1. Introduction

Dynamic photoacoustic spectroscopy (DPAS) was recently demonstrated with remarkable sensitivity to trace gases in a study by the Massachusetts Institute of Technology, Lincoln Laboratory (MIT-LL) [1,2]. To gain a better understanding of aerosol photoacoustics, several chemical threat reduction studies [3–5] and environmental monitoring studies [6–8] have sought to uncover the feasibility of a photoacoustic-based aerosol detector. However, previous studies have utilized some type of flow-through photoacoustics (resonance cell), which implies an extractive sampling approach toward sensing a threat volume. We aim to show DPAS’s utility as an aerosol detector, enabling the advantages of photoacoustic signature recovery while obviating the need for a sampling system. A field-portable DPAS system may prove useful for the detection and identification of aerosolized chemical agents, providing an early warning system by sensing threats at standoff distances. Such a system may also enable stand-off detection of atmospheric aerosol particulate matter (PM) useful for environmental monitoring.

2. Experimental methods

DPAS functions by sweeping a laser beam at sonic speed through a sensing region, allowing coherent addition of the photoacoustic wavefronts and thus amplifying the photoacoustic signal. The current iteration of our DPAS system is equipped with a 10 W cw CO₂ laser (~5 W on target) from Access Laser Company (L20G) tunable between 9.2 μm – 10.8 μm. The laser has a full divergence angle of 5.5 mrad, and spot sizes of < 18 mm at a distance of 2.6 m from the laser output were routinely measured. A 5 mW 635 nm diode laser (Laserglow Technologies LBD-635-TD-5) is collinearly aligned with the CO₂ laser beam to aid downrange alignment. A 1” optical cube mounted on a Scitec 300CD optical chopper wheel sweeps the two laser beams in a clockwise motion towards an omnidirectional microphone (Earthworks M30; bandwidth = 50 kHz; sensitivity = 18.0 V/Pa) placed in the far corner of an aerosol/gas chamber, ~1.2 m downrange of the optical cube (see Fig. 1).An Earthworks Microphone Preamp model 1021 set to 55 dB gain powered our microphone. A more detailed description of our DPAS system can be found elsewhere [1,2].

 

Fig. 1 Experimental setup of the DPAS system and aerosol chamber.

Download Full Size | PDF

Experiments were performed at MIT-LL as well as Edgewood Chemical and Biological Center (ECBC). There were several differences between aerosol chambers at each location. At MIT-LL, an anechoic sample chamber was available (length, width, and height of 2.4 m, 0.9 m, 0.3 m respectively) which allowed for a laser beam integration path of 0.45 m prior to reaching the microphone, see “integration path length” in Fig. 1. Aerosols were disseminated into the chamber by means of an eductor tube backed by a cylinder of dry nitrogen. The aerosol chamber at ECBC was a double walled acrylic chamber (interior length, width, and height of 1 m, 0.5 m, 0.5 m respectively) which did not include anechoic foam. Space constraints at ECBC limited the laser beam integration path to 0.2 m. Aerosols were delivered in the ECBC sample chamber by a powder coating gun (Chicago Electric Power Tools 94244).

Software written in-house for the DPAS system enables an IR spectrum to be collected by stepping through the CO₂ laser’s ~58 wavelengths and measuring the DPAS signal. The DPAS signal at each wavelength is obtained by averaging successive waveforms over a given integration time and then calculating the peak-to-peak microphone signal within the expected signal timeframe. Each spectrum measured by the system is laser-power normalized (power was monitored by a pickoff detector) to adjust for variations in the laser power at each wavelength. The spinning mirror rotation rate is stabilized at the operating frequency via a PID control loop (in the vicinity of 23 Hz) with a standard deviation of less than 2∙10⁻⁴ Hz over 10 minutes. The TTL reference signal from the Scitec optical chopper is used as the trigger source during data acquisition. Total system jitter has been determined to be < 10 µs, a 10x improvement over the estimated ~100 µs jitter of our previous DPAS system [1]. Time averaging successive DPAS waveforms requires very good temporal precision in order to properly align them. Given the DPAS waveform timescale of ~100 µs, our jitter of 10 µs is a very significant improvement as it allows us to approach a situation in which we are coherently averaging, whereas with 100 µs of jitter, we were effectively incoherently averaging with our previous system. Other notable improvements over our previous system include the usage of an optical cube in place of a one-sided spinning mirror (4x improvement in data acquisition rate), and improved optical throughput by switching to higher quality optical components (~1.5x).

3. Aerosol detection

3.1 DPAS signal versus laser beam sweep speed

An eductor tube backed by dry nitrogen was used to disseminate approximately 100 mg of Syloid 244 (porous silica; average particle size = 3.2 μm) into MIT-LL’s anechoic chamber. The CO₂ laser was tuned to a Syloid 244 absorption feature at 9.2714 µm and swept through the aerosol towards the microphone, placed at a distance of ~1.2 m from the spinning mirror. The laser beam integration path was approximately 0.45 m. The rotational speed of the spinning mirror was increased from 18 Hz to 28 Hz in 0.2 Hz increments and a DPAS signal was collected (2 second integration time) after each increment in rotational speed.

