Development of a Bioaerosol single particle detector (BIO IN) for the Fast Ice Nucleus CHamber FINCH

We present the setup and first tests of our new BIO IN detector [Bundke, et al., 2009]. This detector is designed to classify atmospheric ice nuclei (IN) for their biological content. Biological material is identified via its auto-fluorescence (intrinsic fluorescence) after irradiation with UV radiation. The use of auto-fluorescence of biological material is a common principle used for flow cytometry, fluorescence microscopy, spectroscopy and imaging. The detection of auto-fluorescence of airborne single particles demands some more experimental effort. However, expensive commercial sensors are available for special purposes, e.g. size distribution measurements. But these sensors will not fit on the specifications needed for the FINCH IN Counter (high sample flow of up 10 LPM). The newly developed -low costBIO IN sensor uses for its fluorescence channel a single high-power UVLED instead of much more expensive UV lasers and is designed to be coupled to the Fast Ice Nucleus CHamber FINCH. If one particle acts as ice nucleus, it will be at least partly covered by ice at the end of the development section of the FINCH chamber. The detector combines an auto-fluorescence detector and a circular depolarization detector for simultaneous detection of biological material and discrimination between water droplets, ice crystals and not activated large aerosol particles to identify ice nuclei. Other key advantages of the new sensor are the low costs, compact size, and the little effect on the aerosol sample which allows it to be coupled with other instruments for further analysis. The instrument will be flown on one of the first missions of the new German research aircraft “HALO” (High Altitude and LOng range).


