PBDE Emissions during the Start-up Procedure of an Industrial Waste Incinerator by the Co-Combustion of Waste Cooking Oil and Diesel Fuel

This study examined the effect of using waste cooking oil (WCO) as an alternative of diesel on PBDE emissions during the start-up of an industrial waste incinerator (IWI). The co-combustions were designed with 0, 40, and 60% WCO injection and become D100, W40D60, and W60D40 multi-fuel combustions. The flue gas was sampled during 4 temperature stages of the furnace: Stage A (< 200°C), Stage B (200–450°C), Stage C (580–700°C), and Stage D (> 850°C). The highest PBDE level was found in Stage A and sharply declined in Stage B by using diesel. The reduction of total PBDE was a competitive result between residue releasing and thermal decomposition in Stage B. The WCO were found to slightly increased the PBDE emissions during the Stage C and D, which provided the suitable temperature for PBDE formation (600–800°C). Therefore, the viscosity became an important control factor when the WCO were utilized as an alternative fuel in IWI operation. The accumulated PBDE emissions during the start-up procedure were 1,099, 1,253, and 1,207 μg by using D100, W40D60, and W60D40, respectively. Additionally, the annual PBDE emissions contributed by start-up procedures increased up to 4.60%, 5.47%, and 5.20% by three fuel combinations, respectively, if the IWI restarted once per month, and became a noticeable issue. Therefore, avoiding unnecessary start-ups was an essential criterion for IWI operation. The small increases (< 1%) of PBDE emissions by altering 40% and 60% diesel with WCO provided a useful information for WCO treatment. This new disposal for waste oil also created a good demonstration of Circular Economy. The overall life-cycle analysis was suggested to be investigated in the following research.

The major emission sources of PBDEs involve several combustion sources, including waste incinerators, metallurgical processes, power-heating systems and open burning of household wastes, such as electronic parts, paints, solvents, and textiles (Altarawneh and Dlugogorski, 2013;Yang et al., 2013b;Chang et al., 2014).As developed countries embark on waste-to-energy incineration as a way forward in treating municipal solid wastes, emission of PBDEs and PBDD/Fs into the surroundings from poorly controlled incineration processes is anticipated, which may thus contribute to the PBDE environmental load (Wyrzykowska et al., 2008;Altarawneh and Dlugogorski, 2013).A study characterizing the PBDD/F and PCDD/F emissions into the stack flue gas of the nine IWIs found elevated PBDD/F levels about 8 times higher than that of MSWIs.This is probably due to the higher bromine content in the industrial waste than in the municipal solid waste, therefore suggesting that IWIs are an important PBDD/F source of emission to the atmosphere (Wang and Chang Chien, 2007).Laboratory waste, another waste stream which is relevant to this study, is generated from experimental, testing, or analysis processes and are known to contain a complex of toxic chemical compounds and they have high heating-value and a high chlorine content (> 9%).Such wastes usually contain liquid wastes which include solvents, metal ions, anions, organics, oxidants, reductants, complexes ions, other impurities and water (Wu et al., 2014).
Numerous researches on applications of waste cooking oils such soap making, biodiesel production, co-combustion in diesel engines, furnaces and burners (Ramadhas et al., 2004;Zheng et al., 2006;Lu et al., 2013;Chang et al., 2014;Mahfouz et al., 2015;Tsai et al., 2015a, b;Tsai et al., 2016) as a raw material and because waste cooking oil is relatively cheaper (Sabudak and Yildiz, 2010).On the other hand, waste cooking oil has high free fatty acids (FFA) content, high water content, and impurities which affects their beneficial uses like in soap making and biodiesel production (Leung and Guo, 2006;Predojević, 2008;Sabudak and Yildiz, 2010), while high viscosity values poses problems in pumping and injection in diesel engines (Kalam et al., 2011;Mwangi et al., 2015).Furthermore, the significant cost of producing biodiesel from raw waste cooking oil remains a major challenge to successfully exploit waste cooking oil as a resource.
The start-up of continuous MSWIs is usually a cold start-up, which occurs after a plant revision and follows certain procedures: (1) Oil burning with a maximum fuel feed rate until the suitable temperature for incineration (850°C) is reached.
(2) Start of waste feeding and increasing feed rate until design load is reached (Gass et al., 2002;Wang et al., 2007).Several studies have investigated the effect of the combustion of a WCO-based biodiesel on the emission of toxic air pollutants from a diesel engine (Muralidharan and Vasudevan, 2011;Chang et al., 2014;Tsai et al., 2016).These studies dealt with the emissions from industrial boilers and diesel-engine generators.To the best of our knowledge, there have been no studies on the emissions of PBDEs from IWIs during the cold start-up period of waste combustion with WCO as a mixed fuel.
In this study, we explored the effects of co-combusting WCO and a diesel fuel in an industrial waste incinerator (IWI) on the emissions of PBDEs into flue gases using three different mixed ratios of WCO and diesel.Gaining a better understanding of the characteristics of PBDEs formed during the co-combustion processes, it may be possible to develop appropriate control strategies to decrease their emissions from IWIs.This study examined in particular whether WCO added to a diesel fuel in combustion could partially alter the levels and patterns of PBDEs emitted into the flue gas during the start-up of the IWI.

