16 Stg.Houtstook 24-0083684 (geanonimiseerd)

husks, wood chips, and olive pits. In addition to the system design itself, fuel type may also impact emissions, efficiency, and performance. Fuels that contain bark for instance are known to contain many contaminants as trees will push the “unwanted” matter to their leaves or bark. In summary Jensen found mineral compounds are up to 10 times more abundant in bark than in wood, with bark ash contents up to 600Xiwt, predominantly calcium, silica, and phosphorus [35]. Additionally, phenolic acids in bark are mostly found in outer bark cells. Low mo­ lecular weight extractives are often very abundant in bark such as al­ cohols, resin acids, alkaloids, and terpenes. There can also be small amounts of essential oils (volatile low molecular-weight terpenes) (e.g. incense cedar, ponderosa pine, and eucalypts, among many others) [36]. Moisture content of biomass fuels are also another key factor in the ignition time and devolatilization rate [37], as well as the overall combustion performance [38,39]. The elements that may be found in a solid biomass fuel, and the emission from their combustion, also depend on where the tree was grown. Specifically, trees living in areas with toxic air pollution or ground water contaminants may have toxic metals accumulated in the bark and leaves [40]. In the sections below various common fuel types are described in some detail. 4.1.

Cordwood Cordwood or as more commonly referred firewood, is un-treated, with or without bark and air-dried wood. Cordwood can vary in terms of species (fuel density) and pieces of the same species further may vary in moisture content, the amount of bark covering the surface, weight, and size. To complicate this fuel further, combustion of cordwood is subject to natural chimney draught, cannot be ‘turned on or off like automatic fuels can with instantaneous ignition, and is subject to var­ ious lighting techniques. To minimize this variability and increase best combustion practices, stove or boiler manufacturers typically recommend their systems be fired with either two main types of cordwood: deciduous or hardwoods and coniferous or evergreens or softwoods. Examples of hardwoods include oak (red or white), maple, or beech and softwoods include pine, spruce or Douglas Fir. Traditionally, round logs are cut into four to six pieces to improve the surface area and combustion performance. Manufacturers typically recommend seasoned (air-dried) cordwood for use in their devices with moisture ranges on a dry basis of 16-200Xi, however some units have claims to the ability to burn green wood (moisture levels above 350XO. In addition, piece size, amount, and loading procedures are listed in manufacturer's written instructions and device manuals.

There are currently no known cordwood fuel specifi­ cations to ensure consumer's fuel meet the recommendations of man­ ufacturers in either the US or Europe, however a technical committee in Europe is currently drafting standards for cordwood (CEN/TC 335) [41].

4.3. Wood chips Biomass fuels with particle sizes typically smaller than 5 cm x 5 cm x 5 cm are commonly classified as wood chips. Wood chips have 2-3 times lower bulk density than pellet fuel [39], they may contain bark and their moisture content is typically higher than that of pellets. Higher quality or lower moisture wood chips typically are chipped from seasoned cordwood by a fuel grade chipper with set screens to achieve a uniform desired size. As biomass tends to re-absorb moisture if left uncovered, proper care is necessary to maintain lower moisture content chips after chipping. In some cases solar heat (hot air for drying) has been employed to produce high quality [46-48]. Lower quality, higher moisture content chips may be chipped by a general agricultural chipper directly after removing a tree (allowing for no seasoning of logs and in some cases leaves, grass and dirt may also be included) which tends produce a shredded material, producing longer pieces of wood, twigs, dust, and a variety of non-uniform sizes. Wood chips produced in this type of condition would have a much lower heating value than those of uniform size, lower moisture, and free of dust.

Similar to cordwood there are no known specifications to ensure consumer's fuel meet the recommendations of manufacturers in either the US or Europe, however a technical committee in Europe is currently drafting standards for chips as well (CEN/TC 335) [41].

4.2. Pellets

4.4. Briquettes

Pellet fuel or more commonly referred to as just pellets, are solid biofuels made of compressed organic matter or biomass. Biomass feedstock for pellet production includes wood shavings, wood chips, grasses, straws, lumber mill scrap, and wastes from energy crops. Typically pellets use a sawdust type feedstock with little to no bark so the use of a hammer mill is common as a first step to treat the raw material. The raw material may be dried if exposed to outdoor weather elements or if the raw material was green. During the processing state high temperatures cause the lignin to break down and melt which will later solidify, holding the pellet together. In some pellet plants a small amount of cornstarch or flours may be added to act as additional “glue”. Finally, the processed material is pushed through a pellet die at high pressures where the pellet is cut to the desired length (typically on the order of 0.6-3.8cm [42]).

Biomass briquettes similar to pellets, use biomass sawdust or fiber type feedstock (straw, paper, agricultural waste, leaves, and so on), and use a densification process making them uniform in size and easy to store and transport but are pressed into a rectangular shape resembling a traditional red building brick or larger disk or solid or hollow log shape of 7-25 cm in length and 10 in 30 cm in diameter. If the fines are heated to a temperature which breaks down the biomass lignin under high pressure no binders for the briquettes are needed [49] but in some cases, if the fibrous material is not heated or processed under high pressure, a binder material is used such as molasses, starch, or even cow dung and cactus [50]. Depending on the type of raw material, the first steps of production are sieving, drying, or crushing, followed with preheating, densification, cooling, and packing the briquettes [49]. Studies show higher 519

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

Manual and automatic feed heaters (specifically boilers) rely on different mechanisms to meet the heat demand. Automatic feed boilers can modulate down to a lower burn if they sense they are approaching the heat limit and then go into a burn-out period to be restarted again by an electric ignition when there is a call for heat again. Manually fed boilers on the other hand, cannot restart automatically and so are forced to ‘slumber’ their fuel bed by shutting air dampers and slowing the rate of fuel combustion. In some advanced technologies the boilers have the ability to modulate down slowly and maintain a 300Xi of nominal output without going into a full ‘slumber’. Once there is a call for heat again, these boilers will open the air dampers and adjust the fans to achieve their full output once again. Gaegauf et al. [56] reported the particle size distribution of batch-wise fired appliances (wood stoves) varied significantly during a burn cycle, while wood log and continuous fed boilers showed a fairly constant particle size distribu­ tion. Beyond the difference in size distribution of the emitted particles, differences in the particle's properties exist which will be discussed in detail below. These differences are likely to influence the biological effects induced by wood smoke particles [10].

