Characterization of renewable diesel particulate matter gathered from non-premixed and partially premixed flame burners and from a diesel engine

Abstract

Particulate matter coming from the combustion of renewable diesel (RD), ultra-low sulfur diesel (ULSD) and a volumetric blend of 30% of RD with ULSD (RD30) were collected and physico-chemically characterized. Soot samples were generated in two flame burner types (non-premixed flame, NPF, and partially premixed flame, PPF) trying to simulate the diffusion and premix regimes found in diesel engines. The impact of both fuel nature and burner type was assessed on soot mass, particle size and morphology, particle nanostructure and surface functional groups. In general, although the results of HRTEM and SMPS suggested that the addition of RD reduced the average particle size and increased the concentration of ultra-fine particles, the mass emission of soot was drastically mitigated regardless of the burner used. The results also suggest that the changes in the chemical characteristics of the soot were slightly more sensitive than the changes in the internal nanostructure of the particles, since the graphitic character (as showed by Raman and infrared analysis) increased as the RD content increased, being stronger for the PPF system. Comparisons between engine soot and flame soot confirmed that the addition of RD into ULSD produced smaller and more carbonized particles. In fact, some engine results were located in between those obtained in PPF and NPF burners, suggesting that both combustion regimes are contributing to soot characteristics in engines. This consistency suggests that a first assessment of the impact of alternative fuels on the characteristics of particulate matter can be conducted through the basic approach offered by laboratory flames, thereby avoiding the costs associated with generating large quantities of fuel and the complexities of in-cylinder physical interactions and engine parameters.

1. Introduction

Approximately 20% of the world primary energy is consumed by the transportation sector, and this energy is almost entirely provided by fossil fuels (85%) [1]. It is well known that their use in vehicles produces large quantities of pollutants with corresponding negative impacts on the environment and health. Particularly, diesel particulate matter (PM) not only alters the air quality in urban centers (visibility impairment and smog), but also increases the risk of acute and chronic respiratory diseases. In fact, diesel PM has been classified by the International Agency for Research on Cancer (IARC) as “carcinogenic to humans” (group 1) since June 2012 [2]. Carcinogenicity of the diesel PM has been related not only to strong genotoxicity of the adsorbed polyaromatic hydrocarbons (PAH) on the particles surface but also to mutagenic effects showed by extracts of PM [3, 4]. In consequence, government agencies have introduced strict regulations regarding to the emission of PM into the environment, and the scientific community has proposed several alternatives to address those regulatory requirements. Advanced concepts in diesel combustion, new diesel engine technologies, exhaust aftertreatment devices and alternative fuels are commonly proposed strategies to meet the new emission requirements. With regard to the study of non-conventional fuels in diesel engines, oxygenated fuels like alcohols and biodiesel, have proven to be successful in reducing PM mass emissions [5, 6]. Nevertheless, they have some technical issues (blending limitations, due to fuel stability and changes in diesel fuel properties) and emission concerns (some biodiesel fuels increase NOx and alcohols increase unregulated emissions like aldehydes) that need to be assessed.

Recently, synthetic paraffinic fuels such Fischer-Tropsch and renewable diesel -RD- fuels (obtained by hydrotreating vegetables oils or animal fats) have gained worldwide interest due to their good engine performance and low emissions characteristics [7, 8]. RD is a high-quality bio-based fuel that consists of a mixture of liquid paraffinic hydrocarbons (C₁₅-C₁₈) that does not contain sulphur or aromatics. Although RD is considered a premium fuel due to its high cetane number (CN > 80) compared to conventional diesel (CN ≈ 53) [9], it should be blended up to 50% with diesel fuel to meet the requirements of lubricity and cold-flow properties [10]. Several authors have stated that, compared to diesel, RD not only promotes better engine efficiency [11–14] but also reduces regulated emissions (CO, THC, NOx, and PM) [15–24] and non-regulated pollutants such PAH and aldehydes [25–29]. Some researchers have also tested RD to gain an understanding not only of PM behavior in diesel particle filters [30–32] but also of the physicochemical feature of the particles [24, 31–35].

Although RD demands high pressure and temperature reactions with hydrogen in the presence of sophisticated catalysts, the infrastructure is commonly available in conventional refineries [15]. The RD exhibits similar fuel properties than high quality diesel fuel, avoiding the major concerns related to conventional biodiesel such as oxidation stability, the tendency to block fuel filters and the capacity to absorb water [26, 36]. Since RD is an ash-free fuel, the life time of aftertreatment devices can be enlarged in comparison with conventional biodiesel fuel [18]. On the other hand, since RD is oxygen-free, the carbonyl emissions (formaldehyde and acetaldehyde) could be potentially reduced in comparison with biodiesel fuel [37].

Laboratory flames are a valuable tool for establishing the impact of fuel chemical composition on soot properties that removes the thermo-fluid interactions occurring inside the engine cylinder. There are several studies that use surrogates in flame burners, but few deal with the evaluation of the morphology and nanostructure characteristics of the soot particles produced by commercial diesel fuels. For instance, Daly and Horn [38] compared the structure of toluene soot and its reactivity to ozone with soot produced from diesel and kerosene. Toluene and kerosene particles were produced in a co-flow diffusion flame burner, while diesel soot were gathered from a pressurized diffusion flame. Results showed that, although toluene and kerosene samples were generated under the same combustion conditions, they differed in reactivity due to marked differences in both the nature of organic carbon and the structure of elemental carbon. In contrast, they observed that the chemical composition before and after combustion was similar for diesel and toluene soot samples, despite both soots being produced in different burners. Kerosene soot was the least reactive due to its much more ordered structure (higher presence of aromatic species and higher C/H ratio) and diesel soot was the most reactive because of its amorphous structure (higher aliphatic nature and lower C/H ratio).