Incinerator, Fuels, and Start-up Operation
The basic information of the employed industrial waste incinerator (characteristics and operating conditions) used in this research are outlined in Table 1.The incinerator included a primary combustion chamber to preliminary combust the waste at 850°C and completely destroy the unburned organics in the secondary high temperature combustion chamber at 1050-1450°C.The flue gas was fist quenched by a series of two-cooling tower to shorten the cooled down time pass through 400-250°C.This was because of that the feedstock of an IWI had higher content of dioxin precursors, which tend to reform dioxins and furans at that certain temperature window.The air pollution control devices included activated carbon injection (ACI), baghouse filtration (BF), and wet scrubber.More detail incineration flow diagram and air pollution treatment could refer to our previous studies (Wu et al., 2014).
Three kind of multi-fuel combustion were investigated during the start-up processes in this study: The pure conventional diesel (D100), produced by CPC Corporation, Taiwan, was injected to heat up the incinerator as the based case.The waste cooking oil (WCO) has been injected with specific amount when the certain temperature approached (550°C) during start-up operations.The feeding rate curve (in volume) of total fuels provided, including WCO and D100, were remained the same.In other word, there were two testing cases when the combustion chamber temperature reached 550°C: (1) providing 40 vol.%WCO and 60 vol.%D100, abbreviated as W40D60; (2) providing 60 vol.%WCO and 40 vol.%D100, abbreviated as W60D40.During the above three fuel trials, only D100 was combusted when the Remove acid, and cooling (120°C to 80°C) furnace temperatures were below 550°C.The reason for this is the flash point and viscosity of WCO are higher than those of diesel fuel (as shown in Table 2).Therefore, injecting WCO to partially alter the diesel fuel at temperature around 550°C may improve and enhance flow characteristics and lead to more complete combustion based on some preliminary tests and thermogravimetry (TGA) results in our previous study (Chen et al., 2017) and Fig. 1.

Physical and Chemical Properties of Testing Fuels Fuel Stability
The thermodynamic stability between WCO and diesel were first confirmed by a standard testing procedure.Three fuel blends, including D100, W40D60, and W60D40, were prepared by a gentle mixing with magnetic stirrer to investigate if a low strength mixing process during the multi-fuel injection could suddenly distribute the fuels molecules uniformly before the combustion.The stability tests took place by using a centrifugation at 5,000 rpm for 15 minutes and observed for cloudiness and separation of layers.The one-phase clear liquid is the most stable apparent and followed by one phase-cloudy, two-phase cloudy, and two-phase clear liquid.

