An improved Dual-wavelength-excitation Particle Fluorescence Spectrometer (DPFS) has been reported. It measures two fluorescence spectra excited sequentially by lasers at 263 nm and 351 nm, from single atmospheric aerosol particles in the 1-10 μm diameter size range. Here we investigate the different levels of discrimination capability obtained when different numbers of excitation and fluorescence-emission wavelengths are used for analysis. We a) use the DPFS to measure fluorescence spectra of Bacillus subtilis and other aerosol particles, and a 25-hour sample of atmospheric aerosol at an urban site in Maryland, USA; b) analyze the data using six different algorithms that employ different levels of detail of the measured data; and c) show that when more of the data measured by the DPFS is used, the ability to discriminate among particle types is significantly increased.
© 2010 OSA
There is an increasing interest in characterizing biological and other organic carbon containing aerosols. Atmospheric aerosol contains complex internal and external mixtures of organic and inorganic compounds. A large fraction of the organic component of atmospheric aerosol consists of Primary Biological Aerosol (PBA). PBA indicates airborne particles that include bacteria, bacterial spores, fungal spores, pollens, viruses, algae, and parts of plants, fungi, or animals that may have been directly injected into the atmosphere by, e.g., wind ablation. PBA occur over a large size range, varying in diameter from a few 10’s of nm (e.g., small viruses) to 100 μm (aggregates, or some pollens). PBA may be generated from what may sometimes be thought of as biological waste. For example, rotting wood may contain fungal hypae (parts of the body of the fungus). Frass, the excrement produced by insects as they eat plants, etc., may consist largely of bacteria and fungi. PBA aerosol particles may be mixtures of biological and nonbiological material. PBA is important in: a) transmission of diseases of humans, other animals and plants; b) generating allergies; and c) acting as cloud condensation nuclei. PBA tend to absorb more light than most inorganic aerosols, especially at shorter wavelengths, and may therefore affect the heating and cooling of the atmosphere. PBA has been reported to comprise 15 to 80 percent of the atmospheric aerosol [1–3], depending on geographic site and season. Elbert et al.  have estimated that the global average fungal-spore mass concentration in the lower atmosphere is 1 μg/m3, and that 50 Tg/yr of fungal spores are emitted each year. There is an enormous range of molecules and combinations of molecules in PBA and other organic-carbon aerosol [4–11].
The intrinsic Laser-Induced-Fluorescence (LIF) of aerosol particles has been used for detection and classification of PBA and other organic-chemical aerosol particles. The first attempts to exploit the LIF of single airborne particles for their classification began in the mid-1990s. Pioneering efforts concentrated on measurement of the undispersed fluorescence within one to three bands [12–17]. Various sensors for airborne particles that measure fluorescence in one to three bands have been commercially available for years (see review by DeFreez ). Almost concurrent with the early developments, more capable LIF detectors that measure the dispersed fluorescence spectrum in many channels [19–30] were developed in an attempt to increase the potential for discrimination and provide information on particle composition. Also, LIF-based sensors have been developed with dual-wavelength excitation with the capability to detect emission in two or four bands along with elastic scattering [25,31–35]. One LIF sensor measures the time-dependent fluorescence in four bands . Preliminary reports of a sensor capable of dual-wavelength excitation and measurement of fluorescence spectra for each excitation wavelength have appeared [36,37].
The number of different types of fluorescent molecules occurring in PBA and other organic-carbon aerosol is enormous. The great majority of the fluorophors of the fluorescent molecules in PBA include one or more aromatic rings, which are commonly heterocyclic and substituted. Some fluorophores, e.g., tryptophan, tyrosine, nicotinamide adenine dinucleotides, and flavins, occur in all living cells, but not necessarily in all PBA. Many common fluorophors in PBA do not occur in all cells, e.g.: chlorophylls in plants; lignins, lignans, sinapyl alcohols and many others in woody plants; and compounds such as ferullic acid (with a fluorescence spectrum similar to NADH) in cellulose. Other fluorophors from secondary metabolites of plants and fungi  can occur in PBA and may occur in a smaller number of species, e.g., fluorescein is made by Pseudomonas aeruginosa and P. fluorescens; coumarin and its derivatives such as umbelliferone are found in several plant families. Although studies of the genetic material from microorganisms in atmospheric aerosol are in their infancy [4,44], such studies are expected to eventually illustrate in more detail the immense diversity of microorganisms transported in air. A diversity in genomic material is likely related to a diversity in secondary metabolites. Many non-biological compounds found in the atmosphere, including a very significant fraction of polycyclic aromatic hydrocarbons (PAHs) and their oxidation products, are highly fluorescent . Although PAHs can occur primarily in combustion-generated particles smaller than 1 μm in diameter, these small particles can agglomerate to particles larger than 1 μm diameter , and may agglomerate with nonfluorescent particles . Humic materials and humic-like substances (HULIS) are other fluorescent materials that occur in atmospheric aerosol [42,43]. HULIS can be formed from biomass burning, anthropogenic sources, and marine sources, and may arise as secondary organic aerosols from lignin pyrolysis products, etc.
Given the extra complexity and cost of multiple excitation and/or multiple emission LIF systems, an important question is: What additional discrimination capability is added when more excitation and/or emission wavelengths are used for analysis of aerosols? The question is open ended because there are: a) so many different discrimination problems (different target aerosols, different ambient background aerosols containing a highly diverse array of microoranisms, different excitation and emission wavelengths, etc.), and b) so many different fluorescent molecules in atmospheric aerosols.
In this paper we investigate the increase in discrimination capability obtained when more excitation and/or emission wavelengths are employed. The data are taken using a Dual-wavelength-excitation single-Particle Fluorescence Spectrometer or DPFS, which can measure the fluorescence spectra and elastic scattering excited at both 263 nm and 351 nm laser wavelengths for single atmospheric aerosol particles (in the 1-10μm size range) as they flow through a sampling cell. We use the DPFS to measure fluorescence spectra of: a) several types of test particles, including a preparation of Bacillus subtilis (a spore forming bacterium); and b) of atmospheric aerosol measured during a 25-hour period at an urban site in Maryland, USA. Then we analyze the data using six different algorithms that employ successively more features of the measured data. Each algorithm determines the fraction of atmospheric particles that are similar to the preparation of B. subtilis spores. We find that when both spectra and elastic scattering signals measured by the DPFS are used, only 0.06% of the atmospheric particles measured have fluorescence spectra and scattering signals consistent with (i.e., within two standard deviations at each wavelength) the specific preparation of B. subtilis studied. We employ B. subtilis, not because anyone is concerned specifically about the occurrence of B. subtilis in the atmosphere, but because it contains some fluorophors known to occur in a significant fraction of PBA, it is frequently used as simulant for bacterial agents, and it is relatively innocuous.
We also describe improvements in the DPFS system which provide for higher-quality measurements of single-particle fluorescence spectra and elastic scattering. This work demonstrates the increased capability for discrimination provided by the DPFS and similar fluorescence-based approaches for measurement and classification of atmospheric aerosol, particularly for differentiation of certain biological aerosol particles from the natural background.