Merchan-Merchan et al. [39] used a wick-fed diffusion flame to study the structure and the size of soot particles generated from diesel and biodiesel fuels. Particle nanostructure analysis revealed that biodiesel-derived soot particles possessed a highly ordered onion-like structure compared to diesel soot, which presented less graphitic structure consisting of concentrically oriented graphene segments with non-crystalized amorphous material. Their results also showed that biodiesel particles were smaller than those produced by diesel due to higher temperatures of biodiesel flames (higher oxidative environment). In a recent work using vaporized co-flow diffusion flames, Merchan-Merchan et al. [40] confirmed that the diameter of the particles increased as the diesel fraction in the blend increased. Moreover, when the biodiesel fraction in blended fuel was increased, the chain-like structures became more complex, i.e., long-branched aggregates composed of a high number of primary particles with higher degree of networks in contrast to the short-branched agglomerates formed in the diesel flame.

Witkowski et al. [41] studied the soot volume fraction and morphology of particles sampled in a laminar co-flow methane-air diffusion flame seeded with diesel and various surrogate fuels. They found very similar soot volume fractions when toluene was blended with n-tetradecane in a similar concentration as that of the aromatic content of diesel (30 %vol.). However, the primary particle size and radius of gyration were larger for diesel than for surrogate fuel, indicating a lower surface growth rate for the surrogate relative to the real fuel. Barrientos et al. [42] studied the effect of fuel-bound oxygen in fatty acid esters on soot oxidation behavior. They burned methyl esters, alkanes, biodiesel and diesel in a co-flow laminar diffusion flame burner. Thermogravimetric analysis showed that the oxidation behavior depended on the length of the alkyl chains, i.e., soot generated from methyl esters with longer carbon chains (biodiesel-like fuels) exhibited lower reactivity compared to those samples derived from shorter alkyl chains. They also performed Raman spectroscopy on soot samples from methyl esters and n-dodecane confirming that lower soot structural order improved the soot oxidation (methyl esters > n-dodecane). They also stated that particles generated from conventional diesel revealed higher reactivity than the soot produced by a high-cetane and low aromatic diesel fuel.

Paul et al. [43] performed a comparative study of the soot precursor nanoparticles (particles in the nucleation mode with size below 10 nm) from both the exhaust of gasoline engine (running at no load and 3000 min⁻¹) and the non-sooting zone of partially premixed gasoline flame. TEM images and UV spectra from both sources confirmed the presence of particles with size below 10 nm and also showed that these nanoparticles are not aggregated. IR showed the presence of C-H aliphatic and oxygenated functional groups in the spectra of the two sources, but aromatic functionalities are only present in engine samples. Huang et al. [44] collected soot from jet engines fuelled with Jet-A and compared qualitatively against the soot generated in a normal diffusion ethylene flame (considered as “model soot” for atmospheric and climate studies). TEM images showed that flame aggregates have open and branched structure, in contrast to the compact morphology observed in the jet samples. In addition, jet primary particles did not display the typical core-shell structure present in the flame soot, instead a divergent nanostructure was observed (authors stated that this is a consequence of a highly turbulent conditions in liquid-fueled combustors).

According to this literature review and as far as the authors are aware, there are very few studies in flames that evaluate the effect of paraffinic fuels on the physicochemical properties of PM, and none of them compare PM gathered in flame burners with that obtained from a diesel engine. In the present work, two types of flame burners were used to generate soot samples that were characterized through Raman spectroscopy, high-resolution transmission electron microscopy (HRTEM), Fourier-transform infrared spectroscopy (FTIR) and scanning mobility particle sizer (SMPS). Neat RD and a blend with diesel fuel were used to evaluate the effect of paraffins on soot particle emission behavior and physicochemical nanostructure features. The findings will help increase the understanding of the effect fuel composition on soot emissions and will also provide useful information to understand the implications on the use of renewable resources in the transportation field.

2. Methodology

2.1. Fuels

Ultra-low-sulfur diesel (ULSD, supplied by the Colombian petroleum company) was used as reference fuel for comparison with neat renewable diesel (RD, home produced according to ref. [45]) and its volumetric blend at 30% (RD30, which accounted for around the half of paraffin content in the fuel blend). RD was produced by hydrotreating 2.7 L (2.5 kg) of palm oil in a 5.5 L Parr high-pressure reactor using 250 g of (NiMo)S_x/Al₂O₃ catalyst. The system was heated to 350 °C, pressurized with H₂ at 50 bar and mechanically stirred at 350 rpm for 4 h. Table 1 shows the properties of the different fuels tested in this study. The addition of RD into ULSD not only enhanced the autoignition quality of commercial diesel (increased the cetane number) by increasing its paraffinic fraction, but also reduced the boiling range, the final boiling temperature and the sulfur content. The low sulfur and low aromatic content obtained by blending the diesel with RD could help to reduce particle nucleation and soot tendency [46].

Table 1.