Physical Properties
The specific gravities of the blends were determined using the simple method of volumetric-weight, when the flash point was measured by using Pensky-Martens closed cup method.The heating values of the fuel blends were determined using the IKA C500 automated bomb calorimeter.The pH readings were determined using a SUNTEX SP-701 pH meter, when the acid values were measured using AT-510 N automatic titrator.The ash content in the blend was quantified by heating the fuel samples in a high temperature furnace at 800°C for three hours and obtaining the difference in the weight from after and before heating.Additionally, an infrared moisture analyzer was used determined the moisture content in each of the blends.Furthermore, the combustibles were computed using the following relationship: combustibles % = 100% -(water content + ash content)%.Finally, the viscosities of resulting blends were obtained using a Brookfield DV-I+ viscometer apparatus.Chen et al., 2017; b Refer to Chang et al., 2014.Note: The results of D100 and W40D60 were re-illustrated from our previous research (Chen et al., 2017), when the data of W60D40 were done by the current study.

Chemical Compositions and Properties
The elemental analysis was done using Element Analyzer to quantify the contents of carbon (C), hydrogen (H), nitrogen (N), and sulfur (S).On the other hand, the oxygen (O) content was calculated using the equation as follows: The thermogravimetry analyses (TGA) took place to investigate the reaction temperature of WCO and diesel compounds by using a Perkin Elmer Pyris Diamond Thermogravimetry Analyzer.Air (21% O 2 and 79% N 2 ) was utilised as the oxidizer in order to simulate real combustion conditions.The volumetric flow rate of intake air was then remained by 100 NmL min -1 .The heating was non-isothermal and approximately 10 mg of fuel blends were heated from 30°C to 850°C with the rate of 10°C minute -1 .The residual weight percentages (WP) and differential thermogravimetry (DTG) analyses were finally discussed.

Sampling Strategy
As shown in Table 3, the stack flue gas samplings for PBDE were conducted within four designated chamber temperature stages, A, B, C, and D, during the IWI start-up procedures for D100, W40D60, and W60D40 cases.
Temperature Stage A (< 250°C) displayed the PBDEs concentrations before chamber temperature reached the first temperature range of higher PBDE emission, which was around 400°C (Artha et al., 2011).Furthermore, Stage B was ranged 250-450°C, representing the temperature of PCDD/F and PBDE formation resulted from de novo synthesis.Stage C (580-700°C) were selected to evaluate the impact of using waste cooking oil-based diesel blends as auxiliary fuels on the emission of PBDEs, while and Stage D (> 850°C) would indicate the emission after a stable and more complete combustion.Nevertheless, Stage D was the start of waste feeding and continuously operating temperature for the normal IWI.Consequently, the PBDE samples were collected from the IWI flue gas once per stage when using three kinds of multi-fuel injection ratios.

Flue Gas Sampling
The flue gas PBDE samples were collected isokinetically based on the standard PCDD/F sampling procedure, following U.S. EPA Modified Method 23 by using a U.S. EPA Modified Method 5 sampling train.One stack flue gas sample was collected, and the collection time for each sample lasted for approximately 0.5-2.5 hours.The sampled volumes were normalized to the standard condition, 1 atm and 273 K, with the unit of Nm 3 .Each PBDE sample was composed of particulate and gaseous phases.The particulate PBDEs were collected on a quartz filer filter, while the gaseous congeners were captured by a series of two glass cartridges packed with polyurethane foam (PUF).

PBDE Analyses
The PBDE analyses refer to the standard method of PCDD/Fs published by Taiwan EPA (NIEA A810.13B).Before the pre-treatment processes, each PUF sample was spiked with certain amount of 13 C-labeled internal standard (IS) solutions (as shown in Table S1), to record the sample recovery within whole analytical procedure.The spiked samples were extracted by toluene using Soxhlet extractor for 24 h.The extract was first concentrated by rotary evaporation and a nitrogen (N 2 , purity > 99.99 wt.%) gas stream purging system to less than 2 mL.The cleanup processes then took place with a series of packing columns.The first one is an acid silica gel column eluted with 20 mL n-hexane.The eluate was moved to a following alumina column, which was then eluted again with the order of 10 mL of n-hexane and 25 mL of dichloromethane (DCM) /hexane (40/60, in volume).The third cleanup unit was an active carbon column, when the elution was conducted with 5 mL of toluene/methanol/ethyl acetate/n-hexane (5/5/10/80, in volume) solution.Furthermore, the eluate was finally concentrated to near-dry condition using gaseous N 2 purging.The injection recovery standard (RS, 2,2',3,4,4',5'-Hexabromodiphenyl ether [ 13 C 12 ]), was added to quantify the overall sample volume by 10 µL for further instrumental analysis.