Introduction
Ambient aerosol originates from multiple sources and consists of a wide variety of materials like mineral dust, sea salt, ammonium sulphate, acids, soot, organic polymers, plant debris, pollen, bacteria and spores.Several recent publications focus on the sources, distribution and potential impact of aerosol particles of biological origin on atmospheric processes like the formation of clouds and precipitation (Ariya and Amyot, 2004;Deguillaume et al., 2008;Georgakopoulos et al., 2009;Mohler et al., 2007;Mohler et al., 2008;Pratt et al., 2009;Prenni et al., 2009;Szyrmer and Zawadzki, 1997;Morris et al., 2008).Biological particles have the ability to act as cloud condensation nuclei (CCN) and as ice nuclei (IN) and induce heterogeneous freezing even at high temperatures (Jaenicke, 2005;Jaenicke et al., 2007).Some biological material is known to be active as IN even at relatively high temperatures of up to −2 • C.They can be bacteria (e.g.pseudomonas syringae) (Vali and Schnell, 1975;Vali et al., 1976;Schnell and Vali, 1976;Maki and Garvey, 1975;Maki et al., 1974), pollen (von Blohn et al., 2005;Diehl et al., 2001;Diehl et al., 2002), fungal spores, etc. Secondary ice formation e.g. by mechanical fragmentation, splintering during riming of ice particles (Hallet-Mossop process) and fragmentation of large droplets during freezing will multiply the number of ice crystals in clouds by up to a factor of 10 000 (Heymsfield and Mossop, 1984;Mossop, 1985;Pruppacher and Klett, 1996).Thus, even a small number of IN will have the potential to alter the microphysical structure of a cloud and play a key role in mid-latitude precipitation formation by the Bergeron-Findeisen process.We have designed the new BIO IN sensor to analyze the abundance of IN of biological origin.The sum of the different emission spectra in the 400-600 nm range will give the signal strength measured by the BIO IN detector.Figure modified from (Palumbo and Pratesi, 2004).(Palumbo and Pratesi, 2004).
The interest in the detection of particles of biological origin in ambient aerosol has greatly increased during the last decade.A method to distinguish biological material from non-biological particles utilizes the detection of autofluorescence (intrinsic fluorescence) after irradiation with UV light (Seaver et al., 1999;Pan et al., 2009;Pan et al., 2003;Hairston et al., 1997).Ho (Ho et al., 2000) compared fluorescence measurements (UV-APS) to reference sampler data that provide culturable or living bio aerosol number concentrations.Biological matter consists of various different substances.Some of them are efficient fluorophores.For the excitation of auto-fluorescence of these substances, two UV excitation wavelength ranges are common: 260-280 nm and 340-380 nm.
(a) Wavelength range 260-280 nm Protein fluorescence is induced by absorption of radiation in this wavelength range.The aromatic side-chains of the amino acids tryptophan, tyrosine and phenylalanine are primarily responsible for the inherent fluorescence of proteins.Tryptophan is the main fluorophore among them, since both its molar absorptivity and the fluorescence quantum yield are typically several times higher as compared to the other amino acids.Therefore absorption spectra of proteins often resemble the absorption of tryptophan, and their maxima are mostly located at about 275-280 nm (Wetlaufer, 1962).The corresponding protein fluorescence emission spectra vary with the protein-solvent environment and peaks are typically located in the region around 330-350 nm (Demchenko, 1986).Proteins are ubiquitous in all biological matter including not vegetative forms and cell fragments, which makes them to universal biomarkers.However we did not select this UV range for excitation because false counts caused by non-biological aromatic hydrocarbons (e.g.present in soot) will occur, leading to over-estimation of the number concentration of bioparticles.b) Wavelength range 340-380 nm Apart from the proteins, there are several enzymatic co-factors present, which act as auto-fluorophores at longer absorption wavelengths, namely the reduced forms of the nicotinamide adenine dinucleotide (NADH (Coenzyme 1)) and derivatives (e.g.NADPH) and the flavins (e.g.Riboflavin (Vitamin B2), flavin adenine dinucleotide (FAD) see Fig. 1.Both classes of substances are involved in the metabolism of cells (citric acid cycle).The intermediates with the nicotinamide group in the reduced form (NADH, NADPH) are very efficient fluorophores, while their oxidized forms (NAD + , NADP + ) are non-fluorescent.Therefore the fluorescence emission from vegetative cells is much higher than that from others with minimized metabolism, like bacterial spores.This makes the blue NAD(P)H fluorescence emission band (centered around 450 nm) a marker for viability of cells (Hairston et al., 1997).However, this effect may be partly compensated by an increase of the FAD green fluorescence emission band (centred around 540 nm), because the redox state of the FAD cofactors is in equilibrium to the mitochondrial NAD + /NADH pool, and the most efficient flavine fluorophores lipoamide dehydrogenase and electron transfer flavoprotein are highly fluorescent in their oxidized forms (Huang et al., 2002).
The UV LED in our BIO IN detector operates at 365 nm, and the fluorescence wavelengths range from 400 nm to 600 nm collecting the emission from both NADH and flavins.We do not expect to be able to distinguish cells or fragments which show no or very low fluorescence, which might be caused by deterioration or ageing.When coupled to FINCH it might also be the case that a very small biological particle acts as IN and grows to detectable size, but its overall fluorescence is too small to be classified as biological.For these reasons it is clear that we will be able to detect only a lower limit of viable biological IN.
Although fluorescence from inorganic material like glass or kaolin cannot be excluded, fluorescence from these material is more than one order of magnitude weaker than that of biological material (Sivaprakasam et al., 2004) and is supposed to be negligible.