Properties of the ULSD, RD30 and RD

Properties	Method	ULSD	RD30	RD
Density at 15 °C [kg/m3l	ASTM D4052–11	861.00	837.28	780.90
Kinematic viscosity at 40 °C [mm2/s] ASTM D445–12		4.356	3.792	3.086
Lower heating value [MJ/kg]	ASTM D240–09	42.43	42.82	43.80
Aromatics [% v/v]	-	31.50	22.05	0.00
Naphthenes [% v/v]	-	35.95	25.17	0.00
Paraffins [% v/v]	-	32.55	52.78	100.00
Sulphur content [mg/kg]	ASTM D2622–16	12	< 5	< 5
C [% w/w]	ASTM D5291–16	86.91	86.34	84.86
H [% w/w]	ASTM D5291–16	13.09	13.66	15.14
Mean chemical formula (calculated)	-	C15.06H26.97	C15.44H29.05	C16.53H35.06
Boiling range [°C]	ASTM D86–16a	187 – 389	196 – 382	255 – 327
Derived cetane number	ASTM D7668–14	51.36	67.26	90.94

2.2. Flame burners

In compression ignition engines, fuel atomization and vaporization produce fuel-rich regions that mix with air (coexisting with evaporation) leading to partial premixing [47]. Soot particles are produced in both fuel-rich partially premixed flames and in diffusion flames (formed around droplets of fuel) [48]. Based on that, this study selected these two types of flames in order to emulate the sources of soot generation in typical diesel combustion. For the diffusion non-premixed flame (NPF), a conventional wick-fed burner was used, while for the partially premixed flame (PPF) configuration, the technique reported by Love et al. [49] was used. They eliminated engine variables like turbulence, injection timing and fuel atomization and vaporization using a PPF of prevaporized pure diesel-like fuels. The radiant heat fraction, the emission indices of NOx and CO and the soot volume fraction of both petroleum-derived and biofuels agreed with those obtained from engine studies [49]. Although these authors also provided useful information for understanding the dynamics of combustion process of modified fuels, including in-flame profiles of radicals and gas concentration [50], they did not present any information of the impact of fuels on the soot physicochemical features.

A 12-cm length cotton wick (4 mm diameter) was housed by a stainless steel fitting placed onto a 50-mL cylindrical glass reservoir (Figure 1a) to produce the NPF. For all fuels, the wick length was adjusted to 5 mm since this was the minimum wick exposure to produce soot in the RD fuel. The fuel flow rate, calculated from the fuel mass loss in the burner and measured using a scale (Shimadzu AUX220), was about 1.27 mg/s for all fuels. In this case, a constant fuel flow rate allowed comparing at relatively similar carbon flow rate with a difference of 2.7% by weight between ULSD and RD fuels. The flame temperature was measured at different flame heights using an R-type thermocouple (Pt 13% Rh/Pt, with wire and bead diameter values of 75 μm and 150 μm, respectively) which was placed into the flame by a rapid insertion technique in order to reduce the thermocouple exposure time in sooting flame regions. A Labjack U12 data acquisition system connected to an EI-1040 amplifier were used to record the temperature. Radiation corrections due to heat losses were performed for the thermocouple readings [51]. The standard deviation of the replicas was not larger than 10 K and the absolute uncertainty of the measurements is estimated to be ±60 K [52].

Figure 1.

Non-premixed flame burner (a) and partially premixed flame burner (b).

Figure 1b shows the setup for the partially premixed flames consisting of a 2.1m stainless steel tube (12.7 mm OD) wrapped with flexible electric heating tapes used to heat the air up to 400 °C. This temperature was selected because it was over the final boiling point of any of the liquid fuels tested (see Table 1). A 50-cm³ syringe pump injected the liquid fuel into the high-temperature air stream. The resulting air/fuel vapor mixture flows through a 20-cm stainless steel tube (9.5 mm ID) with a beveled rim. The vertical burner was also heated but at a lower temperature (~200 °C) to avoid fuel auto-ignition. The fuel mixture was ignited at the exit of the burner with an external pilot flame which was removed afterwards. In order to validate that this burner was able to fully vaporize the fuel, gas chromatography of both raw and vaporized RD was carried out. GC profiles were obtained using an Agilent Technologies 7820A GC with FID detector and a DB-1MS (60 m x 0.25 mm x 0.25 um) column. The two GC spectra (see Figure S2 in supplementary material) showed maximum differences of 4.5% in the intensities of the corresponding peaks, indicating that the components in the fuel were completely vaporized. Before performing a new experiment, the lack of formation of coke deposits was verified in both the burner and the fuel injector. In all the cases evaluated here, no evidence of coke formation was seen. Test conditions for this burner are shown in Table 2, and these conditions were selected after performing a stability map, which will be discussed in the results section. The fuel and air flow rates used to achieve the same equivalence ratio. The difference in carbon flow rate between ULSD and RD for PPF system was calculated to be below 5% by weight.

Table 2.

Experimental conditions for partially premixed flames.

Fuel	Stoichiometric air/fuel ratioa	Air flow rate [L/min]	Fuel flow rate [mL/min]	Equivalence	C/O ratio
ULSD	14.442	8.5	1.154	1.4	0.485
RD30	14.574	8.5	1.175	1.4	0.483
RD	14.908	8.5	1.232	1.4	0.457

^aCalculated from elemental analysis.

2.3. Sampling procedure and characterization techniques

Samples generated in the two flame burners were collected at the tip of the flames using glass fiber filters (Advantec GC-50, 0.5 μm pore) in line with vacuum system coupled to a 15-cm-length water-cooled stainless steel probe (3 mm ID). The probe has a steel housing that enables support of either filters or TEM grids. The sampling process was carried out without disturbing the flame and keeping constant the sampling rate and time (90 seconds).

