Atomic emission spectroscopy for the on-line monitoring of incineration processes

Spectrochimica Acta Part B 58(2003) 823–836

0584-8547/03/$ - see front matter� 2003 Elsevier Science B.V. All rights reserved.PII: S0584-8547Ž03.00017-X

Atomic emission spectroscopy for the on-line monitoring ofincineration processes

E.A.H. Timmermans , F.P.J. de Groote , J. Jonkers , A. Gamero , A. Sola ,a a a b b

J.J.A.M. van der Mullen *a,

Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlandsa

Departamento de Fısica Aplicada, Universidad de Cordoba, 14071 Cordoba, Spainb ´ ´

Received 15 October 2002; accepted 13 February 2003

Abstract

A diagnostic measurement system based on atomic emission spectroscopy has been developed for the purpose ofon-line monitoring of hazardous elements in industrial combustion gases. The aim was to construct a setup with ahigh durability for rough and variable experimental conditions, e.g. a strongly fluctuating gas composition, a high gastemperature and the presence of fly ash and corrosive effluents. Since the setup is primarily intended for the analysisof combustion gases with extremely high concentrations of pollutants, not much effort has been made to achieve lowdetection limits. It was found that an inductively coupled argon plasma was too sensitive to molecular gas introduction.Therefore, a microwave induced plasma torch, compromising both the demands of a high durability and an effectiveevaporation and excitation of the analyte was used as excitation source. The analysis system has been installed at anindustrial hazardous waste incinerator and successfully tested on combustion gases present above the incinerationprocess. Abundant elements as zinc, lead and sodium could be easily monitored.� 2003 Elsevier Science B.V. All rights reserved.

Keywords: Microwave induced plasma; Atomic emission spectroscopy; Continuous emission monitoring; On-line monitoring;Incineration; Inductively coupled plasma; Combustion gas analysis

1. Introduction, the aims of combustion gasanalysis

During the last decade worldwide several groupshave been studying the feasibility of atomic emis-sion spectroscopy(AES) for the on-line analysisof flue gases. As far as we know, these investiga-

*Corresponding author. Tel.:q31-40-2474043; fax:q31-40-2456050.

E-mail address:[email protected](J.J.A.M. van der Mullen).

tions are without exceptions focused on off-streamexhaust gases, i.e. relatively ‘clean and cold’ end-of-pipe gases just before their ejection into theatmosphere takes place. Main reason for the explo-ration of this research field is the expectation thatwithin a few years European and American legis-lation will enforce continuous emission monitoringof flue gases for several industries, e.g. powerplants, cement industries and waste incineratorsw1x.Clear progress has been made with inductively

824 E.A.H. Timmermans et al. / Spectrochimica Acta Part B 58 (2003) 823–836

Fig. 1. Schematic drawing of DTO9 at AVR-Chemie. Measurements have been performed above the combustion in the secondarycombustion chamber, marked with an asterisk.

coupled plasma(ICP)-AES, in which an ICP hasbeen used as the excitation source of the flue gasesw1–5x and in 1997, the first systems for quantita-tive continuous emission monitoring based on ICP-AES became commercially availablew3x. Otherresearch groups focused their attention onto laser-induced breakdown spectroscopy(LIBS), in whicha laser is directly focused into the off-stream gasflow creating discharges inside the stackw6,7x.Although gas sampling, a critical stage when ICP-AES is used, can be avoided with the LIBStechnique, serious problems have been encounteredwith the calibration due to the very large gradientsin the laser-induced plasma and the presence offly ashes. A third approach to analyze stack gasesis based on non-resonance atomic fluorescence(AF) w8x. As for ICP-AES, remote sensing isapplied. Basically, LIBS gives insight in the ele-mental composition of particulates, AF providesinformation on free atoms, whereas ICP-AES sup-plies information on free atoms as well as partic-ulates(provided that they do not precipitate duringextraction and that they are vaporized within theplasma).Apart from legislation purposes, on-line moni-

toring can also be used to control the combustion

process. Opposite to the groups mentioned, ouraim is to investigate whether it is possible todevelop analytical equipment that can be used forprocess control. Therefore, a collaboration withAVR-Chemie, a large industrial hazardous wasteincinerator in Rozenburg(The Netherlands) wasstarted, with the aim of on-line monitoring ofheavy metals present in the combustion gases justabove the flames of the incineration process itself,i.e. in gases that have not been cleaned yet.The process of waste incineration at AVR-

Chemie is sketched in Fig. 1. Measurements shouldbe performed on the combustion gases in the so-called secondary combustion chamber, which hasthe function of incinerating hazardous waste thathas not been completely burnt inside a slowlyrotary kiln. The solid slag remaining after thecombustion process inside the after burning-cham-ber is removed through various outlets, whereasthe combustion gases undergo several cleaningstages, before they are ejected into the atmosphere,cf. Fig. 1.A potential application of on-line information

on temporal fluctuations of metals in the combus-tion gases present in the secondary combustionchamber could be the adjustment of the incinera-