2.1 Experimental: Improved DPFS aerosol sampling system
The DPFS system for the rapid measurement of dual-wavelength excited fluorescence spectra and elastic scattering from single atmospheric aerosol particles flowing through an optical cell is illustrated in Fig. 1 . The system employs two pulsed UV-lasers (wavelengths 263-nm and 351-nm), fired sequentially, to excite the fluorescence, and a single 32-anode PMT for the detection of fluorescence spectra. The basic scheme of this system has been briefly described previously [36,37]. Briefly, the key improvements over those systems are: i) the use of infrared (IR) diode trigger lasers instead of visible trigger lasers so that the visible fluorescence can be measured over a larger wavelength range; ii) positioning the lasers so that the two spectra are recorded for optimal alignment (the particle is at approximately the same distance from the focal point of the collection optics for each spectrum); and iii) an improved combination of optical filters.
In order to reliably sample micron-sized atmospheric particles at reasonable rates, a virtual impactor concentrator was used upstream from the inlet of the DPFS. The concentrator (MSP model 4220) samples air at a rate of 330 L/min and provides concentrated aerosol particles (over the 2-10μm size range) in the minor outlet flow at a rate of about 1 L/min. This minority outlet flow is fed to the inlet nozzle assembly of the DPFS (Fig. 1B) for aerodynamic focusing which results in further particle concentration. The nozzle assembly consists of two concentric nozzles: an inner nozzle that forms the aerosol jet (having diameter around 400 μm) with particles moving at about 10 m/sec speed), and an outer nozzle for a clean air sheath flow. This nozzle focuses the aerosol into a jet that remains collimated, laminar, and cylindrical for a distance of about 1cm from the nozzle . The nozzle assembly is mounted in a small optical chamber (see Fig. 1A - a cubical airtight cell, 5 cm on each side), where single particles are interrogated with lasers. The chamber is aspirated through an outlet tube concentrically aligned with the inlet nozzle assembly. A piston pump (KNF Neuberger, UN 86) draws air through the nozzles and chamber. A uniform flow is achieved by inserting a critical orifice between the outlet tube and the piston pump, which reduces pressure fluctuations caused by the pump.
Particles flowing near the center of the aerosol jet and 5 mm below the nozzle tip are detected (from their scattering) by two continuous wave (CW), crossed IR diode-laser beams (785- and 830-nm, 10 mW), which are positioned perpendicular to the aerosol jet. The intersection of these diode laser beams forms an approximately 150-μm × 150-μm region that defines the “trigger volume” within the aerosol jet (see detail in Fig. 1A and 1C). Once particles within the trigger volume are detected by the PMTs (one PMT is equipped with an interference filter that passes 785-nm light, and the other PMT equipped to pass 830-nm light), and the signals exceed a preset threshold, a logic AND gate generates a 500-ns TTL pulse to initiate the measurement protocol.
The detailed time sequence of events is shown in Fig. 2 . First, the two diode lasers are turned off coincident with the AND gate TTL pulse (turn-off time is about 1 μs) so that scattering from them will not affect subsequent measurements. Second, 5 μs after the TTL pulse is sent, the first probe laser, a 263 nm laser pulse (10-ns, 0.030-mJ, 1-mm diameter, forth harmonic of a Q-switched Nd:YLF laser, Photonics Industries DC- 150-263) is fired to excite fluorescence in the targeted particle (after 5 μs the particle will have traveled about 50 μm below the trigger volume). The emitted fluorescence is collected by a large-aperture (NA = 0.4) Schwarzschild reflective objective (Newport 50105) and focused onto a spot centered about 1 mm below the middle of the input slit of a spectrograph (Jobin Yvon, CP-140). The dispersed spectrum is recorded by a 32-anode PMT (Hamamatsu H7260). The overall spectral resolution is about 16.5 nm per anode of the multi-anode PMT. Any residual scattering or fluorescence induced by the two IR lasers and electronic noise generated by the turn-off process are mostly eliminated by taking data 5-μs later. A long-pass liquid filter (dimethyl-formamide diluted with water in a 1-cm thick cell) is placed in front of the spectrograph slit to block nearly all the elastic scattering from the 263-nm laser; this filter efficiently transmits the fluorescence with wavelengths longer than 280 nm. The ratio of DMF to water is adjusted so that the magnitude of the elastic scattering leaking through the filter can be used to estimate particle size, but not so large as to saturate the detector for particles within the 1-10 μm size range. Third, the fluorescence spectrum and elastic scattering intensity for each particle measured by the 32-anode PMT are captured and analyzed by a custom-designed readout and processing electronics interface (PhotoniQ OEM, Vtech). This board is triggered 50 ns earlier than the UV laser, and first reads the background charge from the PMT at each anode, and then reads the signal charge that accumulates in 200 ns during the fluorescence emission window. The absolute fluorescence intensity at each anode is obtained by subtracting the background charge from the signal charge. The entire reading, subtracting, analog-digital conversion (ADC), spectral analysis, and data saving process for the spectrum takes 11.6 μs. Fourth, a second probe laser, a 351 nm laser pulse (10-ns, 0.025-mJ, 1-mm diameter, third harmonic of a Q-switched Nd:YLF laser, Photonics Industries DC- 150-351) is fired to excite fluorescence in the same particle 12 μs after firing the 263 nm laser pulse (17 μs after the AND gate TTL pulse, during which time the particle has moved about 170 μm below the trigger volume). The fluorescence excited by the 351-nm laser pulse is focused onto a spot centered about 1 mm above the middle of the input slit. A long-pass filter with cutoff at 380 nm covers the upper half of the slit to block nearly all the elastic scattering from the 351-nm laser, but without attenuating the fluorescence from 280 nm to 380 nm excited by 263 nm laser that was previously focused onto the lower part of the slit. The center of the two interrogation zones for the two UV-laser probe pulses and the optical axis of the collecting objective are centered at the middle of the slit. The corresponding 351-nm laser induced fluorescence spectrum and elastic scattering is measured by the same 32-anode PMT and are again captured by the PhotoniQ board.
The two IR diode lasers for triggering are turned on again, once the data reading process performed by the PhotoniQ board is completed. The system does not accept any trigger from the AND gate until 5 μs after the recording window for the second probe laser; this ensures the board has enough time to finish the acquisition and data transfer process for the 351-nm excited fluorescence spectrum and is ready to take the next 263-nm excited fluorescence spectrum again for the next trigger event. This dead time is also sufficient to avoid false triggering induced by scattering, fluorescence, or electronic noise that might arise from the turn-on of the two IR lasers. The entire process for obtaining two fluorescence spectra excited by two sequentially-fired laser pulses (at 263 nm and 351 nm) illuminating a single aerosol particle takes about 25 μs. Therefore, the maximum frame rate allowed by the electronics is on the order of 40,000 per second. However, our requirement that there be only a small probability of sampling multiple particles at once limits the actual useable frame rate to some number much smaller than that - around 4,000/s. In the work reported here the peak sample rates are less than 500/s, well below this rate.