Studies have shown combustion conditions and the resulting parti­ cles strongly depend on fuel type, combustion technology, and type of operation [3,10,20,21,56-59]. One study in particular detailed the emissions of pellets in a pellet boiler and pellet stove, wood chips in a wood chip boiler, and two different cordwood species in a cordwood boiler and cordwood stove. The study found significant differences in terms of emissions factor (mg/MJ), even in the cases when spruce or beech cordwood was used in the same stove run at steady state; 67 and 93mgZMJ, respectively [60]. Additionally, the study found the period of combustion, cold start or steady state had a significant impact in terms of emission production, for some cases ten times greater emis­ sions output (mg/MJ) were observed. Fig. 5 shows emission factors for various type of wood-heating ap­ pliances in comparison to gas and oil boilers, district heating and heat pumps. It is shown that older, conventional wood combustion tech­ nologies have the highest emission factors. The highest emissions seen in older units are due to smoldering conditions and limited heat-storing, so combustion gases are less burnt than in modern units [31]. A review in terms of wood smoke health perspectives by Bqlling et al.

show that smaller particles (20-125 nm) are more common amongst logwood, pellet and chip advanced boiler units while open fireplaces and con­ ventional (old) stoves produce nanoparticles in the range of 50-600 nm [10]. This suggests that as conventional units are displaced by modern, more efficient and “cleaner” systems, although the total particulate emissions are reduced, the number of NPs may increase. However, this is not accurately represented by current regulatory gravimetric mea­ sures as NPs contribute very little mass even when present in significant numbers.

Source types

Fig. 5. Particle pollution of different heat sources compared with diesel truckdata from Denmark adopted from [5].

quality briquettes are produced under higher processing pressures and temperatures, longer dwell times, and increased binder contents [51,52]. Briquettes may be used in automatic or manually fed systems; the typically automatic systems only include primary heaters or com­ mercial grade systems. 5. Combustion conditions influencing nanoparticle emissions Wood-heating appliances have highly variable combustion condi­ tions which cause large variations in the physicochemical character­ istics of the emitted particles [21]. The four main combustion condi­ tions explored in studies typically include start-up (initial burn period), steady state (efficient), intermediate (a modulating stage with an output lower than the unit's maximum) and burn-out (final phase of the burn). These periods are most often identified by the carbon monoxide (CO), carbon dioxide (CO2), oxygen (O2), and mass of fuel values. The steady state phase is where the maximum efficiency is often achieved and PM emissions are reduced. Burn-out is often described as the charcoal phase. Some heaters have the ability to modulate down via control strategies such as those mentioned in the advanced boiler systems or manually controlling the draught damper in a stove for instance al­ lowing for 300Xi of full output or a ‘low’ burn.

The most common op­ erating conditions and their characteristics are listed in Table 1 above. One additional common operating characteristic, but certainly a less desirable, is incomplete combustion. Manufacturers often include in­ structions to users which would prevent such practice. However, many units will operate in an incomplete combustion mode if poor practices are followed. Incomplete combustion is caused by small values of the three most important ‘T's’ in combustion; time, temperature, and tur­ bulence. Too short of a residence time to allow for proper mixing, in­ adequate mixing of combustion and fuel in the combustion chamber due to a lack of turbulence or oxygen, and low combustion tempera­ tures or cold zones in the combustion chamber are examples for in­ complete combustion parameters [53,54]. Characteristics of incomplete combustion are also an increase in volatile hydrocarbon emissions and CO. The constituents of fine particle emissions during incomplete combustion are organic matter, elemental carbon, and fine ash [55]. Typically, biomass heating systems are able to operate at a high efficiency and low emission level in their steady state, full output mode. However, when inefficient systems are used, poor operating techniques are employed, or poor fuel choices are made, all causing for an in­ complete combustion, higher PM emissions occur.

A Swiss study pointed out emissions could be up to 100 times higher if a stove was not operated properly [5]. In addition, if an advanced system's heating load is decreased, cordwood and automatic wood-fired boilers (pellet and chips) may have trouble modulating (controlling the burn rate/output through use of fans and or dampers), causing them to cycle frequently resulting to higher PM emissions and lower efficiency.

6. Nanoparticle emissions during residential wood heating Current emission regulations from wood combustion devices in the US and EU are based on an overall mass concentration of particles and therefore NP levels may not be fully reflected in these emission reg­ ulations due to their low mass contribution. Further, as mentioned above NPs are of significant interest due to their risk of exposure in terms of health and the environment and so many groups have begun investigating (measuring, counting, and characterizing) the NPs asso­ ciated with RWC. Results from a study involving 14 different appliance types, from cordwood stoves to automatic feed boilers showed 950Xi (particle number) of the particle emissions were smaller than 400 nm [56]. Hata et al. found in laboratory studies that more than 300Xi (of cumulative mass) of the smoke particles from the burning of biomass fuel had a mass that fell within a range of í 100 nm and this con­ tributed greatly to the total levels of toxic PAHs and water soluble OC [22]. Bqlling et al. [10] described three typical classes of wood 520

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

combustion particles based on chemical composition and morphology; spherical organic carbon particles, aggregated soot particles, and in­ organic ash particles. Their review shows the smallest (aggregated soot particles, 20-50 nm) stem from combustion in conventional stoves, open fireplaces, and boilers fueled with cordwood, wood chips and pellets during high temperature and incomplete combustion phases. Saarnio compliments Bplling et al.’s description of particle classes by defining the main components of ultrafine particles as EC, OC, sulfates (SO42-), and trace metals [3]. The following sections provide details on nanoparticle type, size, and or concentration found in their respective studies.

