Laser-induced incandescence (LII) measures the thermal emission from soot particles heated by a laser to temperatures in the 2500–4500 K range (Bachalo et al., 2002). Assuming that all the volatile and semi-volatile OC that condenses on the BC particles will be evaporated promptly at these temperatures, the LII signals are directly related to the non-volatile particles, primarily refractory black carbon (rBC), a form of carbon directly related to EC (Baumgardner et al., 2012). The LII 300 instrument calculates a mass concentration from the particle emission signal and the temperature determined from the signal at two wavelengths (446 and 720 nm) using two-colour pyrometry. As the optical properties of the particle may vary from source to source, a correlation factor may need to be applied to this optical determination of the mass concentration to relate the result to that of EC determined from TOA. A multimode pulsed Nd:YAG laser, operating at a fundamental wavelength of 1064 nm, with a pulse duration of 7 ns full width half maximum (FWHM) and a repetition rate of 20 Hz was used as the excitation source. The laser beam passed through a set of optics and formed into a square cross-section light beam uniformly incident throughout the probe volume. The measurement volume is approximate 14.7 mm3, and the range of fluence levels is typically 0.6–3.2 mJ mm-2.

At low laser fluence levels, the peak LII signal intensity rises with the laser fluence as peak particle temperatures increase. As laser fluence is increased, a threshold is reached where the measured mass concentration becomes independent of laser fluence. This plateau region extends for a range of increasing laser fluence until a level is reached where soot sublimation begins to dominate. At these very high laser fluences, the mass loss associated with soot sublimation has an influence, and a reduction in the mass concentration may be observed (Michelsen et al., 2015). Various soot fragments evaporate from the particle surface when sublimation occurs (Schulz et al., 2006). Models usually assume these to range from C1 to C7 (molecules having one to seven carbon atoms), dependent on the particle temperature (Leider et al., 1973). Additional research is required to understand the sublimation loss, which is complicated by uncertainties in physical parameters of the carbon species, limiting the ability of the models to fully predict phenomena in this regime (Michelsen et al., 2015).

In the AC-LII technique (Snelling et al., 2005), the soot temperatures are determined by the two-colour pyrometry method (in the LII 300, narrow bandpass filters centred at ∼446 nm and ∼720 nm are used with bandwidths of 40 and 20 nm, respectively). The peak soot temperature and absolute intensity of the LII signal were used to calculate soot volume fraction (fv). With soot particle material density (ρs) from the literature, the mass concentrations of the non-volatile particles, or EC, are obtained from fv × ρs. The primary particle sizes are determined from the temperature decay rates with the assumption that conduction dominates the particle cooling rates for conditions below the sublimation limit (Smallwood, 2008).

Theoretically, the mass concentration obtained from the two-colour method is valid regardless of the laser fluence applied, and a lower laser fluence level might be preferred to avoid the reduction in particle volume (sublimation) at high laser fluence. Practically, care should be taken with low laser fluence levels for particles from unknown sources with unknown combination of size distributions and morphologies as the assumption of the uniform temperature in the sampling volume could be invalid (Liu et al., 2016). The particles with various sizes could reach different peak temperatures at separate times, which induce uncertainties with the determination of the effective peak particle temperatures and in turn uncertainties with the derived mass concentration values. For a typical effective soot temperature of 4000 K, a 100 K error (2.5 % of the soot temperature) can lead to 15 % error in the estimated soot volume fraction value (Boiarciuc et al., 2006). It is therefore important to operate the LII 300 with optimum laser fluence, in a region of laser fluence where the reported mass concentration is relatively independent of fluence, to minimise uncertainties. In the current work, laser fluence sweep experiments (varying laser fluence from low levels to high fluence conditions with all other parameters held constant) were performed for each source at different operating conditions. This was performed to examine the relationship between the measured mass concentrations and LII fluence levels for multiple sources and to subsequently determine the optimum laser fluence level for both calibration and application to real-time nvPM mass measurements from aircraft engines.