Because the depth-of-field of the collection optics (a reflective objective with numerical aperture 0.4) is less than 10 μm, we try to limit the trigger region and the two interrogation regions to be within 150 x 150 μm horizontally (perpendicular to the direction of flow), which is defined by the crossed diode laser beams. The particles move about 120 μm vertically between the times the two probe lasers fire. This distance is set by the shortest time separation (12 μs) in which the electronics is able to record two spectra and by the speed of the particles (10 m/s). To keep the two interrogation regions as close as possible to the focal point of the Schwarzschild reflective objective, the first interrogation region (with illumination by the 263-nm laser) is positioned about 60 μm above the focal point of the lens, and the second interrogation region (with illumination by the 351-nm laser) is positioned about 60 μm below the focal point of the lens. The small depth-of-field of the collection optics is a main reason for the need to use the double-nozzle assembly to form a laminar, cylindrical, aerosol jet that remains somewhat uniform over the 170-μm distance from the trigger region to the second interrogation region. Uniform trajectories are needed so that targeted particles can be more reliably probed in the interrogation regions of the two sequentially-pulsed UV probe lasers. Even with the aerodynamically focused particle stream, optical alignment is challenging, and variability of particle position during interrogation by the probe laser still causes large signal differences.
The DPFS does suffer from some background noise resulting from the relatively small optical cell, and stray light from laser beams passing through the cell windows. To largely eliminate these unwanted background signals, before taking data on test or atmospheric aerosol particles, 1000 “background” spectra were routinely recorded in a “free-running” mode with the DPFS triggered by a pulse delay generator (SRS3500) at 100 Hz, with no particles present. Spectra for both probe lasers (263-nm and 351-nm) were taken under the same experimental conditions as data in the normal “particle-triggering” mode. The DPFS spectra recorded in the normal mode were then corrected by subtracting the average of the 1000 “background” spectra. Corrections were also made for including the scattering from particles that have negligible fluorescence, but these are better described below, in Section 3.5.
2.2 Test particles and test site for atmospheric measurements
Various solutions/slurries of non-biological and biological samples were aerosolized: polystyrene latex spheres (Duke Scientific); kaolin, a common clay mineral occurring in atmospheric aerosol in arid regions and having very low intrinsic fluorescence (Particle Information Services); tryptophan (Fluka); riboflavin and albumin (Sigma Chem.); Bacillus subtilis spores that were not extensively washed to remove culture media (prepared by Dugway Proving Ground); and MS2, a bacterial phage in a preparation that was not extensively washed to remove culture media (prepared by the Armed Forces Institute of Pathology).
An inkjet aerosol generator (IJAG ) is used to obtain test aerosols. A solution or suspension of the material being aerosolized is placed into the IJAG cartridge. The IJAG generates fairly uniform liquid droplets of approximately 50-μm diameter; smaller satellite droplets are mostly removed by a secondary airflow. The droplets pass through a heated drying column forming the residual dried aerosol particles. For aerosolizing the uniformly-sized PSL calibration spheres, the suspension placed into the IJAG is diluted sufficiently so that the probability of two PSL spheres occurring in the same droplet is very low. For aerosolizing other test materials, the test material is dissolved and/or suspended in water and placed into the IJAG, and the dried particle size depends on the concentration of the material dissolved and/or suspended.
The test site for measuring atmospheric aerosol particles is in Adelphi, MD, USA (39° N latitude, elevation 75 m). This site, located in the Baltimore-Washington metroplex, was also used for a previous study . Ambient air is drawn through a 15 cm-diameter, 10 m-long metal duct protruding through the 25 m high roof of our laboratory (see Fig. 1B). This duct is aspirated by a large squirrel-cage fan attached to a chemical-hood. Part of the air passing through this duct is fed to the virtual impactor concentrator inlet with the minority flow fed to the DPFS.
2.3 Preliminary analysis – check for validity of spectra
After data collection is complete, and before the analyses described later are done, a computer program is used to confirm that each channel of each spectrum is within the correct range and that each pair of spectra are in the correct order.
Confirm that the charge is within range on each channel. Because of nonlinearities which distort the spectral profile when the intensity is near saturation, spectra within any channel (anode of the 32-anode PMT) having a charge in pico-Coulombs greater than 400 (on a 0-600 pC scale) were discarded. The charge is accumulated during the 200 ns integration period of the PhotoniQ board.
Confirm that spectra are from the same particle. The order of firing of the two lasers for any particle is 263-nm then 351 nm. Occasionally, typically or always because of range errors (which occur when a channel exceeds the 400 pC limit) the spectra become out of order in the data file. To verify whether the two spectra are from the same particle, the program looks for scattering from the first (excitation 263-nm) wavelength in a spectrum, and then the scattering from the second (excitation 351-nm) wavelength in the following spectrum. If the two elastic scattering peaks are not in the correct order the spectra are discarded.
3. Measurement and analysis of test and atmospheric particles
3.1 Polystyrene latex particles of known sizes
Because the DPFS measures light scattered by particles and not their size directly, we performed a crude calibration to convert the light scattering signal to particle size. For this calibration we used standard polystyrene latex (PSL) particles aerosolized by an inkjet aerosol generator (IJAG ). For size calibration, in order to produce particles that include only one primary PSL spherical particle, the suspension was made sufficiently dilute that only about one in five IJAG droplets contains a single PSL particle on average. Figure 3 shows the elastic scattering responses (from the 263-nm laser pulse) for polystyrene latex particles with nominal sizes 1 μm, 2 μm, 3 μm, 5.6 μm, and 8μm (obtained from Duke Scientific). The particles are are specified by the manufactured to have a standard deviation in size of less than 5%.
The square root of the scattering intensity in Fig. 3 is plotted versus particle size because particles in this size range (relative to the wavelenghth of the incident light) are known to have total scattering cross sections that are roughly proportional to size. The best least squares fit for the data was found to beFig. 3.
Figure 4(a) shows histograms of light scattering size (determined from the relation above) for polystyrene latex particles with nominal sizes 1 μm, 2 μm, 3 μm, 5.6 μm, and 8 μm. For comparison, the histograms of aerodynamic paricle size histograms measured by the TSI-APS 3200 particle sizer are presented in Fig. 4(b). The light scattering particle size histograms show reasonable agreement with the aerodynamic size histograms, but exhibit a larger uncertainty, especially in the smaller size range. Better precision in the light scattering measurement might result if a near forward scattering measurements were used, rather than the scattering in the 60-120 degree angular range used here.
3.2 Tryptophan test particles: elastic scattering and fluorescence
The response of the DPFS in terms of elastic scattering and total fluorescence is illustrated in Fig. 5 . Homogenous tryptophan particles were generated with the IJAG. Given the concentration of tryptophan in water, and the typical breakup size for droplets in the IJAG, the dominant particle size should be about 4-μm diameter. Particles generated with the IJAG typically have about 30% standard deviation in size, but there also tend to be some particles much smaller in size.
Elastic scattering at both excitation wavelengths and spectra excited by both lasers were measured as described in section 2. The 263-nm-excited total fluorescence and the 351-nm-excited total fluorescence were obtained by summing the channels having fluorescence. The upper plot in Fig. 5 shows the scatter plot of the elastic scattering at 351-nm vs. the elastic at 263-nm. The lower plot shows the fluorescence excited by 263-nm and by 351-nm vs. the elastic at 263 nm. Each dot represents one particle. The light-scattering particle size was estimated using Eq. (1).
Part of the scatter in the data in Fig. 5 is attributable to dispersion in size for particles generated by the IJAG, and part arises from the variability in the response of the DPFS. The variability in the elastic scattering at one wavelength can best be ascertained from Fig. 4 which shows the scattering results for quite uniformly sized PSL spheres. The upper curve in Fig. 5 illustrates that even at a single scattering intensity at 263 nm, the scattering at 351 nm can vary by a factor of 5. This relatively large variation arises because the elastic scattering has such a strong angular dependence, and it is being collected at varying positions relative to the focal point of the reflecting lens. A second reason for the variation is the differences in the intensity of the illumination beam.
