to rapid transfer of HOCl to the gas phase.The act of scrubbing mechanically generates increased air movement and high surface area droplets, both promoting HOCl transfer to the gas phase.Cleaning solution is also moved across more of the tub surface during scrubbing, particularly onto the bathtub walls where thin lms can form that would undergo rapid evaporation and release of HOCl.The rapid decay in HOCl following scrubbing is due to the removal of the reservoir as the bleach was washed down the drain.
Assuming that HOCl at the mid-room and near-ceiling locations is well-mixed (see discussion below), we used the rate constant for HOCl loss and the steady-state concentrations (shown in Table 2) to calculate an HOCl emission rate of 0.8 ppbv s −1 .The total number of emitted HOCl molecules depends on the duration of the experiment; if the tub is rinsed 10 minutes following application, 8.8 × 10 19 molecules will be emitted, while 3.5 × 10 20 molecules will be emitted over 40 minutes of exposure.Wong et al. 1 reported that 4.2 × 10 21 HOCl molecules were emitted following mopping 37 m 2 of a oor with ∼200 mL of a bleach solution.As discussed above, the temporal proles in the two studies were quite different, with no sustained steady-state observed by Wong et al. 1 Based on our calculations, ∼4% of total HOCl molecules partition to the air over 40 min exposure in our experiments.This assumes that all OCl − is converted to HOCl due to exposure to CO 2 in the air and interaction with acidic species at the air-solid interface.The extent of protonation following application to surfaces is not known, although it has been reported to be much more extensive than in the headspace of an aqueous NaOCl solution. 1For comparison, Wong et al. 1 predicted that ∼60% of available HOCl molecules were emitted to the gas phase in their study.In that work, the number of HOCl molecules emitted increased with each consecutive mopping experiment. 1 The authors hypothesized that each cleaning event removed reactive materials that acted as HOCl sinks. 1 Emitted HOCl levels did not depend on the order of cleaning experiments in the current study.This is likely due to the longer period of at least 9.5 h between cleaning leading to random variations in the level of soiling on the bathtub surface for each experiment.
While the general shapes of the HOCl temporal proles are similar at each inlet location, Table 2 shows that there were differences in magnitude and timing.Mean steady-state HOCl mixing ratios were the same at the mid-room and near-ceiling positions within experimental uncertainty based on comparison by paired Student's t-test, and were within the range of peak mixing ratios reported elsewhere following oor mopping with bleach solutions (15-370 ppbv). 1,17The mean steady-state mixing ratio of 21 ppmv measured near the bathtub exceeds all prior reported mixing ratios following bleach surface cleaning by at least a factor of 50, and is larger than the mean steadystate mixing ratios measured at the other inlet locations in this work by over a factor of 100.The high HOCl steady-state mixing ratios at the near-tub inlet location suggest that HOCl emitted from the bathtub surface is not well-mixed at this distance.The high decay rate constant of 204 h −1 measured at the near-tub inlet location is likely dominated by dilution, compared to lower rate constants of ∼23 h −1 at the mid-room and nearceiling locations.Loss at these locations is likely dominated by air change and surface partitioning and/or reaction.The similar steady-state mixing ratios and decay kinetics at the midroom and near-ceiling locations lead us to hypothesize that air in the bathroom is relatively well-mixed within 2.1 min following cleaning.
While steady-state HOCl mixing ratios were similar at the mid-room and near-ceiling locations, differences were noted in the spikes observed during scrubbing ("peak HOCl during scrubbing" in Table 2).The magnitude of the peak was generally larger at the near-ceiling location with a mean of 910 ppbv compared to 220 ppbv at the mid-room location.The relative increase in HOCl mixing ratiothe peak level compared to the mean steady-state levelwas a factor of 8.5 larger at near-ceiling compared to a factor of 1.5 at mid-room.These differences may be due to air being drawn toward the vent mounted near the ceiling, but we do not wish to speculate further here due to the large variability in the magnitude of the peaks at the nearceiling location (76% relative uncertainty).In comparison, peak HOCl levels agree at the 95% condence level for the midroom and near-ceiling locations based on comparison by paired Student's t-test.