QA/QC Information
The PBDE contents in field and laboratory blank samples were at least 5 times lower than the detection limit in the current study.The relative retention time (RRT) of each congener just matched (± 0.5%) that of corresponding standards.The ratios of at least two characteristic ions are consistent with theoretical values (± 15%).The limit of detection (LOD) was defined as which the signal/noise (S/N) ratio was higher than 3; Meanwhile, the limit of quantification (LOQ) was identified as the mass which had relative signal with > 10 S/N ratio.Calibration mixtures that contained isotopically labelled ISs were used in the PBDE quantifications with relative response factor, RRF Sample .Nevertheless, these ISs could also be calibrated by RSs and derive RRF IS .Finally, both recoveries of samples and IS were ranged 70-130%, satisfying the criteria of U.S. EPA Modified Methods 23.

Fuel Properties Physical Properties
The major properties of testing fuels were listed in Table 2.For the basic physical properties, the waste cooking oil (WCO) had 12.9% higher specific gravity, lower pH, and about 2.0 wt.% of water content than those of pure diesel (D100), which were also reported in the previous studies (Predojević, 2008;Fajriutami et al., 2013).The significant high combustible contents (> 99%) and low water and ash contents pointed out that the WCO had high potential to be a derivative fuel instead of waste.However, the inconsistent (with D100) properties, especially water content, might lead to unsteady mixture of multi-fuels.Fortunately, the stability test results shown that both W40D60 and W60D40 blends were stayed in one-phase clear liquids, ensuring that the fuel homogeneity would not strongly affect the further combustion operation.Nevertheless, the small amount of water was reported to enhance the combustion and provide the benefits for engine performance and emission.Additionally, the flamibility would decrease by WCO additions, since the flash points increased 10.7% and 33.3% from D100 to W40D60 and W60D40, respectively.The above reduction would not result in seriously incomplete combustion, instead, the longer ignition delay before combustion would provide higher premixed fuel-air mass flow and had higher peak heat release rate after a short delay.The NO x emission per equivalent input energy was expected to slightly increased.On the other hand, the heating values of W40D60 and W60D40 were found 5.14% and 7.54% lower, respectively, than that of D100, meaning that more fuel injection was required to approach the same heating-up efficiency.The lower heating value could also balance the increasing NO x emission tendency aforementioned, since the Zelditch effect of thermal NO x formation could be inhibited by lower combustion temperature.The above finding could provide a higher capacity of treating more WCO from the market.The storage could be one of the problems that WCO should face to, since the acid value were obviously increased by adding WCO into diesel and all over the preferable diesel fuel storage condition (< 1 mg KOH g -1 ).Additionally, the WCO, W40D60, and W60D40 had apparently 2.2 and 3.1 times higher viscosities than that of D100 at the room temperature (as shown in Table 2).The higher viscosity would lead to worse fluidity for transfer and dispersivity of fuel spray, which further increased the operation cost and lowered the combustion efficiency, respectively.Fortunately, the viscosity differences between WCO blends and D100 could be narrowed down with the increasing temperature of fuels.Therefore, the fuel preheating (50-70°C) before injection was suggested.