Experimental setup
The single particle fluorescence detector for the in situ measurement of biological ice nuclei is based on the optical detector of the Fast Ice Nucleus CHamber (FINCH), which has been previously described in detail (Bundke et al., 2008).Briefly, in FINCH a sample flow of ambient aerosol is mixed with a warm and moist air flow as well as with a cold and dry flow in the mixing region of the instruments chamber, where IN particles are activated at well-defined freezing temperatures and supersaturations (see Fig. 2).The particles grow while flowing through the processing chamber of 1.35 m length.All ice particles pass a virtual impactor, which removes 90% of the gas flow, including non-activated aerosol particles and small water droplets.The optical detector, which discriminates water droplets from ice crystals, is mounted behind the outlet nozzle of the impactor.Here, the different depolarization behavior of water and ice with respect to scattered circular polarized light is used by analyzing the P 44 /P 11 ratio of the scattering matrix.The collimated beam of a 30 mW 635 nm cw (continuous wave) diode laser passes a circular polarizer and is mildly focused to an elliptical spot of 2.5×1 mm, which crosses the 2 mm diameter gas stream perpendicular (see Figs. 3 and 4).Light which is backscattered by droplets and ice crystals is collected in the range of 100 • up to 130 • scattering angle.After a condenser lens a quarter waveplate converts the circular polarized light back into linear polarized light, which is either 45 • or 135 • polarized from the principal axis, depending on the direction of circular polarization.A beam splitting cube separates the two components P RC and P LC of the linear polarized light.The signals P RC and P LC are detected separately with two photodiodes.It is possible to discriminate water droplets and ice crystals by determining the normalized P 44 /P 11 ratio of the scattering matrix, which is given by Eq. ( 1) for this particular case (see Bundke et al., 2008 for details).
The only modification in this section of the detector is an additional 600 nm long-pass filter, which is placed between the quarter waveplate and the beam splitting cube in order to avoid miscounting by scattered UV and fluorescence light.We use the total scattered signal (sum of both channels P RC + P LC ) as trigger for the fluorescence detection.The UV source used for the excitation of the bioaerosol particles is a fiber-coupled high-power LED with 250 mW output at 365 nm, operating in cw mode.The UV light is focused by an aspheric two-lens collimating focusing assembly with 360 nm band-pass emission filter to a circular spot of about 2.5 mm diameter, which overlaps with both the 2 mm diameter cross section of the gas stream and the elliptical spot of the circular polarized 635 nm scattering laser (see Fig. 3).
The fluorescence light is collected perpendicular to the UV beam axis by an aspheric condenser lens and an additional spherical reflector.Two optical filters separate the fluorescence light from the scattered light of both light sources.After this a second lens focuses it to a photomultiplier detector.Additionally a photodiode sensor is mounted in the forward scattering direction (35 • ±5 • ) behind a 400 nm shortpass filter and a focusing lens.The signals obtained in this UV scattering channel are broader and less sensitive than that of the depolarization scattering channels.Therefore we use it mainly for coincidence measurement with the depolarization scattering channels to check the alignment of the beams.Additionaly the background signal can be used to monitor the intensity of the UV LED and normalize the signals to the LED Power.
The signals are sampled in parallel by using a NI PCI 6132 data acquisition module with up to 3 MS/s per channel and are analyzed by a peak detection algorithm using the Lab-VIEW data acquisition software (National Instruments Corporation, Austin (TX), USA).While particles are detected using the circular depolarization detector, bioaerosol particles are identified and counted, if the fluorescence signal extends a significant threshold value (5σ of the noise, see example signal Fig. 6).The ratio of the number of bioaerosol particles to the total number of particles is calculated.

Results
In this section we describe the initial laboratory tests of the newly developed sensor with silica test particles as well as a first sampling of ambient aerosol.

Initial laboratory tests
Initial tests were performed on aerosol produced from a mixture of ca.50 % fluorescent 10 µm spherical silica particles ("Fluo-Blue", Spherotech, Lake Forest (IL), USA) and 50 % non-fluorescent 10 µm spherical particles ("Silica particles", Spherotech, USA) in water suspension.The fluorescent dye has its excitation maximum at 350-360 nm, and the fluorescence emission wavelengths range from 400 to 560 nm.Particles were suspended using a home-made atomizer and were dried by mixing with a dry carrier gas.The total gas flow was about 6 LPM (liter per minute), which is consistent with the flow rates of FINCH after the development chamber and virtual impactor in normal operation mode.For the ini- tial tests the detector was decoupled from FINCH, because silica particles are hardly IN active (See Fig. 5 for details of the setup).
Figure 6 shows a signal graph of two particles passing the detector within one second.Here the photomultiplier was operated with a lower gain.The first (smaller) particle shows a significant fluorescence signal and the second particle almost no fluorescence signal.
In general the scattering intensities of the test particles were found to be relatively broadly distributed in the frequency of occurrence histogram.The two histograms in Figs.7 and 8 show the particle numbers in relation to their scattering and fluorescence intensities, respectively.The relatively broad distribution in Fig. 7 is a little surprising on the first look.However, a check of the aerosol size distribution using an aerodynamic particle sizer (APS, TSI Inc. type 3321) showed that our test aerosol was not monodisperse.The size distribution in front of the detector was found to be similarly broad, with its maximum at about 8.5 µm and a small secondary maximum at 13 µm.This is consistent with the specifications of the distributor of the test particles (not monodisperse, average size of the particles 10 µm).
The distribution of the fluorescence intensity seen in the related histogram was similarly broad, which can be explained with both the broad size distribution and the different dye concentrations in the silica particles.Fig. 9 shows a scatter plot of the fluorescence intensity versus total scattering intensity based on single particle analysis.As can be easily seen, there is no clear separation between fluorescing and not fluorescing particles.The separation line drawn at 0.003 is based on an analysis of the noise level (see also Fig. 6 captions).Moreover no correlation between scattering intensity (which is proportional to the square of the particle size) and fluorescence intensity (proportional to the amount of dye) is evident.However in this particular run shown here, 546 of the 916 particles detected emitted fluorescence, which is equivalent to 60% of the particles and reproducibly seen during our lab tests with a 50%/50% mixture of the fluorescing/non-fluorescing samples.Within the large errors of the mixing ratio resulting from uncertainties of www.atmos-meas-tech.net/3/263/2010/Atmos.Meas.Tech., 3, 263-271, 2010    Fluorescence signal in a.u.