2.3.1. Raman spectroscopy

Raman spectra were obtained using a LabRam HR Horiba microscope system equipped with a 632.8 nm He/Ne laser excitation source of 17 mW. For each sample, three different spots were analyzed in a spectral range of 800–2000 cm⁻¹ using a magnification objective of 50x. A source power of 0.17 mW and an exposure time of 20 s were used to avoid burnoff of the sample [53]. Since Raman shifts did not vary significantly when fuels were changed, the location of the fitting functions were fixed at their average positions according to [54, 55]: 1160 cm⁻¹ (D4, Lorentzian), 1340 cm⁻¹ (D1, Lorentzian), 1545 cm⁻¹ (D3, Gaussian) and 1605 cm⁻¹ (G, Lorentzian). D2 band was not identified in any spectrum. To describe the soot nanostructural features, the area of each curve were compared to G band area. The standard deviation of the Raman parameters evaluated in this study associated to graphitic and amorphous content (A_D1/A_G and A_D3/A_G) was about 9.6% of the mean.

2.3.2. High Resolution Transmission Electron Microscopy (HRTEM)

Flame particles were collected directly on the TEM grid placed in the housing coupled to the probe. The exposure time of the probe was 0.5 s. Before TEM analysis, soot deposition on grids was checked by means of an optical microscope (Nikon Labophot-2 with the objective 100x). A FEI Tecnai G2 F20 transmission electron microscope operated at 200 kV was used to obtain images at 29,000x and 43,000x for calculating the mean primary particle diameter (d_pp). To determine the d_pp, particles with distinguishable boundaries were randomly selected and were analyzed by using the image processing software ImageJ®. To obtain information about the length, tortuosity and separation of the fringes (i.e., interlayer distance d₀₀₂), HRTEM images taken at 590,000x of magnification (spatial resolution of 0.019 nm/pixel) were processed using the methodology presented in [56]. d_pp and fringe analysis were made using between 3 and 6 images per sample (see Table S1 in the supplementary material).

2.3.3. Fourier-transform infrared spectroscopy (FTIR)

Functional groups were identified using qualitative FTIR analysis. A small amount of raw collected PM (non-devolatilized) was used to prepare a 0.5 wt% KBr pellet. Infrared signals were recorded using a Nicolet 6700 spectrometer with a MCT/A detector. Each spectrum was the result of a 32 scan accumulation, a value that provided the best signal/noise ratio. A blank spectrum was obtained from a KBr pellet prepared without a PM sample to ensure that infrared signals were only attributable to PM chemical compounds and not affected by the preparation process. Three replicates of each sample were taken to estimate repeatability of the method. In general, the uncertainty in the IR measurements was less than 5%. Each spectrum was normalized by the 2920 cm⁻¹ signal as proposed in references [57, 58]. This avoids making absolute peak height comparisons between different spectra, which can be affected by factors such as the thickness and sample concentration in the KBr pellet.

2.3.4. Scanning Mobility Particle Sizer (SMPS)

Particle size distributions (PSDs) of the PM generated were measured with a with a SMPS TSI 3080 with a Kr-85 bipolar charger), which consisted of a TSI 3080 electrostatic classifier with a nano differential mobility analyzer (nano-DMA, TSI 3085) and a TSI 3025 ultrafine condensation particle counter (UCPC). The particles were drawn through an open-end stainless steel probe (ID of 8 mm) using an eductor (#15, Fox Venturi Mini Eductors, motive air at 5 psig). The dilution ratio was 5:1 (see section S.4 in supplementary material). The sheath flow rate and aerosol flow rate were set at 3 L/min and 0.3 L/min, respectively. For each fuel and burner type, three consecutive SMPS scans were recorded, with the mean and standard deviation values used for reporting the final PSD. The SMPS data was analyzed using AIM (version, TSI) software, which allows the corrections for the particle multiple charges and diffusional losses. Particle mass distributions were calculated from the PSD and particle density with the methodology proposed by Gomez et al. [59]. When the NPF particles were analyzed using the SMPS system, an incomplete PSD was obtained, indicating that the NPF particles are bigger than the upper detection limit of nano-DMA. Therefore, only the PPF burner SMPS data is reported.

3. Results and discussion

3.1. Flame descriptions

All flame images depicted in Figure 2 were taken using the same digital camera (Samsung SM-G950F of 12 megapixels) and using the same camera settings (ISO-120, aperture = f/1.9 and exposure time = 1/60 s). The qualitative visible flame high was determined in a light-reduced room using a millimetre scale ruler. Under the conditions tested, flames have an average height of 13 cm and 4 cm for PPF and NPF, respectively. While the typical yellow region was identified in NPF, two regions were observed in PPF: a bright blue cone (which represents the premixed reaction zone) and a surrounding region consisting of unburned species that continued to burn with ambient air (diffusion zone). This second region had a blue-violet part (soot-free) and a luminous yellow zone (soot). The yellow luminosity was higher for ULSD flames compared to RD flames due to their augmented soot radiation. When RD is added to ULSD, the blue-violet zone increased in PPF and the soot trail became narrower in both the PPF and NPF (see red circles in Figure 2), indicating lower soot formation. In fact, in the NPF configuration the flame tip went from open to closed as the aliphatic content in the fuel increases. It is widely accepted that if the parent fuel has aromatic compounds in its formulation, the soot emission will be higher than that coming from more paraffinic fuels [60]. As expected, Table 3 shows the decrease in soot emission as the RD fraction increased in the mixture. Increasing the paraffinic fraction from 33% (ULSD) to 53% (RD30) resulted in a reduction of soot mass emission of 53% and 28% in PPF and NPF systems, respectively. When neat RD was used in PPF and NPF, these reductions reached an 83% and 70%, respectively.