approximately 107 nm) [63]. They also found that an increase in combustion in air supply impacted the particle size distribution; thus more oxygen caused an increase in the concentration of particles with smaller diameter. In addition, the study showed the particle bound PAHs decreased as oxygen levels increased; organic carbon was found to be 14Z for burning conditions with increased O2 and 57Z when there was an absence of O2 [63]. Particle bound PAHs were likely re­ sulted from adsorption or condensation processes. Kocbach et al. [64] collected particles from two wood combustion sources in Norway; one from a small farm heating from a wood stove and the other ambient sample collected from a residential neighbor­ hood who uses wood as a major heating source, collecting all stages of the burn cycle. To confirm the particles collected were indeed all from wood sources, the group took direct samples from a common single­ stage Norwegian stove with a high temperature combustion chamber (1000 "C) and another during an air-starved combustion condition with a low temperature of 500 "C. The results from X-ray microanalysis (XRMA) the study indicated Potassium (K) and sulfur (S) were the most abundant elements observed in the particles, while Silicon (Si) and Calcium (Ca) were present but slightly less. Kocbach et al.

proposed the S levels in the carbon aggregates were not directly from the source but rather caused from a reaction with gases in ambient air. While the mean diameters for primary combustion particles for both locations remained very similar, roughly 39 nm ±11 and showed consistent number of particles (226-239), the morphology differed from Transmission Elec­ tron Microscopy (TEM) analysis. The high temperature combustion produced carbon aggregations while the low temperature combustion did not have aggregates but carbon dominated spherical particles with diameters between 50 and 600 nm [64]. The NPs produced during the low-temperature combustion were similar to particle from smoldering forest fires. A later study by Kochback et al. [65] collected NPs from a common single-stage Norwegian wood stove in a laboratory at high-temperature burn conditions, following Norwegian standards for testing wood stoves using a dilution tunnel. The group examined the NPs collected with High Resolution Transmission Electron Microscopy (HRTEM) to further determine the microstructure of the NPs. Similar to an earlier study [64], only carbon aggregates were found with the primary particles having a mean geometric diameter was 31 ± 7 nm from TEM analysis.

From HRTEM analysis the group found the wood smoke NPs had one nuclei and a turbostractic structure; a concentric arrangement of layer places with a two-dimensional graphitic crystal structure, lacking the ordered stacking of graphite [66]. In addition, the group found the NPs to have a total PAH count of 9745 ng/mg. Corsini et al. [67] studied an advanced cordwood stove with a nominal output of 8 kW. The stove has state-of-the-art triple air supply system with both primary and secondary air supply. They found the ultra-fine particle (UFP) number concentration was 7.0 x 108 particles/ cm3 with an average GMD of 50 nm for beech logs while the UFP number concentration was 1.9 x 108 particles/cm3 with an average GMD of 99 nm for conifer logs. In addition, the group found the PAH concentration was significantly higher in UFP from wood logs, with conifer wood logs generating approximately 20X more PAH compared to beech wood logs [67].

6.1. Cordwood stoves Hosgood et al. [11] studied stoves from 15 Chinese homes which used coal or wood. Using an aerosol monitor, portable diffuser charger and electrometer, the group measured in real-time concentrations of NPs in the ambient air surrounding these homes. They also measured total PM2.5 one to two meters from the main combustion source. Their findings suggested NP levels were not adequately accounted for in the total mass records due to NPs insignificant mass contribution to PM2.5 measurements; this supports arguments of needed particle type and count regulations in addition to the current mass based emission reg­ ulations. Overall the group concluded there was a clear association between cooking and heating activities and elevated local NP levels [11]. Bologa et al. [61] tested an 8 kW stove with dry beech log wood and characterized the particle size distribution as well as the particle mass concentration. Their study showed the short start-up time of five to seven minutes produced more particles than the steady state and burn­ out phases. Stack temperatures for the start-up, steady state and burn­ out phases were ~100"C, 300-330"C and 160-180 "C, respectively.

The group showed the start-up and steady state phases to have a monomodal particle size distribution of 180 nm while the burn-out phase had the highest concentration of nano-sized particles with a bi-modal dis­ tribution of 180-200 nm and 30-40 nm. Stack emissions were diluted from a welded construction stove and analyzed for their size distribution and total number concentration in the Gaegauf et al. study [56]. The start-up period of the stoves showed the largest particle size (i.e., average 200 nm), followed by the steady state period with an average of approximately 175 nm and the smallest particles (~ 30 nm) were found in the burn-out period. Similar to the study above, the lowest number concentration of particles less than (0.8 pm) was seen in the steady state period and the highest in the burn­ out. Hays et al. studied the polycyclic aromatic hydrocarbons (PAHs) associated with both Douglas Fir and White Oak at high ^ 24Z) and low (13Z) moisture contents in a freestanding steel woodstove. Their experiments were run in compliance with US test standards, measuring only a steady state condition, and used an Electrical Low Pressure Impactor (ELPI) to collect filters and analyze the PAHs. They reported that the geometric mean diameter (GMD) is between 313 and 662 nm particles representing an accumulation mode [62].

The study showed low molecular weight PAHs preferentially segregated to fine aero­ dynamic diameter particles while higher molecular weight PAHs pre­ ferred coarse particles. Hueglin et al. [63] studied the size and particle bound PAHs. A dilution sampling system was used to collect the particles, employing a scanning mobility particle sizer (SMPS) measure the particle size dis­ tribution. The stove had a 15 kW output and was fired with beech wood having a moisture content of 15-180Zo. The results showed the size distribution was strongly dependent on the state of the combustion process; start-up had the greatest particle concentration and largest mean diameter (approximately 260 nm), while the burn-out phase had the smallest mean particle diameter (approximately 60 nm) and inter­ mediate phase had the lowest concentration (mean diameter

6.2. Pellet stoves A study by Commodo et al. found the smallest particles emitted from a pellet stove were on the order of 7-30 nm and were also the most abundant; on the order of 1x104particlesZcm3 [68]. The study also showed the amount of particles emitted after the transient phase de­ creased significantly, showing fewer particles ^ 100 nm) in the steady state or stationary phase. Corsini et al. [67] also studied an 11 kW top-fed pellet stove with two different pellets—beech and conifer. In agreement to the results from the cordwood stove, higher GMD diameter sizes were measured 521

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

with the soft-wood fuel than the hard-wood fuel with diameters of 75 nm and 38 nm. The average UFP number concentration for both the beech and conifer pellets was 1.1 x 108 particles/cm3. These in­ vestigators reported that there were lower quantities of PAHs associated with the particles generated by the pellet stove than the cordwood stove. This was most likely due to more complete combustion in the pellet stove that ultimately produces less inflammatory active particles [67].