825E.A.H. Timmermans et al. / Spectrochimica Acta Part B 58 (2003) 823–836

tion composition, in order to optimize the combus-tion process or to minimize peak loads of metalflows. This could probably increase the lifetime ofthe equipment used in the cleaning stages thatfollow after the two-stage incinerationw9x.The analysis of non-cleaned gases implies that

the merits of the previous groups—achieving lowdetection limits and an accurate quantification ofconcentrations—are less important in our case,whereas flexibility, the withstand against changingexperimental conditions and the presence of acidsand dust particles are of much more importancew9x. As a consequence, it is not straightforward touse commercially available ICP-AES equipment,since this is developed and optimized for theanalysis of relatively clean off-stream gases. Thecombustion gases above the incineration processinside the secondary combustion chamber containhigh levels of pollutants and measurement condi-tions are difficult. Especially due to the hightemperature of the combustion gases(;1200 8C)and the presence of fly ashes and corrosive acidcompounds great care has to be given to thematerials used in the analysis system.Since, to our knowledge, no methods are in use

for monitoring metals inside the incineration ovenitself, it is obvious that new developments in thisarea are of great importance. It is the aim of thisstudy to find strategies to speed up these devel-opments. In Section 2 the used experimental setupwill be given while in Section 3 results obtainedat AVR-Chemie will be presented. Conclusionswill be given in Section 4.

2. The experimental setup

Three largely separated steps can be distin-guished in AES in the case of combustion gasanalysis:

1. Gas sampling, i.e. the extraction of the combus-tion gases out of the waste incinerator andtransport towards the excitation source.

2. Dissociation and the excitation of the combus-tion gases by an excitation source, e.g. a laseror plasma.

3. Acquisition and analysis of the emission spectra.

These three steps will be discussed separately.Since the excitation source plays a central role in

AES, the applied plasma source will be discussedat first. The data-acquisition and gas sampling willbe treated afterwards.

2.1. Limitations of the ICP as excitation sourcefor molecular gases

In analytical chemistry the ICP is by far themost popular excitation source when aqueous sam-ples have to be analyzedw10,11x. CommercialICP-AES systems usually are completely automat-ed and optimized for the analysis of these samples.As experimental settings(as gas flows) can bevaried only to a limited extent, and since it islikely that for gas analysis other experimentalconditions are required, the use of these systemsis not straightforward. For this reason, our initialefforts to analyze gases were done with a non-commercially available 50-MHz ICP made byPhilips w11x, requiring fully manual control.

Within a certain range, increasing the flow ofcombustion gases into the argon-ICP will result inan increase of the emission signals of the presentspecies and thus improve the signal to noise ratios.It, therefore, is important that a considerableamount of combustion gases can be introducedinto the plasma. However, it is well known thatmolecular gases can have a large influence onnoble gas discharges and that the excitation powerof the plasma might be severely alteredw10,11x.When too much molecular gas is introduced, theplasma might even extinguish. Therefore, an exten-sive study on the effects of molecular gases on theICP is needed.The effects of incinerator gases on the excitation

power of the argon ICP are simulated by simulta-neously introducing air and an aqueous analyte.The analyte (30-ppm zinc in 10% HNO) is3

introduced through the central channel of an ICPFassel torchw12x with the use of an argon-fedcross-flow nebulizer(typical aerosol injection rate:12 ml h ). Additional air, being a simulant fory1

the combustion gases, has been added to theargonyaerosol flow downstream of the nebulizer,just before entering the central channel in the ICPtorch. It should be noted that only the air flowQ is varied during the experiments and that theair


Fig. 2. The PDR(a) and the deduced detection limit DL(b) for an atomic zinc line as function of the air flowQ introduced intoair

the argon ICP.

amount of Zn introduced into the plasma is keptat a constant rate.Results are shown in Fig. 2, in which the

dimensionless quantity ‘peak intensity to detectionlimit ratio’ (PDR) is used. It is defined as

c I(c)PDRs s (1)

c z sL A B

with c being the concentration of Zn in thesimulated incinerator gas(in mg m ), c they3

L

detection limit of Zn(in mg m ), I(c) the meas-y3

ured Zn line intensity,z an arbitrary constantA

(taken as 2 in our case) ands the noise in theB

measured background. In Fig. 2a the PDR of themost intense emission line of zinc(at 213.86 nm)is given as a function of the amount of air addedto an argon plasma with a RF power supply of 2.0kW. It is found that the introduction of more than0.21 l min of air leads to strong decrease of they1