Chemical Compositions
For the chemical compositions of testing fuels, the carbon (C) and hydrogen (H) contents were reduced with the WCO ratios, resulting in the heating value reduction.Additionally, the C/H ratios of W40D60 (6.5) and W60D40 (6.5) were both slightly higher than D100 (6.3) and could lead to more CO 2 emission.The overall life-cycle analysis was suggested to be investigated in the following research.Notably, the fuel oxygen content of W40D60 and W60D40 were 5.4 and 7.6, respectively, which were significantly higher than that of D100 (1.0%).The fuel oxygen was reported to improve the oxidation reaction in our previous studies and reduce PM, CO, HC, and PAH emissions (Lin et al., 2010(Lin et al., , 2012;;Chang et al., 2013Chang et al., , 2014)).Additionally, the chlorine, sulfur, and aromatic contents could also affect several emissions, such as PCDD/Fs, PCBs, PCDEs, PAHs, PM, SO 2 , and HCl.The increases of Cl and S content might slightly increase the SO 2 , and HCl emission.Nevertheless, the existence of PAHs could be the precursors and form PBDEs and other persistant organic pollutants (POPs) PCDEs at certain temperature condition and presences of Br and Cl.Therefore, the lower aromatic contents provided by WCO might inhibit the PAH formation from incompletely combustion and further reduced the PBDE and POP emissions, especially during the unsteady start-up operation.The toxicity of PCDD/F and PCB emissions were found 86% and 94% reduced, respectively, by using WCO-diesel co-combustion in our previous study (Chen et al., 2017).

Thermal Reaction Properties
TGA and DTG results of D100 and W40D60 were reillustrated from our previous research (Chen et al., 2017) and compared with the data of W60D40, done by the current study (as shown in Fig. 1).D100 start to react with air from 100°C to 220°C.The reaction might include adsorption/desorption, dehydration, evaporation, and oxidation.D100 was composed of short chain hydrocarbon structures (C 10 -C 25 ), which were thermodynamically steady, and leaded to more rapid and concentrated reaction at about 200°C.In the DTG result of W40D60, diesel degradation peak still appeared but wider (100-290°C), which could be resulted from that the complicated content of WCO blend.The WCO might be composed of free fatty acid, long chain or polymerized derivative hydrocarbons (Fajriutami et al., 2013;Ho et al., 2014), which had better thermal stability and retarded the evaporation and oxidation of D100.Additionally, there were extra two mass degradation peaks occurred at 300-380°C and 400-480°C, which could be resulted from the double-bound broken of monosaturated fatty acids and decomposition of saturated fatty acids (Santos et al., 2004).The intensity of diesel degradation was less than the other two decomposition reaction in DTG diagram of W60D40, because there were more than 50% fuel mass contributed by WCO and have better thermal stability.Therefore, the WCO should not be injected as the supporting fuel until the temperature of combustion chamber reach 500°C during the start-up procedure.The WCO was injected at 550°C chamber temperature in the real operation of IWI in the current research.