The fluorescence excited by either the 263-nm or the 351-nm beam is bunched more tightly (than the 351-nm elastic scattering) at any given light-scattering-size primarily because the angular dependence of the fluorescence is much smaller than the angular dependence of elastic scattering. Consequently, the effect of particle position (with respect to the collection optics) is far smaller in the case of fluorescence.
Because tryptophan strongly absorbs 263-nm light, the actual particle size determined from the 263-nm scattering is expected to be larger than the “light scattering” size. This effect of absorption on scattering size should be less significant with 351-nm light, which is absorbed less by tryptophan.
3.3. Biological and non-biological test aerosols
Spectra of aerosols made from some biological and nonbiological materials are displayed in Fig. 6 . Each spectrum is an average of 100 spectra from nominal 5-μm-diameter particles. All spectra are normalized to have the same peak height as that of B. subtilis. The sharp peaks at 263 nm and 351 nm are the elastic scattering from the lasers, and the peak 527 nm is the second order of the elastic scattering peak at 263 nm. The spectra for all three simulants (B. subtilis, albumin, and the phage MS2) have a strong fluorescence peak around 340 nm, the region of peak emission from the amino acid tryptophan. The spectral shoulders from 400 to 600 nm show some differences which may be attributable to fluorescent compounds of the growth material. Reduced nicotinamide compounds such as nicotinamide adenine dinucleotide (NADH) may contribute in the 400 to 600 nm range for B. subtilis. However, albumin is a protein (obviously not purified extremely well or it would lack the shoulder at 410 nm), and MS2 is a bacterial phage, which should have no NADH if it were washed well.
We and others have found that the fluorescence spectra: a) from one species of bacterium prepared under different growth conditions can exhibit larger differences between each other than those seen in the examples shown in Fig. 6; and b) from different species of bacteria, prepared under the same conditions, can have very similar spectra . This dependence of the LIF-spectra of bacteria on growth conditions and sample preparation, and the similarity of many types of bacteria if separated from any additional materials, makes it appear impossible to discriminate between the many, many different species of bacteria that can be in bioaerosols based only on their LIF spectra. Differences in LIF spectra for atmospheric bacteria may be more related to how the bacteria grew and were aerosolized, and on any other materials they may be agglomerated with. On the other hand some other types of microorganisms, such as fungal spores and plant pollens may have a much higher fraction of species with distinctive spectra. Also, it may be that certain species of bacteria when growing in their natural environment, do have distinctive signatures, either because of their own secondary metabolites, or because of fluorophors in their microenvironment.
Figure 6 illustrates that for the three particle types (B. subtilis, albumin, and the phage MS2, each prepared in such a way that fluorescent impurities remain) improved discrimination can be achieved by using the fluorescence spectra excited by the 351-nm laser (see right side of Fig. 6) along with the 263-nm-excited spectra. The spectral differences are clearly visible to the naked eye. Also, because different samples can have markedly different ratios of fluorescence-to-elastic-scattering, it is also possible to use the relative strengths of the fluorescence signals, and the ratios of fluorescence-to-scattering, to discriminate more effectively between these particles.
3.4 Thresholds for elastic scattering and fluorescence
Using the measurement described above, we decided that for a particle to be considered fluorescent it should satisfy two conditions.
First, the light detected in the fluorescence wavelength region must be at least 3 times the average of background noise (i.e., the intensities recorded when the laser fires when no particle is present). For the rest of this paper the threshold for total fluorescence (summed over all channels used) whether the 263-nm excited or the 351-nm excited, is 4, which also happens to be 1/100th of the maximum allowed charge in pC in any channel.
Second, the ratio of total fluorescence to elastic scattering must be greater than a certain fraction (0.34 for 263-nm fluorescence / 263-nm elastic, and 0.26 for 351-nm fluorescence / 351-nm elastic). This second condition is required because the light scattered by large nonfluorescent particles (e.g., a 10-μm kaolin particle) is large, and some of this elastic scattering can generate a small fluorescence signal on the 32-anode PMT. The elastic scattering may generate some fluorescence from windows or filters of the cell, which may find its way into the spectrometer. Alternatively, a small fraction of the elastic scattering may scatter within the cell and may somehow leak to the fluorescence channels of the multi-anode PMT. These two criteria are used for the analyses for the rest of this paper.
3.5 Atmospheric aerosol
The DPFS has been used to measure the concentration, fluorescence spectra and elastic scattering (from which particle size is estimated) of atmospheric aerosol particles, quasi-continuously during September 2009. Typically there are around ten to a few hundred particles detected per second within the 1-10 μm size range. About one to two million spectra for each excitation wavelength are commonly recorded each day. We first determined, from the DPFS measurements of atmospheric aerosol, the elastic scattering and total (integrated over wavelength) fluorescence excited by the 263-nm and 351-nm probe lasers. A typical data set for 1000 atmospheric aerosol particles (sampled between 5:00-5:04 PM on Sept. 17, 2009) is displayed in Fig. 7 . The fluorescence intensity is plotted on a log-scale to better illustrate the large dynamic range of fluorescence intensity. In this data set, 285, or 28.5% of the particles have fluorescence above the noise floor and satisfy the two conditions stated in Section 2.3.
The measurements indicate that atmospheric aerosol particles have ratios of fluorescence-to-elastic-scattering that vary by more than three orders of magnitude over the entire 1-10 μm size range. From Figs. 4 and 5 we might estimate that a factor of five (out of the factor of 1000) may be attributable to variations in elastic scattering because of position relative to the collection optics. So, even without the spread due to the DPFS, the range of ratios appears to be more than 200. This finding is consistent with the notion that atmospheric aerosol particles are generally complex mixtures that: 1) can contain mostly inorganic carbon with little or no fluorescent organic carbon (inorganic carbon has very weak fluorescence and consequently a majority of particles have fluorescence signals that fall below the DPFS noise floor), 2) can contain both organic and non-organic carbon in a wide range of mass mixing ratios, and 3) can contain significant amounts of highly fluorescent organic carbon compounds.
The complexity of atmospheric aerosol motivates us to pose the question: To what degree can particular aerosols of interest be discriminated from the background atmospheric aerosol using DPFS or some similar fluorescence-based sensors? In the next sections we attempt to partially answer this question. We are not suggesting that any of the example particle materials (B. subtilis, kaolin, or pure albumin, riboflavin, or tryptophan) used in this paper are important materials of particular interest, although tryptophan and flavins are some of the primary fluorophors in PBA. The question we are attempting to address regarding the value of different amounts of information regarding elastic scattering and LIF spectra is more general, and we hope is of some use to researchers interested in various specific types of PBA and other organic carbon containing aerosols, which may have LIF spectra that are more or less distinctive. For the comparison of the algorithms below, we employ B. subtilis, not because anyone cares specifically about B. subtilis in the atmosphere, but because it is relatively innocuous, has a spectrum when well washed that is similar to other spectra of various other well- washed bacteria, and contains some fluorophors that are known to occur in a significant fraction of PBA.