Fig. 6.
Example of signal graphs of two test particles that have passed the detector within one second.On the y-axis the signal intensity and on the x-axis the sample number taken at 200 kHz are shown.The signal was triggered, thus peaks are centered at sample number 100.If more than one particle was detected the signal was added with a gap of 56 samples in advance.The upper panel shows the phase discriminating depolarization channels (P RC red line and P LC black line), the lower the flourescence channel.For all signals displayed the constant background is substracted.Particles are counted as fluorescent if the signal maximum is larger than the photomultiplier background signal plus 5 σ noise .For this measurement σ noise is 0.0006 which corresponds to a detection limit of 0.003.In this example the first particle shows a significant but weak flourescence signal.The second particle shows no or only extremely weak fluorescence and is therefore counted as not fluorescent.The signals in the fluorescence channel are generally broader than the photodiode signals because the circular spot size of the UV source is larger than the elliptical spot of the diode laser.Thus the particles need more time to travel through the detection volume.

Number of Signals
Fluorescence signal in a.u. the concentrations of the two sample solutions, the observed ratio of fluorescent particles to all particles is consistent with the mixing ratio.

Initial tests with ambient aerosol
Ambient aerosol was sampled directly through a 3 m×6 mm I.D. copper tube from the outside of the Geosciences Department building in Frankfurt am Main, Campus Riedberg, Germany, at 10 m above ground level on 26 June 2009, 14:30-16:10.The sample flow rate was 5 LPM.In this example for atmospheric measurements, a total number of 704 particles have been counted, which amounts 7 particles per minute.52 particles (6.3%) have been found to be clearly fluorescent and therefore of biological origin.A nine minute ambient sample taken with the APS directly after the measurements resulted in a total number of 69 particles larger than 3 µm.If one assumes that the particle concentration and Total scattering signal in a.u.
Fluorescence signal in a.u.size distribution did not vary significantly during the time of both measurements, then we may conclude from the total counts/minute that the detector triggered on particles larger than about 3 µm.This observed biological fraction of particles >3 µm is a reasonable result for this initial test.It is consistent with the 75 % quartile range of 1.7-5.5 % based on 5 min samples for fluorescent biological particles >1 µm measured with a UV-APS (TSI Inc.) during a 4 month period (3 August-4 December 2006) at Mainz (Germany) (Huffman et al., 2009).Mainz is located located close to Frankfurt in the metropolitan area named "Rhein-Main region" .
Figure 10 shows an example of a relatively small particle, for which the scattering intensity is only slightly above the triggering threshold, but which shows a strong fluorescence signal.
Figure 11 shows a scatter plot of the signal intensities in the scattering and the fluorescence channels.No correlation between particle size (proportional to the square root of the scattering intensity) and fluorescence intensity is observed.This is not surprising, given the large variance of biological particles in the atmosphere and the different concentrations of auto-fluorescent metabolites among them.
Up to now, no investigations about the influence of ice coverage on the fluorescence properties of the particles have been made.Ice coverage of the particles might have some influence on the detection probabilities, which we can hardly quantify yet, but they should be mentioned in the following.Certainly small biological IN which grow to detectable size can be miscounted as not fluorescing, because the overall amount of fluorophore metabolites is too small to be detected.For the ice cover itself we do not expect a large influence, because the transmission of a 10 µm layer of ice is   In this case a small particle (weak scattering signal) exhibits a strong fluorescence signal.Here it is obvious that particle size is not directly correlated to the fluorescence signal (see also Fig. 8).The small shift of the signal maxima of about 25 samples (=125 µs) between the two channels originates from a small misalignment of the two beams.
larger than 99.999% in the wavelength region from near UV to VIS (Wozniak and Dera, 2007;Warren, 1984).A second point is the loss of UV irradiation on particle by reflection at the ice surface, which can roughly be estimated to be 1-4% (depending on angle and surface structure).The fluorescence emission on the other hand is omni-directional, thus scattering inside the ice shell can be neglected.The fluorescence intensity at low temperatures might be even Fluorescence signal in a.u.
Total scattering signal in a.u.lifetimes and emission quantum yield at low temperatures.The same behavior was found for NADH (Scott et al., 1969).Summing up we expect that the detection probability will not be reduced significantly by the ice shell.