PBDE Emissions in the Flue Gases
The levels of PBDEs in the flue gases upon three multifuel combustion cases were listed in Table 4.During Stage A and B, the total PBDE concentrations in each sample 1 and 2 (S1 and S2) of D100, W40D60, and W60D40 cases were 54.2, 25.8, and 39.3 ng Nm -3 and 9.43, 2.32, and 4.64 ng Nm -3 , respectively.The D100 was the only energy source for the above two stages of start-up operation.The differences of PBDE emissions among three cases could be varied by the different residual PBDE levels adsorbed on the bottom ash and inner surface of combustion chamber.During Stage C and D, the respective levels of the total PBDE concentrations in each sample 3 and 4 (S3 and S4) were 7.31, 3.96, and 7.21 ng Nm -3 and 3.03, 5.03, and 7.38 ng Nm -3 .The highest total PBDE concentrations was recorded with D100 at Stage A (54.2 ng Nm -3 ), whereas the lowest one was found with W40D60 at Stage B (2.32 ng Nm -3 ).The PBDE emission concentrations in all cases started with the highest value of S1 and significantly drop to S2 and remain in the lower steady levels of S3 and S4, which were only 6-19% of the peak values.
The mass levels of individual PBDEs (%) and the corresponding congener subcooled liquid vapor pressures (P L ) were shown in Figs.2(A) to 4(A) (marked with black circles).The observed highly brominated congeners were all dominated by BDE-209, followed by BDE-206 or BDE-207 as the second or third dominant one in D100, W40D60, and W60D40 (Figs. 2(B) to 4(B)).Additionally, BDE-208 was the fourth dominated congener for all three fuel combinations.Additionally, the mass fractions of PBDE congeners had the opposite trend with their P L , indicating the major contribution of PBDE might be from the residues in the ashes and on the inner wall of combustion system.These findings were consistent with the observation of the previous studies that the nona-and deca-BDEs were dominated by BDE-209 based on a profile typical of the deca-BDE technical product, which BDE-209 was followed by three nona-BDE congeners (i.e.,   (Wyrzykowska et al., 2008;Wyrzykowska-Ceradini et al., 2011). Tu et al. (2011) also reported the dominant congener found from two municipal solid waste incinerators was BDE-209, accounting for 65.9% and 77.7% of total PBDE levels in stack flue gases, respectively.The second and third highest congeners were BDE-206 (6.01-6.36%)and BDE-207 (5.93-5.02%),respectively, among all PBDEs.A comparable trend was also found where the most dominant congeners were  in the stack flue gases of the reducing furnace and submerged arc furnace (SAF), suggesting the related combustion processes favored the formation of highly brominated congeners (Liao et al., 2012).
As shown in Figs.2(C) to 4(C), BDE-47 and BDE-99 were the predominant congeners for the moderately brominated congeners.Furthermore, tetra-BDEs and penta-BDEs are approximately equal in their levels in S1 and S2, with BDE-47 and BDE-99 being the most abundant congeners, which is characteristic of the penta-BDE technical product (La Guardia et al., 2006;Wyrzykowska-Ceradini et al., 2011).In studying more than 40 PBDE congeners, Wyrzykowska et al. (2011) found that the penta-and octa-BDE congeners (BDE-47, BDE-99, BDE-100,  in technical formulations are the major PBDE species in raw flue gases.They hypothesized that the highest level of PBDE congeners in the flue gas are probably products from the thermal desorption of wastes containing brominated flame retardants.Figs.2(C), 3(C), and 4(C) showed a pattern of mass contributions of lower brominated congeners that is consistent with the congener vapor pressures.Nevertheless, the mass fraction of moderately brominated congeners within S1, S2, and S3 were increasing as shown in Figs.2(C) to 4(C), while the higher brominated ones tended to reduced their contribution,  especially BDE-209.The above observation might be resulted from the debromination of BDE-209 in to lower brominated congeners, which could be supported by the previous research.Wyrzykowska-Ceradini et al. (2011) reported an enrichment of lower brominated congeners in the flue gas samples, that are minor or not present in the PBDE technical mixtures exhibited which suggesting that debromination occurs.
For more clearly analysing the effects of different fuel provided on the emissions of total PBDEs, the variation among cases should narrowed down.The concentrations of S1 to S4 in each case were normalized by being divided with S1, producing S2/S1, S3/S1, and S4/S1 ratios as nondimensional factors (as shown in Table 4).Total PBDE concentrations for all three start-up cases reached the highest values in S1 and significantly declined to only 9.0-  17.4% of S1, which were measured in S2.The elevated levels during Stage A were probably resulted from the residue in combustion chamber, which is consistent with the finding of the previous studies (Wyrzykowska et al., 2008;Chang et al., 2013).The PBDEs would start to be released from the combustion residues, when the melting points (64-306°C) of congeners were approached during S1 (101-236°C) and S2 (226-450°C).Artha et al. (2011) also indicated the first temperature with higher PBDE emission during fly ash heating-up processes was around 400°C.On the other hand, the decomposition temperature of each PBDE congener was increased with the substitution of bromine and up to 425°C of BDE-209, which was the dominant species of PBDE emissions.Therefore, the decomposition reactions of BDE-209 happened in Stage B apparently reduced the total PBDE emission in S2.The emission concentration of S2 were the competitive result from the residue releasing and thermal decomposition of PBDEs.
The injection of WCO started from Stage C, which could demonstrate the PBDE effect with S3.The S3/S1 (0.13) was decreased from S2/S1 (0.17) by using D100, when the opposite trend was found by utilizing both W40D60 (0.09→0.15) and W60D40 (0.12→0.18).The above results indicated that the PBDE formation by de novo synthesis within 600-800°C (Artha et al., 2011) was not observed in D100 combustion but occurred in two cases of WCO injection.This phenomenon could be resulted from the higher viscosity of WCO spray retarded the combustion ignition.During the ignition delay, the liquid WCO layer could cover the surfaces of inorganic ash or soot produced by primary combustion of diesel, and provide sufficient amount of hydrocarbon to the pyrolysis reaction.The C 2 H 2 could be further transformed from the WCO compounds and become the precursor of PBDEs nearby the particulate catalyst (ash and soot) and enhanced the de novo synthesis.Therefore, the increases of PBDEs in S3/S1 could be found along with more WCO injection.The Stage D showed more steady emissions than the others by using all kinds of fuel combinations.The PBDE emission concentration by combusting D100 was found to had a relatively lower S4/S1 ratio as the normal continuous operation.However, the S4/S1 emission ratios by using W40D60 and W60D40 were both remained as 0.19, showing the reformations and emissions of PBDEs were still undergone.Therefore, the control of viscosity of WCO before injection was important to improve the PBDE emission during start-up operation.