Mavrocordatos et al. [18] collected particles from a 15 kW pellet boiler flue for a detailed analysis of particles which included: particle count, size distribution, morphology, volume, elemental composition, and fractal analysis. Employing a SMPS, they found that the maximum in number concentration measured 80 nm. From Atomic Force Micro­ scopy (AFM) and Scanning Transmission Electron Microscopy (STEM) the group found the individual particles of different shapes and ag­ gregates with a highly compact structure. The group noticed when the particles were exposed to the electron beam in the STEM, after only a few minutes the smaller particles ^ 100 nm) would disappear and the larger would reduce in size and to an inorganic residue consisting mostly of K, Chlorine (Cl) or S. Using an inductively coupled plasma atomic emission spectrometer (ICP-AES) the group was able to do an elemental analysis in addition to a total organic carbon (TOC) mea­ surement by using a TOC-Analyzer. The analysis showed a TOC content of 370 ppm, the presence of K, Ca, S, Magnesium (Mg), Cl, Phosphorus (P), and Sodium (Na) in decreasing amounts (ppm) and traces of Iron (Fe), Aluminum (Al), Mg, and Zinc (Zn). From the analysis the group drew the conclusion that these wood combustion emission particles were likely inorganic salts that could dissolve easily in water [18]. Lamberg et al.

[9] along with Kaivosoja et al. [69] studied a 25 kW modern pellet boiler using pine pellets. The group found the GMD to be 60 nm with a mean mass diameter (MMD) of 137 nm [9]. In addition, the total PAH emissions from particulate matter 1 qm and smaller (PM1) emissions (PAH/PM^ were extremely low, 0.00006Z, with pyrene and fluoranthene as the most abundant. Kaivosoja et al. sup­ ported this work by characterizing the PM1 by elemental analysis. The group found K was the main element found followed by Na, Cl, and S —suggesting that these elements were all in the form of alkali metal sulfates (SO4) and chlorides (Cl-) [69]. Gaegauf et al. [56] also measured the particle mobility diameter distribution of a 17 kW pellet boiler; diluted from a hot stack. They found a peak mobility diameter occurred at approximately 85 nm, but ranged from 12 nm to just below 400 nm.

6.3. Cordwood boilers Schmatloch and Brenn [58] found from two different cordwood boilers of 41 and 35 kW, a modern lambda (oxygen) controlled and fan assisted boiler and a traditional boiler, respectively, that the advanced unit produced slightly smaller particles. Using an ELPI and SMPS, the maximum of the number size distribution laid between 80 and 180 nm but varied significantly for different phases; highest concentrations were found during the start-up. Johansson et al. [31] studied a variety of cordwood boilers which included old, traditional units and modern, advanced systems with power outputs ranging from 6 to 34 kW. Using beech cordwood with moisture contents of 15, 26, or 380Zo moisture content, the group in­ vestigated the effects of varying the load; running the systems at 1000Zo full output or firing the units with a full or half load of cordwood. The group used an ELPI to determine the size distribution and Dekati LowPressure Impactor (DLPI) to determine the number particle concentra­ tion, diluting the samples before measurements. In addition, the group sampled for PAHs as well. Ultimately the group found older wood boilers had higher emission concentrations and un-oxidized compounds than modern wood boilers [31].

In fact, particle emissions were de­ creased by 70-80 times when comparing a modern wood boiler with a ceramic-lined burn-out zone, connected to an external heat storage tank. The PAHs were also found to be the highest in the case of the older wood boilers. Size distribution data indicated a peak mean diameter slightly greater than 60 nm and a decreasing trend in terms of con­ centration as particle diameter grew for both the older wood boilers and fan assisted combustion modern wood boilers. Interestingly, the dia­ meter range shifted towards smaller particle sizes in the case when modern wood boilers were fired with a natural draught.

6.4.2. Chip boilers Using a 40 kW wood chip boiler, Torvela et al. [21] studied the chemical composition link with particle morphology and the effects of combustion conditions on particle type. These investigators looked at three combustion phases that they classified as efficient (low level of CO: 0-100 mg/MJ), intermediate (elevated CO: 100-1000 mg/MJ) and smoldering (high CO: 1000-5000 mg/MJ) [21]. Using a SMPS the mode number size distribution was 37, 150, and 250 for the efficient, inter­ mediate, and smoldering conditions, respectively. They calculated a GMD for the efficient, intermediate, and smoldering conditions to be 25, 65, and 160 nm, respectively. However, even during the immediate and smoldering phases, a smaller number of particles with a dimension of 40 nm were observed. Torvela et al. also identified via TEM the three main wood com­ bustion particle types discussed in Bolling et al.’s review article [10]; ash, soot, and particulate organic matter (POM). The TEM images showed the ash as crystalline structures smaller than 100 nm, the soot particles chained as agglomerates with a primary particle size of 30-50 nm and the POM as an amorphous type with a course surface [21]. Table 2 below summarizes the findings of particle morphology and composition from Torvela et al. Similar to the study of Mavrocordatos et al. [21] reported above, Torvela et al.

found many of their species vaporized, decomposed, and changed during TEM analysis as well to the form crystals or bubbles inside particles, suggesting there may be semi-volatile hydrocarbons adsorbed on the surface of the soot. The group included an EC/OC (mg/ MJ by thermal-optical), PAH (qg/MJ) by gas chromatography-mass spectrometry (GCMS) and PM^mg/MJ) by gravimetric analysis. In all cases, the largest counts for each were found in the smoldering condi­ tion and the lowest in the efficient burn. Gaegauf et al. [56] also studied a wood chip boiler with a 70 kW

6.4. Automatic feed boilers 6.4.1. Pellet boilers In addition to the cordwood boilers studied by Johansson et al. [31], the group included pellet boilers ranging from 3 to 22 kW in power output. The pellet boilers were operated at full or intermittent loads (less than 1000Zo output) with pellet fuel having a moisture content of 7.60Zo or bark pellets having a slightly higher moisture content of 7.8Z but significantly higher ash content of 0.5-3.7Z, respectively. Two pellet boilers run at their full output of, 11 and 22 kW power output ignited with a pilot flame and electrical ignition, respectively, showed very similar mass distributions and maximum particle sizes of 130 nm. Experiments that involved the bark pellets showed an increase of mass concentration and a slight shift to larger particle sizes (approximately 220 nm). No obvious trend was observed between full and intermittent output in terms of particle size, only a very slight increase in mass concentration. Bologa et al. [61] also characterized the particle size distribution from 20 kW and 32 kW pellet boilers with flue gas temperatures during steady state operation in the range of 170-180 "C and 120-140 "C, re­ spectively.