Some typical DPFS results showing atmospheric fluorescence and elastic scattering, along with some known biological materials and kaolin are displayed in Fig. 8 . Shown is a scatter plot of the ratio of total fluorescence to visible fluorescence excited by 263 nm versus total fluorescence excited by 351 nm and scattering particle size. Data points representing single particles are for B. subtilis, albumin, riboflavin, kaolin, and a sample of atmospheric aerosol particles. We see that the data points for biological agent stimulant particles (here B. subtilis and albumin) have only a small overlap with this sample of atmospheric particles; they have essentially no overlap with kaolin and riboflavin.
4. Analysis methods: Algorithms investigated for discrimination
We constructed six detection algorithms which use varying levels of detailed data measured by the DPFS. The two that use the fewest number of measured intensities (minimal capability of the DPFS) are based only the elastic scattering at one wavelength (263-nm or 351-nm) and the total (integrated) fluorescence excited by laser at that wavelength. The one that uses the most measured intensities (more near the full capability of the DPFS) uses both elastic scattering signals and both fluorescence spectra. Some key aspects of the algorithms are summarized in Table 1 .
Algorithm 1 discriminates biological particles based only on their 263-nm elastic scattering (E263) and total fluorescence (F263-Tot) excited by light at 263-nm only. That is, algorithm 1 requires that particles have: a) total 263-nm-excited fluorescence greater than 4; and b) ratios of UV-fluorescence-to-elastic-scattering greater than 0.34.
Algorithm 2 discriminates biological particles based only on their 351-nm elastic scattering (E351) and total fluorescence (F351-Tot) excited by light at 351-nm only. Thus, algorithm 2 requires that particles have: a) total 351-nm-excited fluorescence greater than 4, and b) ratios of UV-fluorescence-to-elastic-scattering greater than 0.26.
Algorithm 3 discriminates biological particles similar to B. subtilis based on their elastic scattering (263 nm) and two-band fluorescence (ultraviolet and visible) excited at 263 nm as illustrated in Fig. 9 . In this algorithm the UV (263 nm) excited fluorescence is divided into ultraviolet (290-400 nm) and visible (400-600 nm) bands. For simplicity we list this algorithm in terms of its ability to detect B. subtilis-like particles. To facilitate the description of this algorithm consider the 263-nm-excited normalized fluorescence spectra of B. subtilis shown in Fig. 9(a). The particles were selected in an attempt to create a distribution with sizes fairly equally distributed in the 1-10 μm size range as shown in the inset of Fig. 9(a). Plotted is the fluorescence intensity at each anode of the 32-anode PMT from B. subtilis aerosol particles that are normalized by the elastic scattering. We can sum different channels (different regions of the spectrum) to obtain three parameters for each spectrum: elastic scattering, total UV fluorescence (integrated over 290 nm – 400 nm), and total visible fluorescence (integrated over 400-nm to 505-nm and 540 nm to 600 nm). By summing 1000 such B. subtilis spectra we can compute elastic scattering, UV fluorescence, and visible fluorescence parameters, ratios of these parameters, their averages, and standard deviations about the average. The resulting particle scattering parameters for 1-10 μm B. subtilis particles are: (a) the average ratio of UV fluorescence to elastic scattering (F263-Band1 / E263 = 1.17 ± 0.39 (SD) and the average ratio of visible fluorescence to elastic scattering F263-Band2 / E263 = 0.31 ± 0.15 (SD). For this algorithm, a spectrum is considered B. subtilis-like if the two ratios are within 2 standard deviations of the average.
Algorithm 4: Discrimination of biological particles having elastic scattering and two-band fluorescence (ultraviolet and visible) excited by 263 nm and 351 nm that are B. subtilis-like. This algorithm is somewhat similar to sensors developed by Sivaprakasam et al, 2004 and Kaye et al, 2005. In this case we use both elastic scattering channels and sum the fluorescence into blue and visible bands. As with algorithm 3 above we tailor the algorithm for detection of B. subtilis-like particles and use B. subtilis data to determine parameters for the algorithm. Part of the algorithm (for 263-nm excitation) is the same as algorithm 3 above. To define the remainder of the algorithm, fluorescence for 351nm excitation is integrated over the visible (400-600 nm) band. The B. subtilis data have the ratio of visible fluorescence to elastic scattering F351 / E351 = 1.03 (average) ± 0.62 (SD). This algorithm requires that this ratio (for 351-nm excitation) must be within 2 standard deviations of the average.
Algorithm 5 discriminates based on 263-nm elastic scattering and the 263-nm-excited fluorescence spectrum to find particles similar to B. subtilis. The 263-nm-excited fluorescence spectra of B. subtilis particles in Fig. 9(a) shows the fluorescence intensity at each anode of the 32-anode PMT from B. subtilis aerosol particles that are normalized by the elastic scattering. We use 1000 such B. subtilis spectra to calculate an average B. subtilis spectrum and standard deviation about the average for each anode of the PMT (Table 2 ). For this algorithm a particle’s normalized fluorescence spectrum must have signals in each anode of the PMT within two standard deviations of the normalized average B. subtilis spectrum.
Algorithm 6 discriminates based on both 263-nm and 351-nm elastic scattering and both 263-nm- and 351-nm-excited fluorescence spectra to find particles similar to B. subtilis. Discrimination of biological particles having fluorescence spectra (normalized by elastic scattering) excited by both 263 nm and 351 nm that are B. subtilis-like. We use the fluorescence spectra and relative intensity of fluorescence excited by both 263 nm and 351 nm. Again we consider 1000 normalized B. subtilis spectra and determine average B. subtilis fluorescence spectra (normalized by the elastic scattering), and standard deviations about the average, for each anode of the PMT (Table 2). For this algorithm a particle’s fluorescence spectra (normalized by the elastic scattering intensity for each) at both excitation wavelengths must have signals in each anode of the PMT within two standard deviations of the corresponding normalized average B. subtilis spectrum.
There are many other algorithms for classifying bioaerosol particles besides the six described here. Some, for example, may employ principal component analysis (PCA). We chose to analyze the data here by simply comparing the ratios of fluorescence to scattering intensities for certain integrated bands or spectral channels, partly because with these schemes the electronics in the DPFS are fast enough to analyze each detected spectrum in 12 μs. We are especially interested in these very fast algorithms partly because we are interested in real time sorting of airborne particles . There are of course other techniques that combine other features such as laser-induced breakdown spectroscopy (LIBS) , multiple elastic scattering measurements , or mass spectrometry  with LIF measurements. Those are beyond the scope of this paper.
5. Discrimination of B. subtilis-like aerosol particles against atmospheric aerosol using the six algorithms
To investigate the capability of the DPFS sensor to discriminate among different particle types, we choose as a test problem the determination of the number of particles in an atmospheric sample that are indistinguishable from one particular type of particle, the preparation of B. subtilis spores used above in Figs. 6, 8, and 9. We analyzed an atmospheric aerosol data ensemble collected at the Adelphi site during the time period from 5:00 PM on Sept. 17, 2009 to 6:00 PM on Sept 18, 2009. Altogether, 1,419,127 particles were sampled during this period. Of these, 351,943 particles, or 24.8%, had fluorescence intensity above the noise floor.
Figures 10 and 11 demonstrate the capability of the DPFS to discriminate among particles. The numbers and percentages of particles versus time that are assigned to be either biological (for algorithms 1 and 2) or B. subtilis-like (for algorithms 3-6) are displayed in Fig. 10A and 10B. Figure 11 illustrates the average percentage for the total 25 hr period.