TOA measures the mass of both organic carbon (OC) and elemental carbon (EC) of the particulate matter sampled on a quartz filter mounted in a stainless steel filter holder. This is not a gravimetric measurement where the mass of the elemental carbon was measured, but rather a chemical analysis of the black carbon on the filter. The NIOSH 5040 (NIOSH, 2003) standard method was used for the TOA, with a modified temperature ramp as specified by ARP 6320 (SAE, 2018), to obtain the EC mass concentrations. The uncertainty associated with this method for EC mass concentration is stated to be 16.7 % (NIOSH, 2003) due to additional uncertainty in assigning the split between OC and EC. The uncertainty associated with the total carbon (TC) mass concentration is 8.4 % (Conrad and Johnson, 2019). A basic overview of the process is presented here. The sample flow through the quartz filters was set to 22.7 slpm for all measurement conditions, which was controlled by a calibrated mass flow controller. Mass flow controllers were calibrated using a Sensidyne Gillibrator-2 bubble flow meter and corrected to the standard temperature and pressure reference conditions of 0 ∘C and 1 atm. The total volume of sample on the quartz filters varied from 360 to 1820 L, depending on the loading for a particular condition to achieve a target mass density of >10 mg  cm-2. The particulate matter (mostly black carbon) in the sample flow was captured by the filter. After collection, a 1 cm × 1 cm punch from the filter was cut out and mounted in an analyser (Sunset Laboratory OC-EC Aerosol Analyser) to obtain the EC and OC content. The filter punch was heated in the analyser oven in stages. From 0 to 425 s, the sample was surrounded by an inert helium atmosphere, and the heating process drove off any embedded volatile organic compounds (VOCs) from the sample. The desorbed VOCs were further drawn into a series of catalyst beds, where they were converted to CO2 and then to methane, whose concentration was measured by a flame ionisation detector (FID) to give a measure of the organic carbon fraction on the filter punch. As the VOCs are heated, it is possible that some were pyrolysed and converted to char, which stayed on the filter punch, instead of being driven off the surface. The quantity of the char was measured via laser extinction propagated through the sampled filter punch. From 425 to 800 s, the sample was surrounded by a reactive oxygen–helium atmosphere. As the oven heated the sample, the elemental carbon reacted with the oxygen to form CO2, which was again converted to methane and measured by the FID. The elemental carbon fraction of black carbon on the filter punch was then measured by subtracting the char (which was originally organic carbon) content from the total. With the two mass values quantified, i.e. the OC mass fraction and the EC mass fraction, the total carbon (TC) on the filter punch can be obtained. By multiplying by the stain area of the filter, the total mass of EC, OC, and TC on the entire filter was quantified. Having also measured the total volume of sampled gas, the mass concentration of the EC, OC, and TC was obtained. One uncertainty in the OC quantification from the TOA measurements was the artefact effect caused by the gas-phase semi-volatile organic compounds absorbed on the filter (Durdina et al., 2016). To correct this artefact, the OC determined from a secondary filter that contained exclusively absorbed gas-phase OC was subtracted from the OC mass determined from the primary filter.

In this study, a photoacoustic extinctiometer (PAX; Droplet Measurement Technologies, Inc.) and a micro soot sensor (MSS; AVL, model 483) were used to provide additional measurements for nvPM mass concentrations. Both the PAX and MSS are based on the photoacoustic method (Adams et al., 1989), measuring the acoustic wave generated by the heated gases surrounding the light-absorbing particles due to their increased temperature after interaction with a laser. Unlike filter-based absorption methods, where uncertainties related to organic aerosol coating and light-absorbing properties are issues (Baumgardner et al., 2012; Corbin et al., 2019), photoacoustic measurements can potentially be affected by the humidity via latent heat and mass transfer from volatile droplets (Arnott, 2003) and absorption by molecular species present in the gaseous medium. The PAX and the MSS measurements rely on the light-absorption properties of the PM sources, for which a mass absorption cross-section (MAC) is usually assumed. In this work, a MAC of 7.5 ± 1.2 m2 g-1 at a wavelength of 550 nm was applied to the PAX following the recommendation by Bond and Bergstrom (2006), noting that the MAC will vary with the black carbon source and operating condition. The MAC for the PAX (with a measured absorption at a wavelength λ of 870 nm) was converted using MAC(λ) = MAC(550 nm) × (550 nm/λ), assuming the refractive index was the same at those wavelengths, equivalent to stating that the aerosol Ångström exponent (AAE) is 1. The PAX operates at a wavelength of 870 nm and the MSS at 808 nm. The MSS was previously calibrated with a soot source (CAST) of known concentration (determined by comparison with EC from TOA), while the PAX was calibrated separately with ammonium sulfate and Aquadag solutions.