Similarly, to the cordwood stove tested by the same group, the steady state pellet boilers also had a mono-modal particle size distribution; however, there was a shift towards a smaller particle size, peaking at about 60 nm. Furthermore, the group showed a decrease in particle mass concentration for the pellet boilers as well; pellet boilers at 10-20 mg/Nm3 vs the wood stove at 25-40 mg/Nm3. 522

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

Table 2 Summarized Findings from Torvela et al. [21] Study Investigating NP PM emissions from Wood Chip Boiler. Combustion conditions

Particle size and composition

Efficient Intermediate

Particle diameter < 100 nm, irregular shape. Most particles were ash with crystalline nuclei; some identified as zinc oxide. Particles appeared to be more spherical in shape with coating of organic matter. Sizes ranged from 40 to 150 nm with the largest particles making agglomerates that contained ash, soot and organic matter. Almost all large branched soot agglomerates. Some inorganic elements were detected, but were mostly a mixture of soot and POM with a minor amount of ash.

Smoldering

Dispersion Spectroscopy) and OC, and the efficient burn produced particles which were merely composed of ash (mainly K, O, with larger amounts of P, Zn, Si and S compared to the intermediate burn).

output; diluted from a hot stack. The group varied the excess air during the operation of the chip boiler to study its effects on particle size and found that lower excess air values corresponded to lower particle counts and a shift towards higher diameters; H.6% O2 and 2.70Xi O2 corresponded to diameters of 60 nm and 90 nm, respectively. Hueglin et al. [63] continued their size distribution and particle bound PAH study focusing on a chip boiler with an output of 60 kW, and enriched the study by looking at the particles morphology as well. The group coupled the SMPS with an aerodynamic particle sizer (APS) to gain more information about the particles and measure larger dia­ meters. The wood chip boiler was almost always run in a steady state full output mode as it is computer controlled and burned chips with a moisture content between 400Xi and 700Xi. Similar to the studies dis­ cussed already, particle emissions were dependent on operating con­ ditions. Since the study only looked at steady state operation, the only varying parameter was the combustion air supply to mean the excess oxygen concentration target. The GMD size distribution had its lowest value when O2 values were at 6.20X) (68.6 nm) and its largest mean diameter size at an O2 value of H.6% (95.7 nm); the concentrations also followed a similar trend [63].

Overall, the NPs concentrations were reduced when the boiler was operated at an optimized O2 value. The group found particles with a diameter greater than 1000 nm to be negligible when using the APS. Particles emitted from the chip boiler contained low amounts of semi-volatile compounds as well. SEM images revealed different types of particles; agglomerates of spherical compact particles, perfectly spherical particles representing coal fly ash, or irregular shaped crystalline platelets. Another study by Leskinen et al. [70] focused their investigation on a wood chip-fired appliance. The group created a novel modular com­ bustion unit which included a reactor and moving grate (not a marketed boiler with given output) which allowed them to produced three dif­ ferent combustion conditions: efficient (representing optimal condi­ tions), intermediate (representing a partial load or poor air-stage set­ tings) and smoldering (burn-out or smothering situation observed in batch combustion) with similar CO levels as seen in the work done by Torvela et al. [21]. The group provided detailed information on the system and fuel type which was both bark and stem wood chips from spruce and typical Finnish broadleaved trees with a moisture content of 24.30Xi.

The group was able to measure the number and mass con­ centrations as well as the size and mass distribution in addition to characterizing the particle morphology, EC/OC, elemental and ion analyses, PM1 and PAH compounds, similar to Torvela et al. [21] re­ ferenced above. Data from each of the burn periods indicated the smallest mass emission concentration (mg/MJ, (amount of particles per kilogram of dry wood combustion)) and smallest particles were pro­ duced during the efficient period (average diameter of 35-75 nm) and the largest particle diameter size and mass emission concentration were observed during the smoldering condition (average diameter ~200 nm). From the groups chemical analysis, they found Zn was the most abundant metal with the highest relative fraction occurring during the efficient conditions. The fraction of PAHs were found to be lowest during the efficient burn and highest during the smoldering. TEM analysis showed the smoldering burn produced particles were made up of mostly large agglomerates of soot (EC and OC), intermediate burns produced a variety of particles from soot agglomerates and ash or composite particles of ash (mainly K, O, Zn determined from Electron

6.5. Fireplaces and masonry heaters Fireplaces and masonry heaters are often used in studies due to their lack of controls and so it is easy to study the effects of different types of biomass without being skewed by device controls. As early as 1982, emission measurements were to study the size and composition of particles were taken from two residential fireplaces and one free-standing fireplace to determine the influence they potentially had on Denver's air quality. Muhlbaler Dasch [71] made both in-field and laboratory measurements, investigating five softwoods, nine hardwoods and two synthetic logs (similar to a biobrick construction). Laboratory and residential test followed EPA method 5 protocols, clo­ sely, so flue gas was diluted. A Sierra Series-220 cascade-impactor was used to determine the size distribution of particles collected. While the log type and size seemed to have the greatest effect on particulate emissions in terms of total particulate emissions (g/kg), there were no obvious differences in terms of size fractionation for the three types of wood; the mass median diameter was found to be 170 nm [71]. From SEM imaging, Muhlbaler Dasch also found the particles to be agglom­ erates and spherical in shape.