We term the fluorescent particles found by algorithms 1 and 2 “biological” because: a) in many situations these particles will be primarily of biological origin, i.e., particles such as fungal spores, pollens, bacteria, viruses, proteins, and cellulose; and b) many researchers assume that they are primarily biological. Kanaani et al. , used a UV aerodynamic particle sizer (UVAPS) to excite fungal spores at 355 nm and measure fluorescence at 420 nm to 575 nm. They could detect the spores as biological particles. They did not think they could differentiate in an ambient air sample the different species of fungi they examined. Huffman et al. , used the UVAPS to measure fluorescent PBA for periods of months. They summarized arguments for why the particles measured by the UVAPS are primarily PBA. Some of these arguments also hold for why algorithm 1, which uses the total fluorescence excited by 263 nm, primarily measures biological aerosols. Although PAHs and HULIS (which are both common in the atmosphere) also fluoresce, the argument is that these fluorophors tend to occur in smaller particles, ones that are too small to be detected by the UVAPS as it is typically run. That argument would also apply to the DPFS. However, as noted above, very small particles can agglomerate into larger particles [40,41]. Gabey et al. , used two xenon light sources with peak emission around 280 nm and 370 nm (using a WIBS-3) to obtain total UV-fluorescence, visible fluorescence and elastic scattering to measure PBA in Indonesia. They argued that the great majority of the particles they measured were fungal spores, and they noted that there were no significant burn sites near their sampling points. A case where fluorescent particles may not be primarily biological is in the smoke from biomass burning. At some distances downwind from a significant burn site, the majority of the fluorescent particles having diameters greater than 1 μm will not be “biological.”
The relative concentrations and percentages of particles that are: a) fluorescent, with an assumption of likely being biological, for algorithms 1 and 2; or b) B. subtilis-like for algorithms 3-6, are shown to differ greatly. They differ by about a factor of 250 for the overall averages (Fig. 11) and by as much as three orders of magnitude for some time periods (e.g., 3 AM to 11 AM in Fig. 10). These six algorithms use increasingly more information available from the DPFS as illustrated in Table 2.
For algorithms 1 and 2 the percentages of particles detected as fluorescent are 16.2% and 16.5% on average. It may be that fewer particles are biological, as discussed above. For algorithms 3 to 6, which are constructed to detect B. subtilis against the natural atmospheric background, the corresponding percentages of particles that are B. subtilis-like are: 2.49%, 1.43%, 0.46%, and 0.06%, as summarized in Fig. 11.
The very large differences in specificities obtained with the six algorithms (as seen in Figs. 10 and 11 above) illustrate the diagnostic power and potential versatility of the DPFS. These six algorithms use successively more information measured by the DPFS regarding the fluorescence intensities, fluorescence spectral shapes, and elastic scattering, and have an ability to discriminate against an increasingly larger fraction of the atmospheric aerosol background in the case of particles having known fluorescence spectra. As discussed in Section 6.1, we envision concurrently running algorithms such as the six discussed here, along with others, so that multiple particle types (or preparations) can be examined at one time with different levels of specificity.
6.1 The DPFS can analyze aerosols in many ways at once
LIF measurements of atmospheric and other environmental aerosols can be used for a variety of objectives. For some applications, total PBA may be of primary interest. For such applications algorithm 3-6 may be too selective, and algorithms 1 and 2 might be more appropriate. For many applications, concurrent monitoring of multiple specific particle preparations, as well as total PBA as determined by algorithms 1 and 2, may be what is desired. For example, one may be interested in airborne allergens such as different types of pollens, fungal spores, feces of cockroach and dust mites, and protein allergens from cats and dogs. Or, for example, one may be interested in several different types of bacteria that have been generated in different ways, and/or in one species of bacteria grown or generated under several different conditions. For such applications, the versatility, diagnostic capability, and concurrent monitoring capability of the DPFS could be especially useful. A separate algorithm similar to algorithm 6 could be developed for each fluorescence spectral “signature” associated with each type of allergenic particle, or bacterium, or for each preparation of bacterial particle of the same type or species. These algorithms could be modified to be more or less specific depending upon the specificity desired for each type. Given the capabilities of modern computers it would be very feasible to search for many different target-particle LIF spectra at once. At the same time as particles are being classified with these quite specific algorithms and spectra, the particles could additionally be classified with other algorithms (maybe ones more similar to algorithms 1-5) to obtain the total numbers of fluorescent particles, or the number of particles having fluorescence within certain emission bands, etc., and possibly to include in these counts the number of bacterial particles of a particular species that may have been prepared in ways different from any of those used for the development of the highly specific algorithms.
Algorithm 6 was designed for one specific preparation of spores of one specific bacterium (B. subtilis). Because this preparation was not highly washed it would contain other molecules from the growth media or other sources. By constructing multiple algorithms similar to algorithm 6, one for each preparation or source of interest, even for one species of bacteria, it may be possible to distinguish among these different preparations, if the concentrations of particles of the specific type(s) are sufficiently above the ambient background.
6.2 The DPFS can be especially sensitive because it is so selective
For specific applications where well defined spectrum is known, e.g., for a specific preparation of bacteria or viruses, or for fungal spores grown under similar conditions, the DPFS can be highly sensitive. In Fig. 10 there is a period of about seven hours where the average particle count found by Algorithm 6 is less than 0.2 per minute. That background count is so low that a small increase in the counts for that type of particle could be noticeable. For example, if the count rose to 2 per minute for 5 minutes, that increase would be 10 times the average rate, and would rise above the background for that type of particle. Whether or not this high selectivity is considered a benefit or a problem depends on the objectives for the measurements.
Atmospheric aerosol is known to contain complex external and internal mixtures of organic and non-organic compounds. There is an enormous variability in the composition of particles containing organic carbon, especially in the biological fraction of organic carbon aerosol. Secondary plant and fungal metabolites probably play a large role in the diversity in the fluorescence of bioaerosols. Here we show that using a fluorescence-based sensor that measures the elastic scattering and UV-excited fluorescence spectra of particles at two wavelengths, certain target particles (a specific preparation of B. subtilis bacterial spores in the example here) can be largely discriminated against the natural atmospheric background. These tests are only for a certain site during a particular 25 hour period, under specific meteorological conditions. Further measurements will have to be carried out at sites with different regional climates and atmospheric environments to more fully assess the diagnostic power of the DPFS aerosol sensor. We envision the DPFS running multiple algorithms concurrently, with one or more algorithm for each particle type of interest.
This research was supported by the Defense Threat Reduction Agency under the Physical Science and Technology Basic Research Program, and by US Army Research Laboratory mission funds.