PDR. This corresponds to the experimental find-ings that forQ )0.21 l min the intensity ofy1

air

the Zn line drops sharply, whereas the backgroundnoises remains more or less constant. Apparent-B

ly, the excitation power of the plasma is severelydisturbed ifQ )0.21 l min .y1

air

From Eq.(1) it can be seen that if the PDR andthe concentrationc are known, the theoreticaldetection limit of Zn in air(or combustion gases)can be calculated. The concentrationc can simplybe deduced from the known mass flow of zinc andthe known volume flow of air into the plasma.Results are given in Fig. 2b and obtained usingthe PDR values of Fig. 2a. Fig. 2b once moreshows that best detection limits are achieved if0.21 l min of combustion gases would bey1

introduced.The measurements show that air introduction

has a substantial impact on the excitation powerof the argon ICP. Assuming that combustion gaseshave the same effect on the excitation power ofthe plasma as air, it is obvious that the combustiongases can only constitute a little fraction of thetotal plasma gas(maximally 0.21 l min of com-y1

bustion gases vs.)25 l min of argon). As ay1

result, concentrations of several trace elements caneasily become too low for detection.For the detection of some elements in combus-

tion gases a higher gas flow might be needed inorder to ensure that sufficient atoms are present inthe plasma. Moreover, due to the risk of conden-sation of the hot combustion gases and the pres-


ence of particulates, which would easily obstructcapillary gas handling systems, it is very difficultto control small gas flows. Apart from the flowcapacity control problems, it can be expected thatthe composition of combustion gases fluctuatesstrongly with respect to the molecular species,particulates and acid compounds. Especially, forplasmas that are sensitive for molecular injection,this might strongly affect the excitation power asa function of time. All these factors suggest that aplasma is required, which is less sensitive tomolecular injection and in which significantlymore than 0.21 l min of molecular gases can bey1

introduced.

2.2. The ‘torche a injection axiale’`

As discussed in Section 2.1, the investigatedargon ICP shows several shortcomings as excita-tion source for combustion gases. In literature,several ICPs have been reported, which, unlike theICP we studied, have a very high resistance againstmolecular gases or can even be operated on puremolecular gases. However, these plasmas usuallyrequire a high power input()3 kW) and conse-quently a large and heavy setup. Because a com-pact and modular setup is desired for themeasurements at the secondary combustion cham-ber of AVR-Chemie, we have looked for a plasmasource in the category of microwave inducedplasmas(MIPs).MIPs have already been used several times in

analytical chemistryw13x. Besides the well-knownsurfatronw14–16x and Beenakker cavityw17x, sev-eral other field shaping applicators have been usedas well w18,19x. However, due to their limitedoperating powers(-400 W), these plasmas showa strong sensitivity to water introduction and con-sequently are less popular for the analysis ofaqueous samples than the ICP.Recent developments in the field of MIP tech-

nology, however, have resulted in new compactplasma sources that have overcome these powerrestrictions and can be operated at power levelsup to several kilowatts. One of these new MIPsources is the so-called ‘torche a injection axiale’`(TIA) w20,21x, producing needle-like plasmas thatseem very suitable for AES. Due to its design, the

TIA is rather insensitive to impedance changesand therefore fulfills the demand of a high resis-tance to molecular gasesw22x. Dependent on thegeometry of the TIA-nozzle even pure moleculargases can be used as plasma gasw23x. This is oneof the main reasons why this plasma source hasbeen used in the measurements at AVR-Chemie.The TIA is based on the original design from

Moisan et al., who also studied the tuning andmatching properties and characteristics of the torchwhen creating atmospheric helium or argon plas-mas w20x. Remarkable is the large power rangeover which the TIA can be operated: from 150 Wup to at least 5 kW. The plasma has a diameter oftypically 1–2 mm and a height of 6 cm(at amicrowave power of 1 kW and 5 l min of argony1

gas supply). Typical maximum electron densitiesand temperatures in an argon plasma at theseconditions are 2=10 m and 2=10 K, respec-21 y3 4

tively, while the maximum heavy particle temper-ature is approximately 5=10 K w24,25x.3

The TIA and the accompanying waveguidestructures are shown in Fig. 3. Microwave radiationwith a frequency of 2.45 GHz is introduced into arectangular waveguide structure(WR-340, innerdimensions 86.4=43.2 mm) by means of an2

antenna and propagates inside the waveguide inthe dominant TE mode. Inside the TIA-structure,01

this wave propagation mode is converted into thecoaxial TEM mode. The microwave energy can befocused in the vicinity of the gap between thenozzle and the outer cover. A plasma breakdowncan be created above the nozzle at sufficientlyhigh power densities after supplying a gas flow.The power supply(Muegge, MW-GIR2M130-2K)is limited to 2 kW maximally.Two different nozzle designs have been used to

control the gas flow and field shaping propertiesw21,22x, cf. Fig. 3. When using a copper nozzlewith a central aperture for gas injection(CA-nozzle), only plasmas can be created if noblecarrier gases are used, e.g. argon or helium. How-ever, when the plasma is created on top of a nozzlewith a central tungsten tip(TT-nozzle), purelymolecular gases, like air or flue gases, can be usedas well. A limitation of the TT-nozzle is that themaximum applied power is approximately 1 kWas at higher power levels the tungsten tip starts to


Fig. 3. The microwave setup with the TIA in central position. On the right the two different applied nozzles are depicted.