Figure 2.

PPF of ULSD (a), RD30 (b), RD (c); and NPF of ULSD (d), RD30 (e) and RD (f). Dotted circle indicates the soot trail. Camera parameters f/1.9, 1/60 s and ISO-120.

Table 3.

Specific particulate matter emissions in the flame burners

Fuel	PPF			NPF
	Specific PMa [x10−5 mgpM/gf]	SD	Variation [%]	Specific PMb [mgPM/gf]	SD	Variation [%]
ULSD	47.9	8.0	-	152	8	-
RD30	22.8	2.6	−52.4	109	7	−28.7
RD	8.5	1.4	−82.2	45	4	−70.6

^aDerived from the mass distributions (Figure 5b).

^bMass collected in filters.

SD: standard deviation

mg_PM: milligrams of particulate matter; g_f: grams of fuel

Figure 3 shows the temperature profiles obtained at different flame heights in NPF up to 40 mm. The temperature in the PPF system could not be measured because the temperatures reached were by far exceeding the working temperatures of the thermocouples used in this study. The temperature profile along the flame centerline was similar for the three fuels tested, increasing from near the flame base up to 15 mm of flame height. Further downstream, the differences in sooting tendencies reduce the local temperatures due to soot radiation. Similar results were observed by Smooke et al. [61] and Wang et al. [62]. In our case, the effect of radiation was most significant for the USLD fuel because it produces much higher soot levels than RD30 and RD, resulting in a lower flame temperature [63, 64]. Higher temperatures for paraffinic flames compared to flames produced by fuels with aromatics have been reported previously. Velasquez et al. [65] using an inverse diffusion flame burner, showed that independent of the HAB, hexane flame temperature was on average 300 K higher than that of a diesel surrogate flame (which contained about 20 %vol. of aromatic compounds). Botero et al. [66] have also reported that an n-heptane non-premixed flame exhibited higher flame-tip temperature compared to toluene and n-heptane/toluene blends.

Figure 3.

Flame temperature measured at the centerline of NPF using different fuels.

3.2. Primary particle size and particle size distribution

Figure 4 displays the d_pp estimated from TEM images for PPF and NPF soot samples. Irrespective of the burner type, the addition of RD into ULSD led to a decrease in d_pp. In the PPF system, the difference in d_pp was about 3 nm when neat RD fuel was used, while in the NPF burner, the differences in d_pp measurements were even larger and near 17 nm. The reduction in the primary particles size could be associated to i),since comparative contribution of different types of growth reactions seems to depend on the fuel characteristics, under similar residence time, paraffinic fuels must follow a much more complex and slower path to produce the first aromatic ring compared to aromatic fuels [67] hindering particle growth, and ii) the high temperature of RD flames cause a highly oxidizing environment that shrinks the particles. Botero et al. [68] also found that in their diffusion flame burner, the d_pp of n-heptane soot is smaller than that produced by the n-heptane/toluene blend and toluene fuel. The PPF-generated primary particles were 1.5 to 2 times smaller compared to the NPF-generated primary particles when the same fuel was burned. This result might be related to the shorter residence time in PPF that impedes the surface growth (in contrast with the extended combustion duration in diffusion flames that favors particle growth).

Figure 4.

Mean primary particle diameter (d_pp) of flame samples.

Figure 5a shows the particle number size distribution of the samples gathered in the PPF. The PSDs show that the addition of RD into ULSD reduces the mean mobility diameter of the emitted particles. It seems that the overall lower emission level of RD leads to less chance for agglomeration. The PSDs also shows that the number of smaller particles is higher for RD flames (Figure 5a). The PSD results of Botero and co-workers demonstrated that paraffinic fuels produced smaller particles than aromatic fuels [66] and cycloalkanes [69] (these compounds represent about two thirds of diesel fuel composition). Mauss et al. [70], using a PPF burner (counterflow configuration) and three different C/O ratios (0.85, 0.75 and 0.70), showed that flames with lower C/O ratio produced smaller particles, which is also in agreement with the results presented here.

Figure 5.

Particle number (a) and particle mass (b) size distributions for the PPF.

It is important to mention that this high concentration of smaller particles is compensated by a lower mass emission for RD (Figure 5b). Ng and collaborators [71], using a single-cylinder diesel engine, reported that Sun Diesel (a paraffinic fuel produced through biomass-to-liquid technology) emitted much more ultra-fine particles (< 70 nm) compared to diesel fuel. Soriano et al. [72] showed that, in comparison with diesel fuel, the combustion of GTL and farnesane (a paraffinic fuel derived from sugar cane) in a diesel engine, led to higher number concentration of particles smaller than 150 nm. In a recent study, conducted by Shukla et al [73], higher concentration of nucleation mode particle was evidenced for RD at 2000 bar of engine rail pressure, compared to diesel fuel. They explain that the higher cetane number of RD resulted in lower accumulation mode (this implies smaller surface area for adsorption of organic species that may form nucleation mode), which ultimately can increase the concentration of nucleation mode particle in the exhaust.