Although the focus in this work was on the quantification of nvPM mass concentration, the particle size distributions from different engine sources were measured using a scanning mobility particle sizer (SMPS; TSI, USA). The SMPS consisted of an electrostatic classifier (model 3082), a differential mobility analyser (model 3081), a soft X-ray neutraliser (model 3088), and an ultrafine condensation particle counter (model 3776). The particle size distribution data of the nvPM emissions from the various sources provide additional information on the physical properties of the particles. This will also provide information for the LII model development in terms of particle size distribution functions. The typical size range for a SMPS scan was 7–206 nm.

Figure 4 shows the LII 300 laser fluence sweep vs. normalised mass concentration (solid symbols) from the Gnome engine exhaust (Rig A) fuelled with kerosene at two different steady-state engine operating conditions – (1) idle and (2) high power output (HPO). It is anticipated that the idle and HPO conditions are two extremes in terms of the properties of the nvPM being emitted from Rig A, with those at idle being less mature and with more volatile organic compounds and the opposite for the HPO condition. Performing fluence sweep tests at these two conditions aided in determining an optimum laser fluence level to cover the full range of conditions from this source for real-time nvPM mass measurements. At laser fluence levels below 1.7 mJ mm-2 (Fig. 4), the measurements showed an almost monotonic increase with fluence between the measured nvPM mass concentrations and the laser fluence. The measurement of nvPM mass concentration was independent of the laser fluence levels in the range of 1.9 to 2.5 mJ mm-2 for the HPO condition and 2.2 to 3.2 mJ mm-2 for the idle condition. The shaded area in Fig. 4 corresponds to the data range that is within 2 % of the peak value of the Loess best-fit curve

The region near the peak of the Loess best fit within the 2 % error band is defined and referred to as a plateau regime (relatively uniform response to the fluence level) in the following sections. The optimum fluence discussed later is in the plateau regime.

It was observed that at the HPO condition for Rig A, the maximum nvPM mass concentration was reached at a laser fluence of ∼2.25 mJ mm-2, beyond which the nvPM mass concentration decreased moderately with increasing laser fluence, attributable to sublimation of the particles. However, at the idle condition, no obvious sublimation was observed over the range investigated for this laser fluence sweep experiment. The peak particle temperatures reported by the LII 300, shown in Fig. 5a, permit further investigation into the difference in response to the nvPM produced at the idle and HPO conditions. The particles reach a lower peak temperature

The temperature is determined by two-colour pyrometry and therefore influenced by the nvPM optical properties, i.e. the relative value of E(m) between the two detection wavelengths. In this study, it is assumed that the value of the absorption function, E(m), is the same at both wavelengths (Snelling et al., 2005). The different particle properties at HPO and idle may invalidate this assumption, but the potential effect would only account for <100 K (Snelling et al., 2004) of the observed difference in the peak temperatures.