Fig. 4. Intensity of a zinc line(introduced as aerosol)measuredfrom an argon discharge created with the CA nozzle as a func-tion of the additionally introduced amount of air. Both plasmasexpanding into open air and plasmas expanding into an argonenvironment are studied.

erode. Typical diameters of the gas channels are1–2 mm for the CA-nozzle and 0.5–1 mm for theTT-nozzle.In order to investigate the influence of molecules

on the detection limit we applied the same methodas described in Section 2.1 for the ICP. In Fig. 4the intensity of a zinc line is shown from an argonplasma (created with the CA-nozzle), with Znanalyte and air additives as a function of theamount of introduced air. If we assume that theair addition has no influence on the noise in thespectral background(s ), we can compare theB

results with Fig. 2a, giving the PDR for theinvestigated argon ICP. It is obvious that in caseof the TIA plasma much more air can be intro-duced as in the ICP, both in absolute as in relativesense (1.5 vs. 0.2 l min and 30 vs. 0.8%,y1

respectively).A remarkable difference can be seen between

TIA plasmas expanding into air and plasmasexpanding into an argon environment(controlledby a large vessel). For plasmas expanding into anair environment, we observed a steady decrease ofthe Zn emission as a function of the amount of airintroduced with the argon carrier gas. However,for plasmas expanding into an argon environment,the emission signal remained more or less constantfor low airflow rates and dropped sharply if the

amount of added air exceeded 1.5 l min . Thisy1

shows some resemblance with the case of the ICPexpanding into air(cf. Fig. 2a), in which theemission intensity also showed a sharp drop. Theresemblance is probably due to an effective shield-


Fig. 5. The optical system. The plasma is focused onto the entrance slit of the monochromator using a parabolic mirror. Both thegrating inside the monochromator and the CCD-camera are computer-controlled.

ing of the ICP from the air environment by theouter argon gas flow.

2.3. The optical setup

The plasma is imaged onto the entrance slit ofthe compact crossed Czerny-Turner monochroma-tor (Thermo Jarell Ash, Monospec 18, focal length15.6 cm, 2400 g mm grating blazed at 260 nm)y1

with the use of a parabolic mirror, see Fig. 5.Using a mirror ensures a large solid angle andavoids problems that the use of lenses would entail(like a wavelength-dependent refractive index).The dispersed emission signal is focused onto

an UV-enhanced CCD-camera(Santa BarbaraInstrument Group, ST6-UV), used in the spectro-scopic mode. In this mode the camera has an arrayof 750 pixels, each 11.5mm=6.53 mm in size.The resolution of the optics is approximately 0.016nm pixel at 210 nm, ample to identify atomicy1

lines and molecular bands. At this resolution, thus,approximately 12 nm is depicted simultaneouslyon the CCD, whereas at 400 nm approximately 16nm is depicted simultaneously. As a result, onlyapproximately 15 measurements are necessary toscan the most relevant wavelength interval(200–400 nm). Apart from the compact size and therelatively low costs, another major advantage ofthe used system is its robustness: re-calibrationafter transport is hardly ever required.

2.4. Combustion gas sampling

The gases are extracted from the oven andintroduced into a remote plasma. Due to the slight

under-pressure inside the oven(fy1 mbar) apump is required for gas sampling. As far as weknow, in literature only gas samplers are reportedfor end-of-pipe measurements on exhaust gas flowsof plants. The most important in these designs isthe iso-kinetic sampling, a feature that is notrelevant in the highly turbulent secondary combus-tion chamber. However, what is important for thesampling of combustion gases is that these gasesare highly corrosive. Because of the extreme hightemperatures and acidity at AVR-Chemie, we hadto find alternatives for these normal probes. Twodifferent gas sampling methods have been devel-oped and tested.The first method is based on a dilution probe

that was constructed and consists of a system oftwo parallel placed Al O tubes that can be insert-2 3

ed through an aperture in the incinerator ovenwall. The principle of the dilution probe is basedon Bernoulli’s theorem (the ‘Venturi effect’), cf.Fig. 6. The incinerator gases are diluted within anincoming argon carrier gas floww9x and a mixtureof both gases is obtained at the outlet of the probe.The dilution process will decrease the dew pointso that condensation of the combustion gases aftercooling probably can be avoided.The second method is based on the creation of

a plasma at an under-pressure. Due to the pressuregradient produced in this way, combustion gaseswill be extracted from the oven and introducedinto the plasma without the need of a pumpbetween oven and plasmaw9x. For the creation ofplasmas at an under-pressure, a chimney-like vesselhas been constructed(Fig. 7), that can be mounted


Fig. 6. The dilution probe at the small orifice ‘1’ an expansion in the argon flow is created. Since this locally creates an under-pressure, incinerator gases are sucked in from the vicinity of the gap. Subsequently, a mixture of argon with incinerator gases isdirected towards the plasma through orifice ‘2’.