Figure 2(a) shows an image plot of the time series data with respect to laser beam sweep velocity. The peak-to-peak amplitude of the DPAS signal versus laser beam sweep velocity is shown in Fig. 2(b). Maximum DPAS signal occurs at a laser beam sweep velocity of ~345 m/s, which corresponds to the speed of sound in air at standard temperature and pressure (STP). At Mach 1, the photoacoustic source and signal propagate in phase, allowing maximum signal amplification [1]. Data for trace amounts of SF₆ gas is shown on the same plot for comparison, and we find the aerosol and gas to display comparable resonance characteristics.

 

Fig. 2 DPAS detection of Syloid 244 aerosol. (a) waterfall plot of DPAS signal. (b) resonance curves for Syloid 244 aerosol and SF₆ gas. (c) a comparison of the temporal waveforms produced by Syloid 244 (aerosol), SiO₂ spheres (aerosol), and MeOH, SF₆ (gases). Resonance curves and temporal waveforms have been normalized to demonstrate the similarities in waveform.

Download Full Size | PDF

The DPAS temporal waveform produced by Syloid 244 (at Mach 1) is shown in Fig. 2(c). Temporal waveforms produced by 0.40 µm silica spheres (discussed in more detail in the sensitivity section), SF₆ and methanol vapor are displayed on the same plot for comparison. All samples tested are found to have comparable temporal waveforms, each with a width of approximately 100 µs. Curves in Figs. 2(b) and 2(c) were normalized to their respective maximum magnitudes.

3.2 IR spectra

Prior to measuring an IR spectrum of Syloid 244, we validated our DPAS system’s capability (with a laser beam sweep velocity of Mach 1) by taking a spectrum of methanol vapor. Less than 1 mL of methanol was placed into a petri dish on the sample chamber floor, approximately 15 cm below the path of the IR laser beam. An IR spectrum was obtained by stepping through the CO₂ laser’s ~58 wavelengths with one second integration time (92 scans through the sample = 23 Hz ∙ 4 cube sides). The resulting spectrum is shown in Fig. 3 (top) and is in overall good agreement with methanol’s absorbance spectrum [9] (overlaid on the same plot). Disagreement in the spectra within the ~10.5 µm – 10.8 µm range is likely due to impurities in the non-spectroscopic grade MeOH used during experiment.

 

Fig. 3 DPAS spectrum of methanol vapor (top) and Syloid 244 aerosol (bottom).

Download Full Size | PDF

After clearing the chamber of any residual methanol, approximately 100 mg of Syloid 244 was disseminated into the sample chamber and an IR spectrum of the aerosol was obtained using a one second integration time at each wavelength (Fig. 3, bottom). The DPAS spectrum of Syloid 244 is in good agreement with the Syloid 244 extinction coefficient spectrum collected via FTIR in previous work [10].

4. Aerosol detection sensitivity

For aerosol detection sensitivity measurements, we used our DPAS system to interrogate monodispersed silica spheres (AngstromSpheres by Fiber Optic Center, Inc.) disseminated into ECBC’s sample chamber while recording particle concentrations with a PALAS Promo 3000 particle counter equipped with a Welas 2100 sensor (500∙10³ particles/mL, 0.2 µm – 10 µm). We experimented with particle sizes varying from 0.1 µm – 10 µm; however, the smaller particles we investigated appeared to clump together to form either ~0.4 µm or 0.68 µm particles, and particles larger than 1 µm tended to settle out too quickly for us to make meaningful measurements. Particle sizes were determined experimentally and are reported as the mode particle size given by the particle counter over each release. For each sensitivity measurement, we tuned our CO₂ laser to a SiO₂ absorption feature at 9.24 µm, disseminated silica spheres of a specific size and recorded the DPAS signal and particle concentration over time as the particles settled in the sample chamber. Figure 4 shows the net particle count and the distribution of particle size over time during the release of 0.39 µm silica spheres. At approximately 1100 seconds, the particle count rapidly drops as we begin to pump on the sample chamber to accelerate the settling of particles.

 

Fig. 4 Particle concentration data during a release of 0.39 um silica spheres. Net particle concentration (top) and aerosol size distribution (bottom) are displayed versus time.