Conclusions
It is known that some aerosol particles which contain biological material activate as IN at relatively high temperatures.
They will thus activate first during cloud icing and will determine the temperature of the onset of cloud icing, and may alter the microphysical structure of the cloud significantly.
The new BIO IN detector allows to classify IN for their biological content and hopefully constitutes an helpful contribution for the understanding of the origin of IN.The performance of this low-cost detector is better than we expected originally but is still limited because of the signal to noise ratio.Thus, the separation between biological and non-biological material is not sharp for particles with low fluorescence activity (e.g.spores).The fluorescence yield is related to the concentrations of NADH/NADPH and Flavoproteins (FAM, FAD, Riboflavin, etc.), which depend also on their oxidation state, varying with the energy metabolism of the living cells.During our initial tests we identified at least 6% of particles larger than 3 µm to be of biological origin.

Outlook
It is planned to enhance the signal to noise ratio by using a a powerful 355 nm solid state Laser, equivalent to the one used by the UV-APS of TSI company (Hairston et al., 1997).It may be possible then to differentiate various types of biolog-ical particles such as spores, pollen, bacteria, cell fragments, by use of a high speed spectrometer to obtain the dispersed fluorescence spectrum, instead of the simple photomultiplier.This will improve the design significantly -but not longer at low cost.
Another possible improvement is the use of photomultipliers in the scattering channels instead of the less sensitive photodiodes.By this we expect to shift the detection limit to particle sizes smaller one µm.
Furthermore we would like to make some comparison experiment with the commercial UV-APS.It is clear that our detector is less sensitive than the UV-APS, especially for the detection of small particles.But we expect a similar detection efficiency for fluorescent bio-particles.
The BIO IN detector as part of the FINCH measurement system is currently in the certification process for use onboard the new HALO aircraft.

Fig. 1 .
Fig. 1.Scheme of auto-fluorescence emission spectra of NADH, Flavins and Lipopigments.The sum of the different emission spectra in the 400-600 nm range will give the signal strength measured by the BIO IN detector.Figure modified from(Palumbo and Pratesi, 2004).

Fig. 2 Fig. 2 .
Fig. 2 Schematic flow diagram of the FINCH counter.The new BIO IN detector is incorporated in the depolarization detector located behind the virtual impactor at the bottom of the development section.Modified from (Bundke et al., 2008) Fig. 2. Schematic flow diagram of the FINCH counter.The new BIO IN detector is incorporated in the depolarization detector located behind the virtual impactor at the bottom of the development section.Modified from(Bundke et al., 2008).

15Fig. 3
Fig. 3 Schematic diagram of the BIO IN detector optics.The numbered optical parts are given in Table1.The aerosol flow intersects the paper plane vertically at the point of intersection of the laser and UV-LED beam (see also Fig.4).

Fig. 3 .Fig. 4 .
Fig. 3. Schematic diagram of the BIO IN detector optics.The numbered optical parts are given inTable 1.The aerosol flow intersects the paper plane vertically at the point of intersection of the laser and UV-LED beam (see also Fig.4).