Fig. 7. The plasma column used for the creation of plasmas atan under-pressure.

on top of the waveguide section of the TIA andevacuated by a vacuum pump. Since the pressureinside the vessel(between 950 and 990 mbar) iskept lower than the pressure inside the incinerator

oven, combustion gases can be automaticallyextracted from the incinerator oven and suckedtrough the nozzle of the inner conductor of theTIA. The stainless steel sampling line is heated toapproximately 1508C in order to minimize con-densation of combustion gases and precipitation ofparticulates. If desired, the combustion gases canbe diluted with argon, just before entering the TIA.With a variable pumping capacity the pressure,and consequently the gas flow, can be controlled.

3. Experimental results and discussion

3.1. General results

Both sampling methods discussed in Section 2.4have been extensively tested at the secondarycombustion chamber of AVR-Chemie.Before using the dilution probe at AVR-Chemie,

it had been tested in the laboratory in a woodfurnace in order to study the dilution ratio. Afterseveral days of testing no instabilities or visibledamage to the probe could be observed. However,at AVR-Chemie, the probe suffered from severalmajor disadvantages:

● The tip of the probe, being exposed to thecorrosive matter inside the combustion oven,was subject to a fast vitrification.

● Obstruction of the off-stream orifice of theprobe by fly ash during operation even increasedthe rate of vitrification and changed the dilutionratio continuously. As a result, it was difficultto keep the total combustion gas flow constant.

● The expectation that condensation of the com-bustion gases would be avoided due to a


decrease of the dew point was incorrect and awet aerosol was obtained at the outlet of thesecond Al O tube. Possibly, part of the heavy2 3

metals present in the combustion gases got lost(due to precipitation at the wall) in the channelbetween the incinerator oven and the plasma.

The corrosion processes reduced the lifetime ofthe dilution probe significantly. Some probes wereseriously damaged after use of only a few hours.This is, thus, in sharp contrast to the laboratorytests in a wood furnace. Apparently, it is verydifficult to simulate the extreme and corrosiveconditions at a waste incinerator plant(e.g. thepresence of fly ashes, volatile acids and otherpollutants).

Much better results are obtained with plasmascreated at an under-pressure using the chimney-like column (cf. Fig. 7) and a heated interfacebetween the incinerator wall and plasma source.This method offered the major advantages thatcondensation of combustion gases in the samplingline could indeed be avoided and that the gas flowwas stable in time and well adjustable. In the caseof ‘hot’ sampling, i.e. sampling with the use ofheating ribbons, no condensation was observedand a representative sample could be obtained.Particulates present in the combustion gases wereintroduced into the plasma without noticeablelosses and evaporated. It is doubtful if large par-ticles are also fully vaporized within the plasmaobservation windoww26,27x, so that some impre-cision and inaccuracy will come from the variationin fly ash size and boiling point. In the case of‘cold’ sampling however, i.e. sampling without aheated interface, condensation precipitated insidethe extraction line, affecting the characteristics ofthe system in time. Therefore, all presented meas-urements were done with a heated sampling line.A limitation is that plasmas created at an under-pressure with the use of the CA-nozzle(cf. Section2.2) showed instabilities if relatively substantialamounts of combustion gases were introduced(f30% at 2 kW). The use of the TT-nozzle,however, resulted in a stable plasma operation overa large range of operating conditions(Ps300–1000 W, gas flows3–6 l min ).y1

3.2. Observed elements

As discussed in the previous section, best resultswere obtained from plasmas created at an under-pressure using a TT-nozzle and the plasma column.Despite the extreme measurement conditions andthe corrosivity of the sample that had to beanalyzed, the analysis system operated stable fora long period(a few weeks), during which it hadbeen used several hours per day. Presented results,are hence, obtained from these type of discharges.A part of a typical spectrum measured from a

discharge in pure combustion gases is given inFig. 8. Many atomic lines and molecular bandscan be distinguished in the shown spectral rangebetween 340 and 425 nm. It should be noted thatthe CCD camera is over-exposed at several spectralranges due to intense molecular or atomicradiation.The elements that have been identified unam-

biguously and their corresponding strongest emis-sion lines are given in Fig. 9. The underlinedwavelength represents the line that can best beused for monitoring, taking into account itsstrength and the absence of interference with otherlines or bands. Several elements, such as oxygen,could not be detected due to the operating rangeof the monochromator(l-600 nm).The list of detected elements has been compared

to the composition of fly ash that was collected atthe waste incinerator plant as wellw9x. Althoughthe fly ash, being analyzed by a sieve analysis ina laboratory was collected during another periodand at a different location, it provided a roughimpression of the abundance of present elements.All elements detected by AES were found to beabundantly present in the fly ashes()2 mg g ).y1

Halogens, like Br, Cl and F were not detected byemission spectroscopy, although they were foundto be abundantly present in the fly ash as well. Itis well known from ICP-AES that a helium plasmais required for the detection of these elementsw10x,since their radiative levels are hardly populated inan argon or air plasma.It should be noted that, although not yet deter-

mined, detection limits for elements present incombustion gases will be rather poor compared tothe detection limits as achieved with ICP-AES for


Fig. 8. A part of a typical spectrum as measured from a discharge in pure combustion gases(at an under-pressure ofy10 mbar).The absorbed microwave power was 600 W and the measurement was done at approximately 10 mm above the TT-nozzle. Molecularbands dominate considerable parts of the shown spectral range.