Emission Rates and Annul Emissions for PBDEs by Using Various Fuels
The trends of the emission rates (ERs) of the total PBDEs emitted in the flue gases of four samples (in green bars) are illustrated in Fig. 5.The emission rate show the extremely high levels by 281, 137, and 206 µg hr -1 in Stage A within the first 2 hours after started-up.Furthermore, the ERs of total PBDEs for all three fuels dropped significantly to 9.4, 2.3, and 4.6 µg hr -1 in Stage B by the thermal decompositions of congeners.Notably, the changed of PBDE emission rates by using two combinations of WCO injections were different from pure D100 during Stage C and D. The ERs of PBDE reduced to 7.3 µg hr -1 from 9.4 µg hr -1 by using D100 but increased to 4.0 and 7.2 µg hr -1 from 2.3 and 4.6 µg hr -1 by injecting 40% and 60% WCO, respectively.We speculated that the WCO having a higher viscosity than diesel, after it was injected at Stage C for W40D60 and W60 D40 may form a blanket on the residue and physically trapped PBDEs causing both lower and higher brominated species to have low enrichment levels at both Stage C and D.
In other studies, Chen et al. (2006) found the PBDE concentrations in the stack flue gases of the incinerators were two to three times higher than the atmospheric PBDE concentrations in other study areas suggesting that incinerators with high PBDE levels and large flue-gas flow rates are important PBDE emitters to the atmosphere (Wyrzykowska et al., 2008;Wang et al., 2010a).Tu et al. (2011) also reported that the highest amount of PBDE mass discharge were observed from their stack flue gases, except in bottom residues of MSWIs' combustion chambers.Therefore, the annual effect of the start-up emissions during the current study were further calculated and discussed.
The annual emissions of pollutants are generally calculated from the measurements made during normal operations; however, the total emissions might be underestimated if those from start-up operation were ignored.This study estimate the annual PBDE emissions, including 0 to 12 times start-up cases.The contribution of start-up to annual emissions were further estimated using the following equation, as well as compared the differences among three various fuel combinations.