Fig. 5 Fig. 5 .
Fig. 5 Schematic diagram of the test-aerosol suspension unit (silica micro-particles, see text for details)

Fig. 6 :
Fig. 6: Example of signal graphs of two test particles that have passed the detector within one second.On the y-axis the signal intensity and on the x-axis the sample number taken at 200 kHz are shown.The signal was triggered, thus peaks are centered at sample number 100.If more than one particle was detected the signal was added with a gap of 56 samples in advance.The upper panel shows the phase discriminating depolarization channels ( RC P red line and LC P black line), the lower the flourescence channel.For all signals displayed the constant background is substracted.Particles are counted as fluorescent if the signal

18Fig. 6 :
Fig. 6: Example of signal graphs of two test particles that have passed the detector within one second.On the y-axis the signal intensity and on the x-axis the sample number taken at 200 kHz are shown.The signal was triggered, thus peaks are centered at sample number 100.If more than one particle was detected the signal was added with a gap of 56 samples in advance.The upper panel shows the phase discriminating depolarization channels ( RC P red line and LC P black line), the lower the flourescence channel.For all signals displayed the constant background is substracted.Particles are counted as fluorescent if the signal

Fig. 7 Fig. 7 .Fig. 8
Fig. 7 Intensity histogram of the frequency of occurrence of the signal strength of the circular depolarization detector using test aerosol.

Fig. 8 .
Fig. 8. Intensity histogram of the frequency of occurrence of the signal strength of the fluorescence detector using test aerosol.

Fig. 9
Fig. 9 Scatter plot of the fluorescence signal versus the total scattering signal intensity (sum of both polarization signals) on the basis of single particle analysis of the 10µm test particles.The line at 0,003 indicates the discrimination between fluorescent and not or only very low fluorescent particles.

Fig. 9 .
Fig. 9. Scatter plot of the fluorescence signal versus the total scattering signal intensity (sum of both polarization signals) on the basis of single particle analysis of the 10 µm test particles.The line at 0.003 indicates the discrimination between fluorescent and not or only very low fluorescent particles.

Fig. 10
Fig. 10 Example of signal graphs where one small ambient particle have passed th On the y-axis the signal in arbitrary units and on the x-axis the sample numbe 200 kHz are shown.The signal was triggered, thus peaks are centered at sample nu The upper panel shows the phase discriminating depolarization chanel, the lower pa the fluorescence channel.In this case a small particle (weak scattering signal)

Fig. 10
Fig. 10 Example of signal graphs where one small ambient particle have passed the de On the y-axis the signal in arbitrary units and on the x-axis the sample number ta 200 kHz are shown.The signal was triggered, thus peaks are centered at sample numb The upper panel shows the phase discriminating depolarization chanel, the lower panel the fluorescence channel.In this case a small particle (weak scattering signal) exh

Fig. 10 .
Fig.10.Example of signal graphs where one small ambient particle have passed the detector.On the y-axis the signal in arbitrary units and on the x-axis the sample number taken at 200 kHz are shown.The signal was triggered, thus peaks are centered at sample number 200.The upper panel shows the phase discriminating depolarization chanel, the lower panel shows the fluorescence channel.In this case a small particle (weak scattering signal) exhibits a strong fluorescence signal.Here it is obvious that particle size is not directly correlated to the fluorescence signal (see also Fig.8).The small shift of the signal maxima of about 25 samples (=125 µs) between the two channels originates from a small misalignment of the two beams.

Fig. 11
Fig. 11 Scatter plot of the fluorescence signal versus the total scattering signal intensity (sum of both polarization signals) on the basis of single particle analysis.The line at 0,004 indicates the discrimination between fluorescent and non or only very low fluorescent particles.There is no significant correlation between particle size (scattering intensity) and fluorescence signal strength.

Fig. 11 .
Fig. 11.Scatter plot of the fluorescence signal versus the total scattering signal intensity (sum of both polarization signals) on the basis of single particle analysis.The line at 0.004 indicates the discrimination between fluorescent and non or only very low fluorescent particles.There is no significant correlation between particle size (scattering intensity) and fluorescence signal strength.

Table 1 .
Physical setup of the BIO-IN detector(see also Fig.3).