Fig. 9. An overview of the elements observed at AVR-Chemie, together with their strongest emission lines(in nm). Atomic linesare labeled with ‘I’ and ionic with ‘II’. Emission lines that can best be used for monitoring are underlined. The Cu and W lines areprobably due to nozzle erosion.

aerosol analysis. Argon ICP-AES normally fea-tures a high accuracy, very low detection andlinearity over several orders of magnitude foraqueous aerosol analysis. This has its origins in

the voluminous argon plasma and a restrictedsample injection, outside of the active zone of theplasma. In the case of the TIA, the carrier gas andanalyte are mixed before entering the plasma, and


Table 1List of the observed molecules, including the wavelength ranges with their most intense emission bands

Molecule Name system Transition Spectra(nm) Relative strength

CN Violet system BS™ P2 2 335–360 Very intense373–389 Very intense410–422 Very intense450–460 Very weak

NH 3360-A system˚ A P™X S3 3 326–338 IntenseN2 First positive system BP™A S3 3 503–600 Very weaka

Second positive system CP™B P3 3 290–298 Intensea

330–500 Weaka

Nq2 First negative system BS ™X S2 q 2 q

u g 380–392 Weaka

420–428 Very weaka

NO g system AS ™X P2 q 2 190–280 WeakOH 3064-A system˚ A S™X P2 2 260–297 Very weak

306–324 Intense

Only present in air plasmas. After introduction of incinerator gases these bands disappear, possibly due to the presence of water.a

consequently, the analyte has a strong impact onthe plasma properties. Laboratory measurementshave shown that in TIA plasmas detection limitsof most heavy metals in aerosols are approximately1-ppm level, substantially worse than most ICP-AES systems. Therefore, it can be expected thatthe detection limits of the combustion gas moni-toring system, are rather poor and worse than thosefor ICPs reported by other authors who focusedon the analysis of cleaned end-of-pipe gases. Onthe other hand, their systems do not benefit froma very compact and modular setup, and moreover,very likely are not suitable for the extreme condi-tions as can be expected when monitoring incin-erator processes for process control, which is thepurpose of our work(cf. Section 1).Since only very short measurement times are

necessary for the detection of the elements listedin Fig. 9 (or even allowed in order to avoid over-exposure of the CCD-camera), the system can beused very well for continuous gas monitoring ofthose elements. The measurement time is almostentirely determined by the data-transfer from thecamera to the computer(f5 s per frame). Furtherdata processing and evaluation still requires moretime, but can be performed afterwards since thespectra are stored on the computer. However, ifdesirable, software can be developed that providesfully automated(on-line) data processing.

3.3. Molecular interference

Several molecular bands are dominantly presentin the measured spectra, as for instance can beseen in Fig. 8. The vast majority of these bandsoriginate from molecular association products thatare created inside the plasmaw23,28x. The observedmolecular transitions are listed in Table 1, includ-ing the wavelength ranges in which the mostintense emission bands appear. Molecular transi-tions labeled as ‘intense’ or ‘very intense’ createsuch a strong spectral background that the corre-sponding wavelength ranges cannot be used forAES due to the over-exposure of the CCD camera.In the spectrum of Fig. 8 it can be seen that the

intensities of the shown Al lines are a few ordersof magnitude weaker than the Ca lines and theq

CN band. Atomic lines in different wavelengthintervals with comparable weak intensities mighteasily be lost in the molecular bands, which inTable 1 are indicated as being ‘weak’ or ‘veryweak’. Therefore, it is important to reduce theintensities of these bands to a minimum. It isobserved that spectra from pure combustion gasplasmas have a very high molecular background.This background can be reduced by the additionof a considerable amount of argon. Probably theargon metastables or ions effectively quench thediatomic species. Experimentally, it is found that


best results are obtained if approximately 2l min of combustion gases and 3 l min ofy1 y1

argon are used. However, it should be noted thatthis mixture still contains much more moleculargases, and thus, sample gases, than the amountthat could be introduced into the ICP withoutdeteriorating the analyte signal(cf. Section 2.1).

3.4. Towards quantitative measurements

The presented measurements are not yet quan-titative and hitherto the measurement system canonly be used to follow relative fluctuations in thegas composition in time. To enhance process con-trol, quantitative measurements could be desirable.A major error source for quantitative measure-ments is related to the loss of particulates duringthe gas extraction. Heating of the joint betweenincinerator oven and plasma torch can minimizethis loss. However, even if the gases can beextracted without losses, and if we assume that theparticulates are totally evaporated inside the plas-ma, several problems remain.In contrast to liquid analysis, where calibration

is rather straightforward, calibration of combustiongas analysis is quite complex, mainly because nogaseous standards containing all relevant elementsare available(yet?). Probably a gaseous standard,preferably air or argon, containing one elementcan be used. For some elements these standardsare commercially available. One step further is acompletely standard-less calibration. For such astandard-less calibration a thorough understandingof plasma properties will be required. Apart frompractical problems such as the difficulties of the‘on-line’ determination of parameters as electrondensity and temperature and the gas temperature,it will be difficult to take into account the matrixeffectsw29x of the so-called easy ionizable elemen-ts. Especially, the highly abundant element sodium(f20 mg g in fly ashes, cf. Section 3.2) willy1

largely influence the argon or air plasma. Sincethe excitation energy from the ground state towardsthe radiative 3p P sodium state is only 2.1 eV2