Mobile measurements during bathtub cleaning
To investigate HOCl exposure during cleaning, mobile measurements with the inlet attached to glasses worn by the researcher were performed to measure near-face HOCl mixing ratios (Fig. 2 and Table 3).The HOCl levels were sometimes below the LOD aer bleach cleaner application because the researcher le the bathroom within ∼1 min, such that the HOCl could not fully mix in the bathroom air.The HOCl levels were below the LOD when the inlet was outside the bathroom for all mobile experiments.The researcher re-entered the bathroom ∼3 min aer bleach application during one experiment; nearface levels of 300 ± 160 ppbv were observed during this period (Fig. 2).Given that only one mid-room measurement was conducted during mobile experiments, we do not speculate on potential differences between mobile and stationary measurements.With that said, it is worth discussing ways in which the presence and actions of a person could affect indoor HOCl concentrations, especially in a relatively small volume of air.6][47] Other human-related factors could have the opposite effect of increasing near-face concentrations.0][51] Further study of near-face concentrations of cleaning product emissions should be performed, as they could be quite different (either higher or lower) than concentrations at similar locations in the absence of people.This could be an important consideration in assessing risks of dermal and inhalation exposure.
As with stationary measurements, the highest near-face HOCl levels were measured during bathtub scrubbing, when HOCl levels spiked to approximately 2.6 ppmv before decaying on the order of 10 min.Peak near-face HOCl levels during scrubbing were approximately 9 times lower than peak HOCl for stationary near-tub measurements.The lower peak HOCl levels measured during mobile experiments are likely due to a combination of the inlet being ∼50 cm above the bathtub surface for mobile measurements compared to 30 cm for stationary measurements, and the inter-experiment heterogeneity observed in the latter (Table 2).These observations suggest that HOCl exposure will depend on the person's proximity to the bleach source, time spent inside the space where bleach is applied, the release of HOCl during scrubbing, and loss to surfaces and rinsing the solution down the drain.

HOCl measurements during sink cleaning
Mixing ratios of HOCl were also measured during the cleaning of a 0.3 m 2 porcelain sink (Table S3 †).One stationary measurement with the inlet at the near-ceiling position (180 cm from the sink) was performed during which the HOCl mixing ratios remained below the LOD.Three mobile measurements were performed, with the inlet 60-90 cm from the sink surface during spraying and scrubbing.The mean mixing ratio during spraying was 17 ± 15 ppbv (over a total of 76 data points), which was below the LOD of 20 ppbv.However, HOCl levels frequently exceeded the LOD, indicating that HOCl was emitted from the sink surface.The low levels emitted from the sink compared to the bathtub are likely due to the smaller surface area being cleaned and the smaller volume of bleach product applied.Mixing ratios increased during scrubbing with a peak level of 370 ± 330 ppbv.This is much smaller than peak levels of 2 ppmv observed during mobile bathtub cleaning experiments.These results suggest that exposure risk can be substantially reduced by cleaning smaller surface areas at any one time and/ or leaving the room.