A study in Italy focused on investigating the particle size distribu­ tion of NP emissions from an 11 kW commercially available fireplace burning beech wood (the fuel for European standards) [2]. The group analyzed 15 tests over three different burn periods: start-up, steady state, and burn-out. Similar to US test methods the group measured from a dilution tunnel and further diluted the flue gas to achieve less particle coincidence. To determine NP concentrations an ELPI and a Fast Mobility Particle Sizer (FMPS) were used. It was concluded NPs accounted for 670Xi of the total particles measured with the mode par­ ticle diameter of the average size distribution to be around 10-20 nm, in all three phases, but the highest concentrations were found in the start-up phase and lowest in the steady state period [2]. The group found particles in the range of 100-326 nm as well but suggested they were likely primary soot particles that underwent condensation in the diluted and cooled fuel gas [2]. A dip in O2 values corresponded to a decrease in number of NPs as well. Interesting in terms of environ­ mental fate of wood combustion NPs, the group showed that due to dilution and cooling of flue gas, stable distinct size distributions during each separate phase were achieved with only slight changes during the transition from one phase to another. In a later study by Ozgen et al.

[72], using the same 11 kW stove as mentioned above [2], nanoparticles with a GMD of 28 nm accounted for 320X) of the emissions whereas larger particles with a GMD of 127 nm accounted for 680Xi of the emissions. Similar sampling techniques were employed using a dilution tunnel. The investigators found the ignition phase was responsible for 460Xi of the NP burden and cold starts, which are common practice in the field, resulted in a magnitude higher in emissions. They reported that larger soot particles emitted during the flaming combustion phases were coated by condensed semi-volatile organic species that were diluted and cooled in the flue gas; this finding suggests that minimizing incomplete combustion is enough for particle number reduction. Similar to other cited studies, the Ozgen et al. [72] 523

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

Table 3 Summary of average particle size from cited sources. Equipment Cordwood Stoves

Pellet Stoves Cordwood Boilers

Pellet Boilers

Chip Boilers

Fireplaces and Masonry Heaters

Summary: average particle sizes are given unless otherwise specified Start up: 180-260nm Steady state: 30; 107; 175-180nm Burn out: 30-60nm; 180-200nm Burn out or start up showing the highest concentrations of NPs and steady state having the lowest concentrations of NPs In some studies, aggregations of NPs with diameters of 313-662 nms Low MW PAHs favored small NP diameters & High MW PAHs favored coarse particles Increase in oxygen in combustion chamber caused for smaller particles Decrease in oxygen caused PAH levels to increase and bind to NPs Image: high temperature combustion produced carbon aggregations with single particles having a single nuclei and turbostratic structure; low temperature combustion produced carbon dominated spherical particles 7-30 nm Steady state showed lowest concentrations of NPs 60-180 nm; but as small as 12 nm to as large as 400 nm Highest concentrations seen during start up Advanced boilers produced smaller particles and lower concentrations than traditional boilers PAHs higher in older boiler technology 60-130 nm (no bark pellets) 220 nm (bark pellets) No obvious trend was observed between full and intermittent output in terms of particle size, only a very slight increase in mass concentration.

Particles were likely inorganic salts Steady state: 25-75 nm; lowest particle concentrations Modulating: 65 nm Smolder: 160-200 nm; highest particle concentrations Lower excess air values led to lower particle counts and larger particles, however another group found greater concentrations with increased oxygen SEM & TEM images indicated different types of particles; agglomerates of spherical compact particles, perfectly spherical representing coal fly ash or irregular shaped crystalline platelets. Efficient was typically salts and crystalline while intermediate was a variety of soot agglomerates and ash; smolder was large agglomerates of soot PAHs bound to NPs lowest during efficient burns and highest during smoldering conditions 10-20 nm; 50-65 nm; 170 nm 70-250 nm during smoldering 100-326 nm range assumed to be agglomerates owed to cooling Start-up had highest concentrations; steady state had lowest Decrease in oxygen corresponded to decrease in NPs Newer technology stoves seemed to produce more NPs Agglomerates with individual particles spherical in shape Normal combustion showed spherical and irregular shapes while smoldering combustion showed more irregular and sintered-like structures Some salts but a presences of heavy organic compounds as well

References [56,61-65,67]

[67,68] [31,58]

[9,18,31,56,61,69]

[21,56,63,70]

[2,55,71 -73]

from both normal and smoldering wood combustion periods in a con­ ventional masonry heater made of soapstone using birch cordwood at 70Xi moisture content. The results from the study showed a higher particle emission number measured in a dilution tunnel by the ELPI and FMPS during the normal combustion and less than half the particle emission number during smoldering state with geometric mean dia­ meters of 56 and 65 nm by FMPS and ELPI measurements, respectively, for normal combustion, and 118 and 160 nm values for smoldering combustion [55]. Surprisingly for cordwood combustion, the repeat­ ability was good. Even if all operating parameters are kept constant, manual fed cordwood appliances are very variable [58]. The study found the average particle number size distribution for normal com­ bustion to be unimodal at 65 nm and the smolder combustion to pro­ duce a bimodal distribution with peaks at 70 and 250 nm [55]. In ad­ dition, SEM and TEM analysis from the hot stack showed both spherical and irregular shaped particles making up larger agglomerates and fine ash particles remained as separate particles. Normal combustion showed spherical and irregular shapes while smoldering combustion showed more irregular and sintered-like structures.

The elemental analysis showed single particles were composed mainly of carbon, but also contained K, S, Zn, and less commonly O, Ca, Fe, Mg, Cl, P and Na were found [55]. Similar to studies mentioned above, particle material volatized under the beam indicating the presence of heavy organic compounds. The group was able to suggest the shifting of particle size during the different combustion conditions seemed to be determined by the amount of condensed organic vapor in the flue gas [55].

found the start-up and flaming phase were characterized by small particle diameters whereas smoldering conditions had larger particle diameters. Number concentrations were greatest during the start-up phases and especially for cold start-up phases but were lower for smoldering phases. A group in France looked at the mass and number distributions of wood combustion particles emitted from beech cordwood (120X) moisture content) during normal and reduced operation from two dif­ ferent fireplaces; one a conventional fireplace (13 kW) with only pri­ mary air intake from the front and the other was a more compact ‘next generation’ fireplace (4 kW) with primary and secondary air intake. Measurements were taken from a dilution tunnel using a gravimetric based filter measurement and ELPI. Surprisingly, the study found the opposite trend of most, in that the older technology was more favorable to less particulates but ultimately showed agreement that more NPs were produced in the newer technology. The study found the next generation fireplace produced much higher particle emission numbers for than the conventional fireplace; specifically, for the reduced con­ ditions, the emission numbers were 50 times greater in the more modern fireplace than the older conventional unit [73].