0

(resulting in 589.0- and 589.6-nm radiation), theelectron energy distribution function will beseverely affected. It is obvious that this influencesthe population of excited states of other elements

as well. As an example, it is observed that inten-sities of argon lines significantly decrease, or evenvanish, if substantial amounts of sample gas withhigh sodium concentrations are introduced.

4. Conclusions

It is shown that plasma-AES can be used forthe on-line monitoring of several elements that arehighly abundant in combustion gases. The maingoal was to construct a compact measurementsystem capable to withstand extreme gas condi-tions as present above an oven of hazardous wasteincinerator plant(high temperature, highly corro-sive environment, the presence of fly ashes andvolatile compounds), rather than to achieve lowdetection limits, which usually is the merit of otherresearch groups. Since several heavy metals, suchas Zn, Fe and Pb, could easily be monitoredcontinuously, the built analysis system in principalcan be used for process control.After extraction from the incinerator oven, the

gases and particulates were dissociated and itsconstituents excited in a microwave-induced plas-ma produced by the TIA. AES finally providedinformation about the present metals. The TIA wasused as excitation source because plasmas createdby this torch could resist high concentrations ofmolecular gases and changing gas compositions.Best results were obtained with discharges cre-

ated at an under-pressure in a mixture of flue gasesand argon. Plasmas created at an under-pressureenabled the use of a very short heated samplingline between incinerator and plasma, so particlelosses during the sampling will be minimal. Theaddition of argon reduced the intensity of molec-ular bands, which otherwise easily overgrow weakatomic lines, and has the advantage that the plasmais less sensitive to changing concentrations of easyionizable elements, such as Na.Hitherto, measurements provided only relative

results. For quantitative analysis, a calibrationtechnique has to be found to overcome this gap.Due to the lack of good calibration standards andthe presence of dust particles and a high concen-tration of easy ionizable elements in combustiongases, it will not be easy to achieve a calibration


method and more insight in fundamental plasmaprocesses will be demanded.

Acknowledgments

The authors would like to thank AVR-Chemieand STW (Dutch foundation for fundamentalresearch) for their financial support and in partic-ular W.J. van de Guchte(AVR-Chemie) for hisassistance during the measurements at the incin-erator site and J. Verwoerd(AVR-Chemie) forreviewing this paper.

References

w1x G.A. Meyer, K.W. Lee, Real-time determination of metalhazardous air pollutants in flue gas emissions: laboratorystudy, Process Contr. Qual. 6(1994) 187–194.

w2x G.A. Meyer, M.D. Seltzer, Real-time analysis of metalsin stack gas using argonyair ICP optical emissionspectroscopy, Proceedings of SPIE Environmental Mon-itoring and Remediation Technologies, vol. 3534, Bos-ton, Nov 1998.

w3x C. Trassy, R. Diemiaszonek, P. Pasquini, R. Meunier,On-line analysis of elemental pollutants in gaseous,effluents by inductively coupled plasma-optical emis-sion spectroscopy, Proceedings of the International Sym-posium on Environmental Technologies: PlasmaSystems and Applications, Atlanta Oct 1995, 401.

w4x S. Hassaine, C. Trassy, R. Diemiaszonek, Trueness andprecision in on-line monitoring metallic pollutants influe gases by ICP-OES, Proceedings of Progress inPlasma Processing of Materials 1999, Begell House,New York, 1999, p. 785–792.

w5x D. Nore, A.M. Gomes, J. Bacri, J. Cabe, Developmentof an apparatus for the detection and measurement ofthe metallic aerosol concentrations in atmospheric air insitu and in real time: preliminary results, Spectrochim.Acta Part B 48(1993) 1411–1419.

w6x L.W. Peng, W.L. Flower, K.R. Hencken, H.A. Johnsen,R.F. Renzi, N.B. French, A laser-based technique forcontinuously monitoring metal emissions from thermalwaste treatment units, Process Contr. Qual. 7(1995)39–50.

w7x H. Zhang, J.P. Singh, F.Y. Yueh, R.L. Cook, Laser-induced breakdown spectra in a coal-fired MHD facility,Appl. Spectrosc. 49(1995) 1617–1623.

w8x G.A. Norton, D.E. Eckels, L.J. Clifford, Applicabilityof Atomic Fluorescence and Atomic Absorption Spec-troscopy for On-Line Analysis of Mercury in CoalCombustion and Gasification Effluents, Proceedings ofthe EPRIyDOE International Conference on ManagingHazardous and Particulate Air Pollutants, Toronto, Can-ada, August 1995.