The impact of the laser fluence on the mass concentration measurements shown in Fig. 4 were also evaluated across multiple LII 300 instruments (LII 1 to LII 4) using the same nvPM source from Rig A to better understand instrument-to-instrument differences and exclude potential single-instrument anomalies in the observed trends with laser fluence levels. The results obtained from the HPO condition are plotted in Fig. 6a. The trends of mass concentration measurements during the laser fluence sweep analysis across the four different LII 300 instruments were similar, with minor differences observed between the individual LII 300 instruments likely due to modest variations in the beam profiles from the lasers, the specifications of the optics, and the resulting shape and size of the laser sheet in the probe volume location. All the instruments demonstrated the existence of a plateau region, identified as the range of fluence levels where the mass concentration was independent of the fluence level applied. The fluence levels from each instrument were estimated from the energy meter in the instruments and laser beam cross-section area in the probe volume, obtained from a beam profile camera, as mentioned previously. It should be noted that while the internal meter in the LII 300 used for determining fluence is not a precision instrument, it does provide representative and consistent fluence values. However, without calibrated energy meters, the fluences reported for each instrument may not be directly compared to the other instruments. To aid in interpretation of the results, fluence shifts were applied for different operating conditions (HPO, Idle in Fig. 5b) and different instruments (LIIs 1 to 4 in Fig. 6b) and can be used for different rigs (Rigs A, C–F in Fig. 11b) to align the fluence sweep data. The data from LII 3 and LII 4 were shifted with an offset laser fluence value to fit with the data from LII 1 in Fig. 6b. The data from LII 2 did not require shifting as the fit with the data from LII 1 was good. The normalised and shifted mass concentration profiles from the four different LII 300 instruments were found to repeatedly demonstrate similar behaviour. While interpretation of the behaviour in the low fluence region (fluence <1.8 mJ mm-2 for Rig A) remains a topic to be explored, it is recommended that measurements be acquired in the plateau region (1.8< fluence <2.7 mJ mm-2 for Rig A; Fig. 6b) and that extra care be taken in interpreting the results from the low fluence and sublimation regions.

During the laser fluence sweep tests, the laser fluence levels were adjusted by changing the Q-switch delay settings. The correlations in between the Q-switch delay settings and the laser fluence levels from the multiple LII instruments are different due to modest variations in the characteristics of each laser used in the LII 300 instruments. The correlation maps obtained during the laser fluence sweep tests at the HPO condition of Rig A are shown in Fig. 7. Each point indicates an average fluence over a period of 15 s duration. Raw data are shown here with no normalisation or adjustments included. A fifth-order polynomial function was fit to the data from each instrument and shown as a solid line in Fig. 7. The fit was essentially linear except at the lowest Q-switch delay settings, where there was evidence of saturation observed for the four different LII 300 instruments. LII 1 was also used for measurements from Rigs C to F. For the test on Rig B (IP rig) the LII 1 was not available, and LII 2 was used instead. LII 1 and LII 2 have a very similar relationship between laser fluence and normalised mass concentration, as shown in Fig. 6.

Figure 8a shows the particle size distributions in electrical mobility diameter (dm) measured using the SMPS in this study from multiple sources. For spherical particles, dm is equivalent to the volume equivalent diameter dve, whereas for aggregate particles, dm is larger than dve (Decarlo et al., 2004). In Fig. 8b and c, example images of the typical soot morphology collected in a previous study (Saffaripour et al., 2017) from a turboshaft engine exhaust are also shown for reference. The TEM images from two conditions are shown here: condition E1-1 (speed of 13 000 rpm, load 70 shaft horse power (shp), and the estimated global equivalence ratio φ of 0.25) and condition E1-2 (speed of 21 000 rpm, load 630 shp, and φ of 0.18). The image shows the primary particles and the fractal-shaped soot aggregate analysed via transmission electron microscope (TEM). In general, the primary particle diameters were small (17.2 nm (condition E1-1) and 20.8 nm (condition E1-2)), as were the aggregates, and they were non-volatile, with no evidence of significant semi-volatile coatings (clear boundaries) (Saffaripour et al., 2017). The aggregate sizes that were determined by projected equivalent-area diameter from TEM images were 33.5 nm (condition E1-1) and 45.7 nm (condition E1-2) (Saffaripour et al., 2017). The geometric mean diameter determined from the particle size distributions from Rig A is in close agreement with the prior results of aggregate diameter from TEM image analysis by Saffaripour et al. (2017).