While 800Xi and 600Xi (total number of particles) of the total PM consisted of particles with a diameter less than 100 nm, for the new and old fireplaces, re­ spectively, it only contributed to a very small extent of the total mass of the aerosol; 5.5-5.80Xi and 2.6-0.30Xi for the new and old fireplaces, respectively during normal and reduced operations [73]. Another group in Finland looked at the PM and gaseous emissions

524

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

In addition to the work done by Lamberg et al. [9] on a pellet boiler, the group also studied three conventional masonry heaters and one modern masonry heater made of soapstone. All masonry heaters burned birch logs with a low moisture content (10-130Xi on a dry basis). The group explored continuous combustion, steady-state periods, cyclic operation, and ignition phases. Overall, GMD sizes ranged from 63 nm to 142 nm. The largest diameters were found during incomplete com­ bustion phases and also produced bimodal number and size distribu­ tions; the ignition phase also produced a larger GMD [9]. Similar to the pellet boiler discussed above, the most abundant PAHs were pyrene and fluoranthene in addition to phenanthrene and cyclopenta[c,d]pyrene for one of the masonry heaters. Total PAH emissions from PM1 emis­ sions (PAH/PM^ were higher than that of the pellet boiler but still reasonably low with a range of 0.15-2.80Xi.

the old-type wood boiler could be reduced if it was connected to a storage tank, similar to advanced modern units, higher quality fuel (lower moisture content) was used, or if fuel amount was decreased and small batches of fuel were loaded instead of a full charge. The study found poorer combustion conditions increased the amount of particles smaller than 100 nm. Finally, the group compared the wood combus­ tion devices to oil burners finding pellet systems rivaled oil burners but old-type wood boilers would have low efficiencies, high PM emissions, and unoxidized gaseous compounds [31]. While advanced systems that employ multiple control devices may decrease the amount of PM emission concentrations as well as reduce the amount of harmful PAHs produced, they tend to be expensive. In 2006, the International Energy Agency commented on using ESPs to reduce submicron particles but such devices were very expensive and not affordable for residential use [27]. Another review suggested the easiest way to control emissions is to use a high quality fuel with low ash content, low moisture, and constant piece size whose options may also not cost competitive, or boilers must use proper combustion chamber designs and adequate air pollution control technology [54]. Therefore, there is a need to develop cost sensitive, yet effective, sub­ micron PM reduction technologies.

One interesting technology dis­ cussed in a review by Lim et al. [77], is the use of a miniature pipe bundle heat exchanger (a technology used in the automobile industry) where particles migrate to the lower temperature surfaces and deposit on the tube walls to remove particles from the flue gas. Other options may also include better combustion designs or controls, such as lambda probes. Lambda probes can detect the oxygen levels and effectively modulate the air to fuel ratio. Several studies above indicated as op­ timal oxygen levels were reached, less particles were formed and a shift towards larger particle sizes was achieved [2,56,63,78]. Perhaps lambda sensors could be used to favor a combustion condition which reduces the amount of harmful NPs produced. Incomplete combustion, smoldering, and transition conditions are known to favor PAH and black carbon production. Further, when temperatures are low, PAHs can condense or adsorb onto particles [77]. Smaller NPs are known to carry larger PAH molecules and are difficult to remove, while larger more coarse particles carry lower molecular weight PAHs [77]. Likewise, Ghiassi et al. found the most reactive soot had an amorphous nanostructure which was composed of short in­ dividual layer planes with no orientation relative to each other [79]. The laboratory study by Hata el al.

indicated biomass type and the heating rate had a significant effect on the water soluble OC fraction and the smaller the NP, the more water soluble OC the particle con­ tained [22]. Very few studies have looked at the microstructure of NPs but this could also provide important information on the NPs reactivity. The microstructure of NPs can give information about particle forma­ tion, condensation, coagulation [65,80,81] and the ability to accumu­ late reactive substances [65,82]. Ideally, a NP with a low mobility is desired to reduce inhalation or atmospheric risks. Ultimately, it is important to understand the rate of environmentally induced changes and how the migration of particles changes their properties [76]. Understanding the aggregation behavior under both atmospheric and aquatic conditions may help further pre­ dict their potential interactions with the ambient environment [26]. Mavrocordatos et al. [18] investigated the affect salt content within the particles play and found their morphology was significantly affected in aqueous environments. Perhaps if a sufficient amount of information was gathered on NPs soot emissions from RWC which included explicit details on fuel type, operating parameters, and sampling conditions, manufacturers and operators could be informed with a ‘best practices’ procedure.

As shown in many of the studies reviewed herein, NPs account for a substantial amount of the particle emissions associated with biomass combustion. However, standards enforced by regulatory agencies are mass concentration based, focusing on size ranges less than or equal to 10 pm (total PM; PMT) or less than or equal to 2.5 pm (PM2.5). Due to

7. Discussion This literature review revealed two earlier review papers and about eighty research articles and reports. An earlier review by Kumar et al. reports only limited information on NP emissions from non-vehicles exhaust sources [74]; the authors reported only two studies associated with NP emissions in RWC. The current review identified and reviewed many more studies that explicitly or implicitly provide information regarding NP emission data from RWC. However, measuring and quantifying the presence of NP in emissions, and how these emissions vary based on fuel, appliance type, and combustion period are not al­ ways well investigated making correlations hard to draw. The second major review by Bplling et al. (2009) pointed out there has been a lot of work done in steady-state operating conditions but there is limited data on particles from conventional cordwood boilers, modern wood stoves, and modern cordwood boilers [10]. The current review shows that this statement still holds true eight years later. Table 3 below summarizes the information collected from the papers above. In addition, cold starts, modulation periods, and in-efficient burns exist and there may be an underestimation of real-life emissions which contain particle frac­ tions that can be associated with health impacts in local areas [56]. As reported by Tom et al.