Download Full Size | PDF

We converted particle concentrations from particles/mL to ppm_mass assuming a spherical geometry and a SiO₂ density of 2.634 g/mL, and by using an absorption coefficient of α = 4.2∙10⁻⁴ (ppm_mass∙m)⁻¹ [10] we calculated absorbance. Peak-to-peak microphone signals are displayed as a function of absorbance in Fig. 5.In the same figure, we compare against SF₆ sensitivity measurements collected at MIT-LL with our CO₂ laser tuned to an SF₆ absorption line at 10.591 µm. SF₆ absorbance values were calculated from gas concentrations reported by an Online Technologies (now MKS Instruments) FTIR spectrometer. Each data point represents ~30 second averages. The noise floor shown in Fig. 5 represents the standard deviation of the background noise.

 

Fig. 5 DPAS signal versus absorbance for SiO₂ aerosols (λ = 9.24 μm) and SF₆ gas (λ = 10.591 μm).

Download Full Size | PDF

In order to estimate the DPAS system sensitivity, we applied a linear fit to the SiO₂ aerosol data and SF₆ data in the region where we have data for both (between 10⁻⁴ – 6∙10⁻³ absorbance). Because of the particle clumping / settling issues we experienced, we were not able to obtain a wide enough spread in particle size to draw any particle size related conclusions. To first order, we consider the absorbance curves produced by each of the SiO₂ releases to be comparable, and we group them together when fitting the data to estimate sensitivity. Linear fits yield nearly identical sensitivity values for SiO₂ aerosols and SF₆ gas, ~98 V/A (volts / absorbance unit) and ~94 V/A respectively. Linear extrapolating the data to the noise floor (~1 mV, 1 s integration time), we estimate the minimum measurable absorbance of our system to be ~10⁻⁵; this minimum measureable absorbance corresponds to a SiO₂ particle concentration of ~500 particles/mL for the range of particles sizes we experimented with. Sensitivity could be improved by lowering the noise floor, either by increasing the integration time, or by utilizing a directional microphone. The effect of particle size on absorbance remains an open question. An alternative dissemination method that minimizes clumping and maximizes settling time could allow for more careful monodispersed aerosol measurements in future studies.

4. Discussion

In order to make performance predictions for an aerosol-sensing DPAS system, we require an understanding of the relevant physical parameters and their effect on the DPAS signal. The most basic question (and focus of this paper) is how does the aerosol DPAS signal compare to the vapor DPAS signal? Figure 2 clearly indicates that for the range and type of particles examined there is very little difference between the vapor and aerosol response (particularly their temporal responses) implying that similar time constants dominate their DPAS responses. Additionally, Fig. 5 shows that the signal magnitudes are very similar for a given absorption (absorptivity ∙ path length). Previous studies of static photoacoustics spectroscopy (PAS) by Gurton [3] have used the term “gaseous equivalence” to describe the similarity between the vapor and aerosol response. The authors analyzed the temporal response of particles of various sizes and compared them to the equivalent gaseous response, modeling the heat flow into and out of spherical particles into a surrounding fluid (air) at STP. Their analysis indicated that particle size and particle thermal conductivity are critical to achieving “gaseous equivalence”. For a given time scale (modulation frequency in the case of PAS), a particle must be above a minimum thermal diffusivity and below a maximum size in order for the efficient transfer of acoustic energy to occur without appreciable time lags. Whereas PAS utilizes a chopped laser to induce a time-varying heat profile, DPAS utilizes a cw laser, relying upon the changing location of the laser beam to induce a temperature change in the gas / aerosol. The relevant timescale for PAS is the period of the chopped laser. For DPAS, we estimate the relevant timescale over which the temperature changes as the time the laser beam takes to traverse a given volume, which is related to the laser spot size by t ~(spot size) / (sweep velocity) ~50 µs. Given our particle sizes (~1 µm) and the thermal diffusivity of SiO₂ (0.0083 cm²/s), the analysis of Gurton predicts that we are operating in the regime of “gaseous equivalence”, consistent with our data.

In order to put our particle sizes in the context of relevant length scales we use the homogeneous heat diffusion equation, which can be written in one-dimension as:

While the minimum measureable particle concentration we have demonstrated may seem low compared to other aerosol detection techniques (such as the TSI Aerosol Particle Sizer Model 3321, which advertises a minimum particle concentration limit of 0.001 particles/mL), we believe DPAS has several advantages over these types of sensors. DPAS is a novel approach for aerosol (and vapor) detection with the added benefit of spectroscopic identification, either as a point sensor or as a remote detection technique. Additionally, DPAS has the potential to sample larger volumes per unit time than commercially available particle counters, without the need for flow control, and without suffering from high particle loads as many particle counters do.

5. Conclusions

We demonstrated the utility of DPAS as a trace aerosol detection technique. We detected a release of Syloid 244 and observed that its temporal response is similar to gases. An IR spectrum of Syloid 244 was recorded using DPAS and found to be in good agreement with the published extinction coefficient spectrum of Syloid 244. Our DPAS sensitivity to 0.39 µm – 0.68 µm SiO₂ spheres (~98 V/A) is comparable to SF₆ gas (~94 V/A). Future experiments are planned for studying the effect of aerosol particle size on absorbance.