3.3. Soot nanostructure

Representative Raman spectrum of the flame samples are plotted in Figure 6. This figure illustrates the clear influence of both fuel composition (Figure 6a) and burner type (Figure 6b) on Raman spectra. For a better comparison, Figure 7 depicts the area ratios of Raman bands related to structural defects, i.e., A_D1/A_G, and A_D3/A_G. The presence of D1 and D3 bands implies, respectively, disordered graphitic structures and amorphous carbon in soot [54]. Regardless the flame burner, the decreasing trend in all the mean area ratios (except for A_D4/A_G in PPF) indicates that the addition of RD into ULSD produces soot particles with a more ordered structure. Additionally, when the same fuel was burned using different flame configurations, a clear reduction in every area ratio indicates that the soot samples gathered from PPF presented more graphite-like structure (reduced A_D1/A_G) and less amorphous carbon content (decrease in A_D3/A_G), thus, a higher internal degree of order compared to NPF. It is likely that both findings are a consequence of the high-temperature environments (based on the conceptual model of Marsh and Griffiths [74]). The RD’s higher flame temperature and lower sooting propensity compared to those of the ULSD flame are related to soot-associated heat loss mechanisms. Similarly, compared to the NPF system, higher flame temperatures are expected in PPF burner not only because of its less-sooting nature, but also due to relatively more complete combustion caused by the supplied air. No changes were found for A_D4/A_G, which implies no variations in ionic impurities.

Figure 6.

Raman spectra: effect of fuel (a) and effect of flame type (b).

Figure 7.

Raman peak area ratios of the flame samples.

3.4. Soot morphology

TEM images of the soot agglomerates from all flame samples are shown in Figure 8. Regardless the fuel and burner configuration, PM exhibited typical sphere chain-like structures but with a complex network of branches. In contrast to the NPF-derived agglomerates, the branches of PPF samples are composed of a high number of primary particles. This led to a high particle shakiness (due to cantilevering of the agglomerates in the grid support) that impeded HRTEM imaging of the primary particles for fringe analysis. Representative HRTEM images of the NPF samples are displayed in Figure 9. All these primary particles exhibited the typical shell/core structure with graphene layers oriented parallel to the outer surface. Qualitative evaluation of these micrographs shows that compared to RD30 and RD particles, ULSD particles have more arranged fringes in the shell region (see red arrows in Figure 9), indicating higher surface growth.

Figure 8.

TEM images of aggregates gathered in PPF and NPF burners. PPF: ULSD (a), RD30 (b) and RD (c). NPF: ULSD (d), RD30 (e) and RD (f).

Figure 9.

HRTEM images of NPF particles: ULSD (a), RD30 (b) and RD (c). Red arrows indicates the extension of shell region.

Figure 10 displays the graphs (with box plots and lines) obtained from fringe analysis of particles collected in the NPF burner. The box plots show the medians, standard deviations (whiskers), 25% and 75% percentiles (upper and lower limit of the box) and means (dots) of the interlayer distance, fringe length and tortuosity (curvature due to defects in the graphene layers). The lines depict the relative frequency of the calculated values. An increase in the paraffinic content of the fuel caused a slight decrease in the interlayer distance, which is associated with the more ordered structure of particles (an increase in pairs of fringes with less than 0.38 nm of separation can be observed in Figure 10a). Compared to ULSD soot, the RD produce particles with shorter fringes (Figure 10b) and, in addition, a reduction in the concentration of fringes with tortuosity around 1.10 but an increase in the frequency of fringes with curvature between 1.15 and 1.25 (Figure 10c). This is in agreement with ref. [68] which argued that high flame temperature produce a highly oxidizing surrounding that reduces the fringe length and increases the tortuosity. The carbon lamellae is bent during oxidation, not only due to disorder introduced by oxygen absorption, but also because of the formation of intermediate 5-member ring species which results in local curvature [75].

Figure 10.

Fringe analysis of particles collected in NPF burner. Interlayer distance (a), fringe length (b) and fringe tortuosity (c). Histograms from the HRTEM analysis were converted into curves by connecting the top of the bars with a cubic spline line.

3.5. Functional groups

Figure 11 displays the infrared spectra for all PM samples. The signals corresponding to the C-H stretching mode associated to CH₂ and CH₃ aliphatic groups in the range of 2850–2950 cm⁻¹ are commonly observed in soot from fuels containing aliphatics in their composition. Since the shape and intensity of these signals remained similar for all soot samples, they were used as reference to normalize the spectra. It can be seen that, although the NPF spectra shows more intense signals compared to PPF spectra (possibly due to more organic material in the NPF samples), an increase of the intensities of the functional groups related to CH₃ plane deformations (1380 cm⁻¹) of aliphatic groups and C=C stretching vibration (1640 cm⁻¹) is observed when RD was added into ULSD. The fundamental pathway for the decomposition of large paraffins proposed by Zhang et al. [67, 76] involves either hydrogen abstraction or thermal decomposition to form alkyl radicals, each of which decomposes via β scission to form alkyl radicals and olefin species. Based on this mechanism, it is likely that the combustion of RD (composed of higher paraffins, C₁₅-C₁₈) yielded CH₃ radicals and olefinic C=C compounds that were subsequently condensed onto the particle surface. On the other hand, the addition of RD raised the flame temperature favoring the reorganization of carbonaceous particle structure (the soot structure became more ordered) [77]. No appreciable changes were found neither for carbonyl functionality at 1735 cm⁻¹ nor for the region of 1000–1300 cm⁻¹ associated to C-O stretching modes of oxygenated compounds.

Figure 11.

Infrared spectra of soot from PPF (a) and NPF (b).