References and links
2. R. Jaenicke, S. Matthias-Maser, and S. Gruber, “Omnipresence of biological material in the atmosphere,” Environ. Chem. 4(4), 217–220 (2007). [CrossRef]
3. W. Elbert, P. E. Taylor, M. O. Andreae, and U. Pooschl, “Contribution of fungi to primary biogenic aerosols in the atmosphere: wet and dry discharged spores, carbohydrates, and inorganic ions,” Atmos. Chem. Phys. 7(17), 4569–4588 (2007). [CrossRef]
4. J. Fröhlich-Nowoisky, D. A. Pickersgill, V. R. Després, and U. Pöschl, “High diversity of fungi in air particulate matter,” Proc. Natl. Acad. Sci. U.S.A. 106(31), 12814–12819 (2009) (PNAS). [CrossRef] [PubMed]
5. S. Zappoli, A. Andracchio, S. Fuzzi, M. C. Facchini, A. Gelencser, G. Kiss, Z. Krivacsy, A. Molnar, E. Meszaros, H.-C. Hansson, K. Rosman, and Y. Zebuhr, “Inorganic, organic and macromolecular components of fine aerosol in different areas of Europe in relation to their water solubility,” Atmos. Environ. 33(17), 2733–2743 (1999). [CrossRef]
6. P. Herckes, M. P. Hannigan, L. Trenary, T. Lee, and J. L. Collett, “Organic compounds in radiation fogs in Davis, California,” Atmos. Res. 64(1-4), 99–108 (2002). [CrossRef]
7. G. Kiss, B. Varga, I. Galambos, and I. Ganszky, “Characterization of water-soluble organic matter isolated from atmospheric fine aerosol,” J. Geophys. Res. 107(D21), 8339 (2002). [CrossRef]
8. H. Bauer, A. Kasper-Giebl, M. Loflund, H. Giebl, R. Hitzenberger, F. Zibuschka, and H. Puxbaum, “The contribution of bacteria and fungal spores to the organic carbon content of cloudwater, precipitation and aerosols,” Atmos. Res. 64(1-4), 109–119 (2002). [CrossRef]
9. A. Gelencser, A. Hoffer, Z. Krivacsy, G. Kiss, A. Molnar, and E. Meszaros, “On the possible origin of humic matter in fine continental aerosol,” J. Geophys. Res. 107, ACH2 1–5 (2002).
10. M. Kanakidou, J. H. Seinfeld, S. N. Pandis, I. Barnes, F. J. Dentener, M. C. Facchini, R. Van Dingenen, B. Ervens, A. Nenes, C. J. Nielsen, E. Swietlicki, J. P. Putaud, Y. Balkanski, S. Fuzzi, J. Horth, G. K. Moortgat, R. Winterhalter, C. E. L. Myhre, K. Tsigaridis, E. Vignati, E. G. Stephanou, and J. Wilson, “Organic aerosol and global climate modeling: a review,” Atmos. Chem. Phys. 5(4), 1053–1123 (2005). [CrossRef]
11. V. Samburova, R. Zenobi, and M. Kalberer, “Characterization of high molecular weight compounds in urban atmospheric particles,” Atmos. Chem. Phys. 5(8), 2163–2170 (2005). [CrossRef]
12. R. G. Pinnick, S. C. Hill, P. Nachman, J. D. Pendleton, G. L. Fernandez, M. W. Mayo, and J. G. Bruno, “Fluorescence Particle Counter for Detecting Airborne Bacteria and Other Biological Particles,” Aerosol Sci. Technol. 23(4), 653–664 (1995). [CrossRef]
13. P. P. Hairston, J. Ho, and F. R. Quant, “Design of an instrument for real-time detection of bioaerosols using simultaneous measurement of particle aerodynamic size and intrinsic fluorescence,” J. Aerosol Sci. 28(3), 471–482 (1997). [CrossRef] [PubMed]
14. J. D. Eversole, J. J. Hardgrove, W. K. Cary, D. P. Choulas, and M. Seaver, “Continuous, rapid biological aerosol detection with the use of UV fluorescence: Outdoor test results,” Field Anal. Chem. Technol. 3(4-5), 249–259 (1999). [CrossRef]
15. F. L. Reyes, T. H. Jeys, N. R. Newbury, C. A. Primmerman, G. S. Rowe, and A. Sanchez, “Bio-aerosol fluorescence sensor,” Field Anal. Chem. Technol. 3(4-5), 240–248 (1999). [CrossRef]
16. L. M. Brosseau, D. Vesley, N. Rice, M. N. Goodell, and P. Hairston, “Differences in detected fluorescence among several bacterial species measured with a direct-reading particle sizer and fluorescence detector,” Aerosol Sci. Technol. 3, 545–558 (2000).
17. P. H. Kaye, J. E. Barton, E. Hirst, and J. M. Clark, “Simultaneous light scattering and intrinsic fluorescence measurement for the classification of airborne particles,” Appl. Opt. 39(21), 3738–3745 (2000). [CrossRef]
18. R. DeFreez, “LIF bio-aerosol threat triggers: then and now,” Proc. SPIE 7484, 74840H (15 pp.) (2009).
19. S. C. Hill, R. G. Pinnick, P. Nachman, G. Chen, R. K. Chang, M. W. Mayo, and G. L. Fernandez, “Aerosol-Fluorescence Spectrum Analyzer: Real-Time Measurement of Emission Spectra of Airborne Biological Particles,” Appl. Opt. 34(30), 7149–7155 (1995). [CrossRef] [PubMed]
20. G. Chen, P. Nachman, R. G. Pinnick, S. C. Hill, and R. K. Chang, “Conditional-firing aerosol-fluorescence spectrum analyzer for individual airborne particles with pulsed 266-nm laser excitation,” Opt. Lett. 21(16), 1307–1309 (1996). [CrossRef] [PubMed]
21. R. G. Pinnick, S. C. Hill, P. Nachman, G. Videen, G. Chen, and R. K. Chang, “Aerosol fluorescence spectrum analyzer for rapid measurement of single micrometer-sized airborne biological particles,” Aerosol Sci. Technol. 28(2), 95–104 (1998). [CrossRef]
22. Y. S. Cheng, E. B. Barr, B. J. Fan, P. J. Hargis, D. J. Rader, T. J. O’Hern, J. R. Torczynski, G. C. Tisone, B. L. Preppernau, S. A. Young, and R. J. Radloff, “Detection of bioaerosols using multiwavelength UV fluorescence spectroscopy,” Aerosol Sci. Technol. 30(2), 186–201 (1999). [CrossRef]
23. Y. L. Pan, S. Holler, R. K. Chang, S. C. Hill, R. G. Pinnick, S. Niles, and J. R. Bottiger, “Single-shot fluorescence spectra of individual micrometer-sized bioaerosols illuminated by a 351- or a 266-nm ultraviolet laser,” Opt. Lett. 24(2), 116–118 (1999). [CrossRef]
24. S. C. Hill, R. G. Pinnick, S. Niles, Y. L. Pan, S. Holler, R. K. Chang, J. R. Bottiger, B. T. Chen, C.-S. Orr, and G. Feather, “Real-time measurement of fluorescence spectra from single airborne biological particles,” Field Anal. Chem. Technol. 3(4-5), 221–239 (1999). [CrossRef]
25. K. L. Schroder, P. J. Hargis Jr, R. L. Schmitt, D. J. Rader, and I. R. Shokair, ““Development of an unattended ground sensor for ultraviolet laser-induced fluorescence detection of biological agent aerosols,” Air Monitoring and Detection of Chemical and Biological Agents II,” Proc. SPIE 3855, 82–91 (1999). [CrossRef]
26. Y. L. Pan, P. Cobler, S. Rhodes, A. Potter, T. Chou, S. Holler, R. K. Chang, R. G. Pinnick, and J.-P. Wolf, “High-speed, high-sensitivity aerosol fluorescence spectrum detection using a 32-anode photomultiplier tube detector,” Rev. Sci. Instrum. 72(3), 1831–1836 (2001). [CrossRef]