 
when "E ST " stands for the emissions per start-up (µg), "N ST " is the number of start-up operations in a year, and "NOR" is the emissions under normal operations with 0-12 times of start-up.The integrated PBDE emissions during one start-up procedure were 1,099, 1,253, and 1,207 µg by using D100, W40D60, and W60D40, respectively.The emission of the normal operation (without star-up) was calculated by multiplying the ER of 12 hours after D100 startup operation (24.1 µg hr -1 ) with 8,760 hr yr -1 and resulted in 211,445 µg yr -1 .Furthermore, the "start-up hours" were calculated by N ST multiplying 12 hrs, which was assumed for one start-up operation in this study, and subtracted from 8,760 hr yr -1 to derive the annul normal operation time (T NOR ).Therefore, "NOR" could be evaluated by the 24.1 µg hr -1 multiplying T NOR .Nevertheless, the star-up emissions of two WCO injection cases were nondimensionalized by that all the concentrations during W40D60 or W60D40 start-up procedure multiplying a ratio of S1 D100 /S1 W40D60 or S1 D100 /S1 W60D40 to normalize the peak emission and exclude the uncontrollable residual effect.Therefore, the annul emissions and start-up contribution among using different fuels then became comparable.The estimated annual emissions and start-up contributions are listed in Table 5.The PBDE emissions for one start-up by using D100 contributed 0.38%, when those were 0.46% and 0.43% by using W40D60 and W60D40, respectively.The contribution seems not so significant in only one starup; however, the normal IWI operation needed to be frequently restarted per month, since the waste composition and characteristics varied significantly and should be carefully control by tuning the combustion system.Moreover, the start-up emissions might become much more important as the number of annual start-up times increases to twelve,  The disposal of waste cooking oil is one of the major waste treatment problem.The WCO was first used as the feedstock to produce biodiesel and further 2% blended with pure diesel to form the conventional "super diesel", which was a multi-component-fuel (MCF) in Taiwan before May, 2015.However, the above MCF was found that their higher viscosity tended to inhibit the performance of nebulization.Additionally, the unsaturated fatty acid contained in MCF might increase the potential of microorganic growth and further leaded to some engine operation problem, such as misfire, unstable power output, and blocking of fuel injection and storage system.Therefore, the WCO cannot be utilized for biodiesel-diesel blends production in Taiwan now and a certain amount (about 40%) of them are exported to Korea for biodiesel production, this might result for more transportation cost, extra energy demand and emissions.On the other hand, the recovery of over 60,000 tone WCO per year could be approached according to the current result with only a slightly increase of PBDEs by being utilized with an incineration process.The circulation economics were further be approached and reduce the overall pollutant and carbon emissions.

CONCLUSION
The total PBDE level was highest in Stage A (< 250°C) and declined significantly in Stage B (< 250-450°C) by using diesel as fuel during start-up operation of a IWI in this study.The reduction of total PBDE was a competitive result between residue releasing and thermal decomposition in Stage B. The debromination of BDE-209 was found to increase the mass contributions of lower brominated congeners in lower temperature conditions.The WCO were found to slightly increased the PBDE emissions during the following Stage C (580-700°C) and D (>850°C), which provided the suitable temperature for de novo synthesis of PBDE formation (600-800°C).The viscous WCO might acted as a cover and provide sufficient unburn hydrocarbon to enhance the aforementioned reaction and extend the start-up emission effect, which could not be found in D100 case in this study.The viscosity became an important control factor when the WCO were utilized as an alternative fuel in IWI operation.The accumulated PBDE emissions during the start-up procedure were 1,099, 1,253, and 1,207 µg by using D100, W40D60, and W60D40, respectively.If the frequency of IWI restarting was once per month, the annual PBDE emissions contributed by start-up procedures would increase up to 4.60%, 5.47%, and 5.20% by three fuel combinations, respectively, and became a noticeable issue.Therefore, avoiding unnecessary start-ups was an essential criterion for IWI operation.The not significant increases (< 1%) of PBDE emissions by altering 40% and 60% diesel with WCO provided a positive information for WCO treatment.This additional benefit from waste oil disposal also created a good demonstration of Circular Economy.Furthermore, the overall life-cycle analysis was suggested to be investigated in the following research.

Fig. 2 .
Fig. 2. Profiles of PBDE congeners in the stack flue gases of D100 and their vapor pressures.

Fig. 3 .
Fig. 3. Profiles of PBDE congeners in the stack flue gases of W40D60 and their vapor pressures.

Fig. 4 .
Fig. 4. Profiles of PBDE congeners in the stack flue gases of W60D40 and their vapor pressures.

Table 1 .
Basic information of Laboratory Waste Incinerator

Table 2 .
Physical properties and chemical compositions of testing fuels a Refer to

Table 3 .
Temperatures in combustion chamber during each sampling and start-up process "Time" represents the time (in hour) spent after the incinerator star-up.2. "Temp."represents the real temperatures of combustion chamber.3. WCO was injected in case W40D60 and W60D40 when the Temp.reached 550°C before Temp.Stage C.

Table 4 .
Ratios of the PBDE concentrations for the 3 fuels at 4 sampling stages