w9x E.A.H. Timmermans, Design of a continuous gas ana-lyzer based on MIP-AES for the on-line monitoring ofmetallic compounds in flue gases, dissertation 2-yearpost-graduate program of Physical Instrumentation,Eindhoven University of Technology, The Netherlands,1997.

w10x A. Montaser, D.W. Golightly, Inductively Coupled Plas-mas in Analytical Atomic Spectrometry, second ed.,VCH Publishers, Inc, USA, 1984.

w11x P.W.J.M. Boumans, Inductively Coupled Plasma Emis-sion Spectroscopy—Part I and II, Wiley, USA, 1984.

w12x R.H Scott, V.A. Fassel, R.N. Kniseley, D.E. Nixon,Inductively coupled plasma-optical emission spectrom-etry: a compact facility for trace analysis of solutions,Anal. Chem. 46(1974) 75–80.

w13x Q. Jin, Y. Duan, J. Olivares, Development and investi-gation of microwave plasma techniques in analyticalatomic spectrometry, Spectrochim. Acta Part B 52(1979) 131–161.

w14x M. Moisan, Z. Zakrewski, R. Pantel, The theory andcharacteristics of an efficient surface wave launcher(surfatron) producing long plasma columns, J. Phys. D:Appl. Phys. 12(1979) 219–238.

w15x M. Moisan, J. Pelletier, Microwave Excitation Plasmas,Elsevier Science Publish. B.V, Amsterdam, The Neth-erlands, 1992, chapter V.

w16x J. Hubert, M. Moisan, A. Ricard, A new microwaveplasma at atmospheric pressure, Spectrochim. Acta PartB 34 (1979) 1–10.

w17x C.I.M. Beenakker, A cavity for microwave-inducedplasmas operated in helium and argon at atmosphericpressure, Spectrochim. Acta Part B 31(1976) 483–486.

w18x H. Matusiewicz, A novel microwave plasma cavityassembly for atomic emission spectrometry, Spectro-chim. Acta Part B 47(1992) 1221–1227.

w19x Q. Jin, C. Zhu, M. Borer, G. Hieftje, A microwaveplasma torch assembly for atomic emission spectrome-try, Spectrochim. Acta Part B 46(1991) 417–430.

w20x M. Moisan, G. Sauve, Z. Zakrewski, J. Hubert, An´atmospheric pressure waveguide-fed microwave plasmatorch: the TIA design, Plasma Sources Sci. Technol. 3(1994) 584–592.

w21x A. Ricard, L. St-Onge, H. Malvos, A. Gicquel, J. Hubert,M. Moisan, Torche a plasma a excitation micro-onde:` `deux configurations complementaires, J. Phys. III´France 5(1995) 1269–1285.

w22x E.A.H. Timmermans, I.A.J. Thomas, J. Jonkers, J.A.M.van der Mullen, D.C. Schram, The influence of molec-ular gases and analytes on excitation mechanisms inatmospheric microwave sustained argon plasmas, Fre-senius J. Anal. Chem. 362–365(1998) 440–446.

w23x E.A.H. Timmermans, J. Jonkers, A. Rodero, M.C. Quin-tero, A. Sola, A. Gamero, J.A.M. van der Mullen, D.C.Schram, The behavior of molecules in microwave-induced plasmas studied by optical emission spectros-copy. 1. Plasmas at atmospheric pressure, Spectrochim.Acta Part B 53(1998) 1553–1566.


w24x J. Jonkers, H.P.C. Vos, J.A.M. van der Mullen, E.A.H.Timmermans, On the atomic state densities of plasmasproduced by the torche a injection axiale, Spectrochim.`Acta Part B 51(1996) 457–465.

w25x J. Jonkers, L.J.M. Selen, J.A.M. van der Mullen, J.van Dijk, E.A.H. Timmermans, D.C. Schram, Steepplasma gradients studied with spatially resolved Thom-son scattering measurements, Plasma Sources Sci. Tech-nol. 6 (1997) 533–539.

w26x J.W. Olisek, J.C. Fister III, Incompletely desolvateddroplets in argon inductively coupled plasmas: theirnumber, original size and effect on emission intensities,Spectrochim. Acta Part B 46(1991) 851–868.

w27x S.E. Hobbs, J.W. Olisek, Inductively coupled plasmamass spectrometry signal fluctuations due to individualaerosol droplets and vaporizing particles, Anal. Chem.64 (1992) 274–283.

w28x C. Truitt, J.W. Robinson, Spectroscopic studies of organ-ic compounds introduced into a radiofrequency inducedplasma: Part II. Hydrocarbons, Anal. Chim. Acta 51(1970) 61–67.

w29x A.C. Lazar, P.B. Farnsworth, Matrix effect studies inthe inductively coupled plasma with monodisperse drop-lets, I: the influence of matrix on the vertical analyteemission profile, Appl. Spectrosc. 53(1999) 457–464.

Documents

Atomic emission spectroscopy for the on-line monitoring of incineration processes