27. Y. L. Pan, J. Hartings, R. G. Pinnick, S. C. Hill, J. Halvorson, and R. K. Chang, “Single-particle fluorescence spectrometer for ambient aerosols,” Aerosol Sci. Technol. 37(8), 628–639 (2003). [CrossRef]
28. Y. L. Pan, R. G. Pinnick, S. C. Hill, J. M. Rosen, and R. K. Chang, “Single-particle laser-induced-fluorescence spectra of biological and other organic-carbon aerosols in the atmosphere: measurements at New Haven, CT and Las Cruces, NM, USA,” J. Geophys. Res. 112, D24S19 (2007). [CrossRef]
29. A. Manninen, M. Putkiranta, J. Saarela, A. Rostedt, T. Sorvajärvi, J. Toivonen, M. Marjamäki, J. Keskinen, and R. Hernberg, “Fluorescence cross sections of bioaerosols and suspended biological agents,” Appl. Opt. 48(22), 4320–4328 (2009). [CrossRef] [PubMed]
30. Y. L. Pan, R. G. Pinnick, S. C. Hill, and R. K. Chang, “Particle-fluorescence spectrometer for real-time single-particle measurements of atmospheric organic carbon and biological aerosol,” Environ. Sci. Technol. 43(2), 429–434 (2009). [CrossRef] [PubMed]
31. P. H. Kaye, W. R. Stanley, E. Hirst, E. V. Foot, K. L. Baxter, and S. J. Barrington, “Single particle multichannel bio-aerosol fluorescence sensor,” Opt. Express 13(10), 3583–3593 (2005). [CrossRef] [PubMed]
32. A. M. Gabey, M. W. Gallagher, J. Whitehead, and J. Dorsey, “Measurements of coarse mode and primary biological aerosol transmission through a tropical forest canopy using a dual-channel fluorescence aerosol spectrometer,” Atmos. Chem. Phys. Discuss. 9(5), 18965–18984 (2009). [CrossRef]
34. V. Sivaprakasam, T. Pletcher, J. E. Tucker, A. L. Huston, J. McGinn, D. Keller, and J. D. Eversole, “Classification and selective collection of individual aerosol particles using laser-induced fluorescence,” Appl. Opt. 48(4), B126–B136 (2009). [CrossRef] [PubMed]
35. G. Feugnet, E. Lallier, A. Grisard, L. McIntosh, J. E. Hellström, P. Jelger, F. Laurell, C. Albano, M. Kaliszewski, M. Wlodarski, J. Mlynczak, M. Kwasny, Z. Zawadzki, Z. Mierczyk, K. Kopczynski, A. Rostedt, M. Putkiranta, M. Marjamäki, J. Keskinen, J. Enroth, K. Janka, R. Reinivaara, L. Holma, T. Humppi, E. Battistelli, E. Iliakis, and G. Gerolimos, “Improved laser-induced fluorescence method for bio-attack early warning detection system,” Proc. SPIE 7116, 71160C (2008). [CrossRef]
36. Y. L. Pan, R. G. Pinnick, S. C. Hill, H. Huang, and R. K. Chang, “Dual-wavelength-excitation single-particle fluorescence spectrometer/particle sorter for real-time measurement of organic carbon and biological aerosols,” Optically Based Biological and Chemical Detection for Defence IV, Proc. SPIE 7116, 71160J–1-8 (2008).
37. H. C. Huang, Y. L. Pan, S. C. Hill, R. G. Pinnick, and R. K. Chang, “Real-time measurement of dual-wavelength laser-induced fluorescence spectra of individual aerosol particles,” Opt. Express 16(21), 16523–16528 (2008). [CrossRef] [PubMed]
38. R. J. Cole, M. A. Schweikert, and B. B. Jarvis, Handbook of Secondary Fungal Metabolites, (Academic, New York, 2003), Vol. 1–3.
39. I. B. Berlman, Handbook of Fluorescence Spectra of Aromatic Molecules, (Academic, New York, 1971), 2nd ed.
40. C. V. Gulijk, J. C. M. Marijnissen, M. Makkee, J. A. Moulijn, and A. J. Schmidt-Ott, “Measuring diesel soot with a scanning mobility particle sizer and an electrical low-pressure impactor: performance assessment with a model for fractal-like agglomerates,” J. Aerosol Sci. 35(5), 633–655 (2004). [CrossRef]
41. R. C. Moffet and K. A. Prather, “In-situ measurements of the mixing state and optical properties of soot with implications for radiative forcing estimates,” P. National Academy of Sciences 106, 11872–11877 (2009). [CrossRef]
42. Z. Krivácsy, G. Kiss, B. Varga, I. Galambos, Z. Sarvari, A. Gelencser, A. D. Molnar, S. Fuzzi, M. C. Facchini, S. Zappoli, A. Andracchio, T. Alsberg, H. C. Hansson, and L. Persson, “Study of humic-like substances in fog and interstitial aerosol by size-exclusion chromatography and capillary electrophoresis,” Atmos. Environ. 34(25), 4273–4281 (2000). [CrossRef]
43. C. L. Muller, A. Baker, R. Hutchinson, I. J. Fairchild, and C. Kidd, “Analysis of rainwater dissolved organic carbon compounds using fluorescence spectrophotometry,” Atmos. Environ. 42(34), 8036–8045 (2008). [CrossRef]
44. V. R. Després, J. F. Nowoisky, M. Klose, R. Conrad, M. O. Andreae, and U. Pöschl, “Characterization of primary biogenic aerosol particles in urban, rural, and high-alpine air by DNA sequence and restriction fragment analysis of ribosomal RNA genes,” Biogeosciences 4(6), 1127–1141 (2007). [CrossRef]
45. J. R. Bottiger, P. J. Deluca, E. W. Stuebing, and D. R. VanReenen, “An Ink-Jet Aerosol Generator,” J. Aerosol Sci. 29, s965–s966 (1998). [CrossRef]
46. R. G. Pinnick, S. C. Hill, Y. L. Pan, and R. K. Chang, “Fluorescence spectra of atmospheric aerosol at Adelphi, Maryland, USA: measurement and classification of single particles containing organic carbon,” Atmos. Environ. 38(11), 1657–1672 (2004). [CrossRef]
47. H. Kanaani, M. Hargreaves, J. Smith, Z. Ristovski, V. Agranovski, and L. Morawska, “Performance of UVAPS with respect to detection of airborne fungi,” J. Aerosol Sci. 39(2), 175–189 (2008). [CrossRef]
48. J. A. Huffman, B. Treutlein, and U. Pöschl, “Fluorescent biological aerosol particle concentrations and size distributions measured with an Ultraviolet Aerodynamic Particle Sizer (UV-APS) in Central Europe,” Atmos. Chem. Phys. 10(7), 3215–3233 (2010). [CrossRef]
49. J. D. Hybl, S. M. Tysk, S. R. Berry, and M. P. Jordan, “Laser-induced fluorescence-cued, laser-induced breakdown spectroscopy biological-agent detection,” Appl. Opt. 45(34), 8806–8814 (2006). [CrossRef] [PubMed]
50. E. J. McJimpsey, P. T. Steele, K. R. Coffee, P. Fergenson, V. J. Riot, B. W. Woods, E. E. Gard, M. Frank, H. J. Tobias, and C. Lebrilla, “Detection of Biological Particles in Ambient Air Using BioAerosol Mass Spectrometry,” Proc. SPIE 6218, Chemical and Biological Sensing VII, 62180B (2006).