3.6. Comparison between flame soot and engine soot

The formation of soot is a rather complex phenomenon that depends on a large variety of combustion parameters, including fuel chemical composition, residence time, burning configuration, flame temperature and pressure. Although soot production in practical combustion systems mainly occurs at high pressure, most of the previous studies on soot formation have been primarily focused on flames (premixed and diffusion flames) operated at normal conditions. Studies conducted on flames at atmospheric pressure and constant fuel composition suggest that although the chemistry involved in the formation of soot does not change regardless of the flame type (diffusion or premixed), differences in configuration impose certain restrictions on the overall process. For instance, the very rich pyrolysis zone where soot growth occurs in diffusion flames cannot be replicated with premixed flames since soot precursor’s formation competes with soot precursor’s oxidation. In diffusion flames, the pyrolysis region is much wider (longer residence time) and oxidative attack on soot precursors and soot may occur subsequent to their formation towards the end of the flame (flame front) causing an increase in both, particle size (surface growth) and soot mass emission compared to premixed flame [78], as has been already pointed out in our study.

In high pressure combustion environments, the soot formation process is affected by changes in the diffusion and reaction rates. As pressure increases, the probability of a collision among the gaseous molecules increases leading to higher reaction rates, but diffusion rate remain almost constant due to the compensation of increased densities of reactants: reactants that diffuse into the flame region are consumed at a higher rate, and as a result, the flame reaction zone gets smaller and the flame shape narrower. Because of that, the soot volume fraction, temperature gradients and residence time increase with pressure [78]. Compared to soot concentration, the knowledge of pressure effect on soot morphology is scarce and sometimes contradictory especially in engine studies. Studies about particle diameter over the engine cycle indicate that particle diameters varied slightly within 30 nm and 50 nm for all crank angles, i.e., pressures [79]. However, for different engine loadings the mean particle size and size distribution changed with crank angle [80, 81], making comparison between different studies difficult. Experiments performed in lab-scale burners suggest that particle size increases due to an increase in surface growth rate at elevated pressure [82], but particle nanostructure seems to be unaffected by pressure and only a slightly increase of graphitization of soot particles was seen due to an increase in resident time [83].

A previous study conducted in an automotive diesel engine [56] suggests that the addition of RD to ULSD promotes smaller primary particles with more ordered internal nanostructure. Some relationships can be drawn when comparing some characteristics identified for particles gathered in flame burners with those evaluated in the engine-derived samples. In this section, mean primary particle size (calculated with TEM images), soot nanostructure (using the fringe features and the Raman parameters), and the chemical functional groups (from the infrared analysis) are discussed, and only data related with neat RD and ULSD are presented. The “2410” label introduced in the following figures was used to describe the engine-related data extracted from [56], which is an engine operation mode at 2410 min⁻¹ and 95 Nm and corresponds to a high-temperature mode and high specific NOx emission point. Figure 12 shows the mean diameter of the primary particles obtained as a function of fuel composition and the combustion environment. According to this graph, the primary particle sizes tends to be smaller for RD compared to ULSD fuel independently of combustion configuration. For the paraffinic RD fuel, the aromatic precursor formation takes longer compared to the already existing in ULSD, and therefore, the particle growth could be affected. It is also interesting to note that the particles generated in diesel engine had diameters between those generated in the PPF and NPF. This could be associated with the hybrid nature of the diesel combustion system, consisting of an initial combustion phase controlled by a partially premixed mixture inside the cylinder and a later phase characterized by a diffusion-driven process. This finding also might support the fact that during diesel combustion, small soot particles are generated in the premixed zone (in the well-mixed section of air and fuel vapor) and they grow as they pass through the plume stem to the head vortex and/or outward the enveloping diffusion flame [48].

Figure 12.

Mean primary particle diameter.

Figure 13 displays the results of the fringe analysis performed on engine soot and NPF-derived soot. While fringe length distributions (Figure 13a) are similar for ULSD and RD, relatively more curved fringes (Figure 13b) and more pairs of fringes with a decreased separation (Figure 13c) were generated when both systems were fueled with neat RD. No effect was observed in the fringe features when the fuel was burned on different combustion system, in spite of the big difference in pressure and turbulence conditions at which combustion takes place. Similar results were reported by Jaramillo et al. [84], who studied the soot nanostructure of partially oxidized soot at 1, 10 and 40 atm. Their lattice fringe analysis algorithm showed that increasing pressure caused no apparent changes in soot nanostructure (fringe length or tortuosity) for the samples produced by non-oxygenated fuels (m-xylene, n-dodecane and m-xylene/n-dodecane blend) in a premixed flat flame burner.

Figure 13.

Fringe parameters of samples from different combustion system: fringe length (a), fringe tortuosity (b) and interlayer distance (c)

The Raman spectra in Figure 14 reveal a clear influence of the combustion system in the structural order of soot particles. Independent of the fuel, the bands related to plane defects in the Raman shift suffer an appreciable increase in their areas, and the G band becomes broader in the diffusion-controlled combustion system suggesting that the internal structure order of particle is reduced (green spectra). Area ratios extracted from Raman spectra through multiple peak decomposition process are plotted in Figure 15 as area ratios. The mean of A_D1/A_G and A_D3/A_G area ratios from engine samples are the lowest, indicating more graphite-like structure with less amorphous domains compared to those generated in flame systems. This is not surprising because the samples gathered at the engine exhaust were subjected not only to high temperatures but also to a high pressure inside the cylinder, increasing their degree of order. The opposite was found in the work conducted by Commodo et al. [83] using a coflow methane-air diffusion flame, who showed that increasing the pressure of the system from 10 to 20 bar did not affect the nanostructure of the gathered soot. Instead, their Raman results indicated that residence time (i.e. the flame height) appeared to be the governing factor of soot nanostructure. The discrepancy with the present work might be related to the fact that they did not compare their results with those at atmospheric pressure, which could exhibit significant changes in the soot nanostructure. NPF samples showed the higher degree of structural disorder (as indicated by their larger A_D1/A_G and A_D3/A_G values), regardless of the fuel tested. This result could be anticipated based on the temperature at which the particles are formed, that is expected to be the lowest among the tested combustion systems (due to incomplete combustion conditions and higher sooting nature of a wick-fed flame burner).