Implications
We have modied a portable H 2 O 2 CRDS to quantify time resolved HOCl levels with a detection limit of 20 ppbv at 3 s time resolution and 10 ppbv for 16 s averages.This instrument is simple to use, easily transportable, and requires little attention over weeks or months of continuous measurements.The small footprint and low noise generated from the vacuum pump also make this instrument ideal for use in occupied spaces.The presented method provides unprecedented capabilities to study indoor atmospheric processes in occupied spaces where bleach solutions are used.
Measurements in an occupied residence provide insight into exposure during common cleaning practices with a bleachbased product.Measurements at three distances from the Fig. 2 Representative time series for a mobile bathtub cleaning event.Dark purple denotes when bleach cleaner was sprayed in the bathtub, grey when the researcher entered the bathroom but did not perform cleaning activities, and light purple when the bathtub was scrubbed and then rinsed.The distances (cm) above the highlighted activity indicate the mean distance of the mobile inlet to the bathtub surface.Unshaded regions are when the researcher and inlet were outside the bathroom.
bathtub illustrate spatial and temporal heterogeneity of HOCl levels within the bathroom.Aer bleach application, HOCl mixing ratios remained elevated at all measured distances for up to 40 minutes until the bathtub was scrubbed and rinsed.Mixing ratios measured more than 1 m from the bathtub surface were consistent with reported concentrations in wellmixed rooms. 1,17However, HOCl levels closest to the point of application reached peak levels of approximately 20 ppmv, more than 50 times higher than those previously reported.Similar results were obtained during mobile measurements during which the instrument inlet was affixed near the researcher's face, with HOCl levels of ∼3 ppmv detected when the researcher's face was within 1 m of the bottom of the bathtub.These results demonstrate that exposure to HOCl and other potentially harmful reaction products from bleach use can vary greatly even within a small room.People who spend time in close proximity to surfaces to which bleach is applied, such as janitorial and cleaning staff, will be exposed to much higher levels compared to other people in the same room, even if they are only separated by a few metres.Our results suggest that the largest exposures will be during scrubbing of bleached surfaces due to close proximity of the person cleaning the surface combined with increased HOCl levels during scrubbing.This work demonstrates the importance of cleaning methods on exposure outcomes.Previous studies, in which oors were mopped with a dilute bleach solution, showed a sharp rise in HOCl levels followed by a rapid decay. 1,17In those experiments, the total duration of elevated HOCl levels was on the order of 5-20 min, whether the experiment took place in a laboratory (ACR = 13 h −1 ) or a test house (ACR = 0.6 h −1 ). 1,17he protocol for disinfecting (rather than cleaning) with bleach, as denoted by the Centers for Disease Control and Prevention (CDC) recommends a surface contact time of at least 1 min, while the manufacturer of the product used in this work recommended a minimum of 10 min. 52Our results demonstrate that high steady-state HOCl levels will be maintained until the bleach solution is removed from the surface through scrubbing, rinsing, and/or evaporation.In the bathroom used in these experiments, the steady-state HOCl levels observed imply a balance between emission and loss processes, such as surface deposition/reaction and ventilation over time periods of at least 40 min.
This work provides important insight into HOCl exposure during common cleaning practices.It will be important to relate exposure to health risk to determine the urgency of implementing harm mitigation strategies.The effects of HOCl exposure on human health are not well known.Hypochlorous acid is suspected of causing skin irritation, and has been reported to react with squalene, forming toxic chlorohydrins at 600 ppbv, which further suggests its potential for skin irritation. 24Inhalation risks for this species have not yet been reported. 2,53Based on our observations, we suggest that more research into the health effects of gas-phase HOCl should be undertaken and that occupational long-and short-term exposure limits should be established.8][19][20][21] A concerning concomitant emission is Cl 2 , which has been reported to attain peak mixing ratios around 20% of those observed simultaneously for HOCl following mopping oors with bleach. 1,17Chlorine gas can be lethal aer ∼1 h exposure at >30 ppmv, generating corrosive compounds capable of dissolving lung tissue; it can also severely damage the eyes and skin. 16,54Recommended/permissible acute 15 minute exposure limits for Cl 2 range from 0.4-1 ppmv. 54ssuming similar temporal and spatial trends for Cl 2 as HOCl in this work, 1,17,55 we might expect sustained Cl 2 levels as high as 4 ppmv near surfaces on which bleach has been applied.This could easily result in 15-minute exposure exceedances when bleach is used as a disinfectant.While further study into exposure risks is needed, we suggest reducing exposure to bleach emissions by limiting the time such solutions reside on surfaces (e.g., 1 min CDC recommendation for disinfection), using products in well ventilated spaces as recommended by the manufacturer, and limiting time spent in rooms where bleach has been applied to surfaces.View Article Online

Fig. 1
Fig. 1 Representative time series of HOCl mixing ratios during stationary measurements of bathtub cleaning events with the inlet at (A) nearceiling, (B) mid-room, and (C) near-tub positions.Time intervals during which the cleaner was sprayed onto the bathtub surface are denoted by dark purple, and subsequent time intervals for scrubbing and rinsing of the bathtub (always in that sequence) are denoted by light purple.

Table 1
Inlet placement for stationary measurements Approximate centre of bathroom, 203 cm above the oor, 20 cm below the ceiling Mid-room 165 Approximate centre of bathroom, 104 cm above the oor, 119 cm below the ceiling Near-tub 33 Directly over bathtub, 39 cm above the oor; 184 cm below the ceiling Environmental Science: Processes & Impacts Paper Open Access Article.Published on 05 January 2023.Downloaded on 11/1/2023 12:09:19 AM.This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

Table 2
HOCl levels, detection time delays, and HOCl decay kinetics for stationary measurements."N" is the number of compiled cleaning experiments a 6 data point (16 s) averages centred on the maximum signal.

Table 3
Mean HOCl mixing ratios during mobile measurements a Numbers in parentheses denote the number of trials in which HOCl levels exceeded the LOD.b Numbers in parentheses denote the total duration (in minutes) of measurements.cReportedHOCl levels are peak levels, calculated the same as in Table2.Paper Environmental Science: Processes & Impacts Open Access Article.Published on 05 January 2023.Downloaded on 11/1/2023 12:09:19 AM.This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.