[75], the majority of published studies were associated with laboratory experiments and field performance mea­ surements were lacking. There is a significant need to characterize the particle types in biomass smoke, since the regional and global climate effects of aerosols from biomass burning depend so strongly on particle properties [17]. However, manually fed cordwood stoves and boilers have a tendency to vary in their burn conditions even if fueled exactly the same way. Schmatloch and Brenn directly observed this when fueling a cordwood furnace and trying to reproduce the size distribution in particles; burn one had much fewer concentrations of NPs less than 140 nm and greater amounts of NPs larger than 220 nm than burn two which showed op­ posite trends [58]. This is owed to the nature of wood- how it burns and falls within the combustion chamber. This makes reproducibility very difficult and particle size distributions to have a larger variation than advanced automatic feed boilers [58]. This statement compliments the findings by Gaegauf et al. [56]. This also makes it difficult to compare particle size distribution among studies. To complicate matters, Baer indicates that nanoparticles are inadequately characterized since they change with time and can be altered or damaged during analysis [76], so precautions must be taken to not bias research.

Some key factors which may cause some uncertainty are sampling locations, sampling temperatures, and dilution ratios [74]. Ozgen et al. point out that the variability of the results concerning the number emissions and the size distribution during different stages of the wood burning cycle suggests that emissions are strongly connected to the combustion conditions depending on the appliance design and operation [72]. The study by Johansson et al. [31] emissions from pellet boilers were closer to those of an oil burner while old-type wood boilers had the highest. An important piece from the study was that emissions from 525

R. Trojanowski, V. Fthenakis

Renewable and Sustainable Energy Reviews 103 (2019) 515-528

NPs small size, current measures may not be appropriate for particles in the nano and ultrafine size fraction since they contribute very little to the overall mass. However, they contribute significantly to the overall number of the airborne particles present [11,83] and, potential adverse health effects posed by NPs may go unknown [19]. Lim et al. [77], reported that dilution sampling affects particle emissions by promoting nucleation but results in an increase in the formation of ultrafine par­ ticles which are better represented by particle number measurements. Torvela et al. investigated the intermittent period when switching from one condition to another, finding the properties of the PM emissions varied significantly even within a short period suggesting that an eva­ luation of environmental and adverse health effects based on average particle property values can be misleading [21]. Further, Lamberg et al. concluded from their study that PM1 mass does not reflect all of the variation affecting possible toxic properties of fine particles originating from combustion [9]. From the Gaegauf et al. [56] study in particular, the highest con­ centration of total suspended particles (TSP) (mg/m3) correlated nicely to the highest total number of nanoparticles ( < 600 nm) collected per MJ of output.

Masonry stoves had the highest TSP concentration along with the highest emission factor of NPs per MJ. In comparison, auto­ matic feed pellet boilers had the lowest TSP and the lowest total number emission factor of NPs. However, when looking at the emission factor (mg/MJ) for both TSP and NPs data in detail, the group noticed while each appliance type typically showed good correlation between the two values, individual cases such as the cordwood boiler showed roughly a 400Xi reduction in the NP emission factor from the TSP emission factor. This supports the suggestion to modify existing test method regulations or add to them a measurement involving particle count, size, or surface area. One positive note from a study which compared vehicle emissions to RWC emissions suggests wood smoke may be associated with less ne­ gative health effects as diesel NPs were smaller in diameter and have 410X) more surface area [64]. Unfortunately another study found PAH content was higher for combustion particles from wood smoke com­ pared to vehicle exhaust, causing risk for a high mutagenic and carci­ nogenic potential [65].

more efficient and “cleaner” systems, an increase in released NPs may follow. However, with more efficient boiler systems PAH levels are being decreased; causing one to argue what is more important NPs or PAHs? Particle distribution of batch-wise fired appliances (wood stoves) varied significantly during a burn cycle, while wood log and continuous fed boilers showed a fairly constant particle size distribution. These differences are likely to influence the biological effects induced by wood smoke particles and therefore it has been recommended to ex­ plore how combustion conditions influence the particle properties, their possible health risks, and reactivity within the environment. Although a lot of work has been done to understand the health effects associated with RWC NPs, but little is known about the environmental fate of RWC NPs and their effect on climate. In addition, NP emission data from wood burning processes under real-world operating conditions are lacking. More data are needed about source, expected quantities or concentrations, and environmental pathways to help identify situations which may be more favorable than others. Several studies show results of size characterization but mor­ phology and chemical composition data are limited.

Predicting or studying the aggregation, agglomeration, dispersion, size, solubility, surface area, charge, and composition are all necessary parameters to better predict the environmental fate and any health concerns of bio­ mass combustion NPs. While biomass is often considered and en­ couraged as a renewable energy source, it is important to not ignore consequences of poor combustion practices and NP production. Several articles have suggested pathways to reduce NPs such as emission control strategies or higher quality fuel sources. Emission control devices suggested for large biomass boilers include ESPs and condensing heat exchangers but their effectiveness in controlling NP has not been assessed. Acknowledgement Rebecca Trojanowski is grateful to New York State Energy Research and Development Authority (NYSERDA) for partial support of this re­ search through contracts # 29697 and # 63038, Dr. Barbara PanessaWarren (Guest Scientist at Brookhaven National Laboratory), Dr. Thomas Butcher (Group Leader of the Energy Conversion Group at Brookhaven National Laboratory), and Dr. John B. Warren (Scientist in the Instrumentation Division at Brookhaven National Laboratory).

8. Conclusions and recommendations Current emission regulations from wood combustion devices in the US and EU are based on an overall mass concentration of particles and NPs are neither measured nor reflected in these emission regulations, allegedly because of their low mass contribution. This review revealed that NPs account for a substantial amount of particle emissions asso­ ciated with residential wood combustion (RWC) devices. A long standing argument has been the need to modify existing test method regulations or add to them a measurement involving particle count, type size, or surface area. NPs from RWC cause local environmental impacts and potential global impacts due to large concentrations in the atmosphere. This review identified combustion conditions that may result in increasing amounts of NP emissions. These are associated with the device type, its operation, and the fuel used. We often see an inverse relationship between the total PAH mass and its NP fraction. It is re­ ported that PAH emissions could be up to 100 times higher if a stove was not operated properly, although NP production would be de­ creased. Also it is reported that, if an advanced system's heating load is decreased, cordwood and automatic wood-fired boilers (pellet and chips) may have trouble modulating, causing them to cycle frequently and generate more mass PAH emissions, but lower NPs.