Figure 14.

Raman spectra of samples from different combustion systems: ULSD (a) and RD (b).

Figure 15.

Raman bands area ratios of samples from different combustion systems.

Table 4 shows the infrared signals of the samples collected in the three different combustion systems. Independent of the combustion system, the functional groups associated with C=C stretch (1640 cm⁻¹) and aliphatic CH₃ bend (1380 cm⁻¹) were more intense when RD was burned. The reasons were discussed in section 3.6. However, it can be seen that the 1380 cm⁻¹ signal was less intense for the PPF system due to lower quantities of alkyl radicals that can condense onto soot particles caused by the better fuel-air mixing, which improves the oxidation of the fuel in a premixed combustion. The relative intensity of oxygenated groups (1240 and 1735 cm⁻¹) was higher for the RD samples, and this occurs because aliphatic species are more likely to be oxidized with oxygen compared to aromatics. In addition, regardless the fuel tested, engine samples have more relative oxygenated groups, and this is related to the lean conditions inside the cylinder, where high concentrations of oxygen increase the possibility of producing much more oxidized species on the particle surface.

Table 4.

Relative intensities of infrared signals of samples from engine and flames.

Fuel	System	Wavenumber [cm−1]
		1240	1380	1640	1735
ULSD	2410	0.138	0.457	0.102	0.302
	PPF	0.090	0.245	0.295	0.153
	NPF	0.085	0.433	0.334	0.269
RD	2410	0.225	1.297	0.207	0.386
	PPF	0.112	0.270	0.339	0.159
	NPF	0.148	0.610	0.607	0.294

Note: Data are normalized to the intensity of the 2920 cm⁻¹ signal.

“2410” data were extracted from [56]

Comparing the soot obtained in the combustion systems tested, it can be note that the chemical characteristics associated with functional groups of the engine soot are similar to the chemical characteristics of the NPF soot. This makes sense, since the diffusion combustion regime in the engine favors the formation of PAH that along with unburned fuel confers the final chemical characteristics to the particles. On the other hand, although the nanostructure parameters of the particles did not show significant differences, the degree of graphitization or carbonization of the engine soot particles resembles what was observed in the premixed flame particles where temperature play an important role in annealing and removing amorphous material. Finally, the size of the engine soot particles is within an intermediate range between diffusion and premixed, so the stage of particle growth will depend besides pressure and temperature, on a combined effect between fuel chemistry and the flame configuration that ultimately establishes residence time for soot particles.

According to literature, faster soot oxidation can be related to smaller particle size (which implies higher specific surface area for oxygen attack) [55, 85], to more disordered nanostructure (which possesses more reactive sites) [86, 87], and to higher concentration of oxygenated surface groups [34, 88]. Based on this, it would be expected that, compared to ULSD particles, RD samples oxidize earlier due to decreased particle size and greater presence of oxygenated groups. Nonetheless, the ordered nanostructure evidenced in the RD particles would hinder their oxidation process. In fact, the aforementioned work conducted by the authors [56] showed that the oxidation behavior of the particles generated in the engine by both ULSD and RD were virtually the same.

4. Conclusions

The present work uses two flame configurations to evaluate the impact of RD on soot characteristics. Neat RD, ultra-low sulfur diesel, and a volumetric blend RD30 (30% by volume of RD in ULSD) were used to evaluate the influence of the paraffinic nature of RD on the internal structure, morphology and chemical composition of the resulting particulate matter. The results suggest that the “bulk nanostructure” of the soot (i.e., amorphous and graphitic domains) and the surface chemical composition were affected by both the fuel formulation and the combustion system. Fuel composition had a bigger effect than the flame combustion conditions on the PM properties. The results also support previous diesel engine study in spite of complex in-cylinder phenomena. Comparisons between engine soot and flame soot confirmed that the addition of RD into ULSD produced more graphitized particles. The high degree of order (determined by Raman parameters) was in agreement with the interlayer distance (performed on HRTEM micrographs) and with the intensity of C=C stretch in FTIR spectra. Particle size distributions measured in PPF showed a decrease in mean particle size with the addition of RD, but they also showed a corresponding increase in the number concentration of ultrafine particles. Although RD generated smaller and a greater number of particles, RD resulted in a reduced mass emission (over 70% when neat RD is used). In addition, although the magnitude of the results was modified by the physics of the combustion environments, the trends imposed by the fuel chemistry were not affected. The agreement between the results of both combustion environments (engine and flame burner) enables the possibility of employing laboratory flames to detect trends in the characteristics of emissions produced by non-conventional fuels, thereby avoiding engine-related issues like in-cylinder physical interactions and the requirements for large quantities of fuel. Thus, flame research can provide a valuable screening tool before fuel evaluation in engine facilities. The findings discussed in this work related to particle size, nanostructure and surface chemistry of flame-generated samples not only support the results of the previous engine study [56] but also help to explain why the soot oxidation of both ULSD and RD particles are practically the same.