1

CH and CH2O in Atmospheric Pressure Methane/Air

Bunsen Flames

Robert J.H. Klein-Douwel, Jorge Luque, Jay B. Jeffries,Gregory P. Smith, and David R. Crosley

Molecular Physics Laboratory, SRI International333 Ravenswood Avenue, Menlo Park, California 94025

Western States Section of the Combustion Institute, Fall 1999, Paper 12F99

Abstract: Laser-induced fluorescence of CH B2Σ-v’=1 is excited inatmospheric pressure Bunsen flames using rotational levels above andbelow the predissociation limit. There is significant rotational, vibrational,and electronic energy transfer even when levels above the predissociationlimit are excited. At atmospheric pressure rotational energy transferproduces a variation in the fluorescence quantum yield with rotationallevel and gas temperature. Laser-induced fluorescence of CH2O excitingA-X 41

0 allows formaldehyde detection at an excitation wavelength quitenear (and often overlapped) with the CH B-X (1,0). The formaldehydestructure is contained inside the CH in the premixed inner cone of theBunsen flame. Excitation of CH2O A-X 41

0 provides an LIF strategy thatminimizes the variation of the population of the excited level over the 700-1800 K temperature range important in the flame.

1. IntroductionPartially premixed Bunsen type flames form the mode of combustion of nearly all

the natural gas consumed in residential appliances and the vast majority burned in smallcommercial applications, such that usage constitutes over 25% of U.S. natural gasconsumption. These flames consist of two stages: a rich premixed flame forms an innercone and a lean diffusion flame constitutes an outer cone. These partially premixedflames are a key target of present and future pollutant emission regulation. NOxemissions will be the subject of increasingly stringent requirements, and it is likely thatCO, aldehydes and aromatics will also come under scrutiny. The consequences ofoperation under marginal conditions are of particular concern. Understanding thestructure of these flames can provide guidelines for burner design efforts to meet theregulatory requirements while still maintaining reasonable efficiency. For example, asimple practical solution to reduction of NOx emissions in some commercially availableburners employs metal inserts; these reduce NOx, probably due to cooling of the flame,but the actual mode of operation is unknown and the amount of toxics such as CO andaldehydes is likely increased.

A model burner has been designed to study a partially premixed Bunsen flameusing laser-induced fluorescence (LIF) diagnostic techniques. The emphasis of themeasurements is an understanding of the flame structure and the chemistry leading topollutant formation. The diagnostics focus on measurements of CH as a marker of thepremixed flame front and the key precursor to prompt NO, and on formaldehyde as a

2

potential air toxic from incomplete hydrocarbon combustion. Our strategy for CHmeasurement uses excitation of predissociative transitions in the B-X (1,0) band withdetection of (1,1) fluorescence, attempting to minimize quenching effects on quantitativedeterminations at atmospheric pressure. The existence of non-predissociative and highlypredissociative rotational levels in the same vibrational state provides a unique scenarioto test the effects of rotational relaxation (and vibrational and electronic energy transferas well) in laser-induced fluorescence measurements. The existence of an overlappinghot band excitation of formaldehyde, A-X 41

0, allows simultaneous or near-simultaneousdetection of this species. The temperature bias encountered with the usual ground stateexcitation strategy is largely avoided by using this band. After discussing diagnosticdevelopment and application, planar LIF (PLIF) images in Bunsen flames are presented.

2. CH MeasurementA variety of strategies have been proposed for quantitative laser-induced

fluorescence measurements of reactive intermediate radicals in atmospheric pressureflames that attempt to minimize collisional quenching corrections using predissociativelevels of the excited radical.1 If the predissociative level has a lifetime short compared tothe time for quenching collisions, then fluorescence from the predissociated state itself isfree from the influence of collisional quenching. However, the collisional energy transferof even a small population to non-predissociative levels complicates this simple picturebecause the fluorescence quantum yield is much larger for the non-predissociative states.In a related previous work, 2 we found a rotational level dependent quantum yield forfluorescence when non-predissociative levels below the predissociation limit in CH Bv'=0 are excited. A non-negligible fraction of the excited molecules is transferred bycollisions beyond the predissociation limit and this fraction becomes larger when thelevel pumped is closer to the onset of predissociation. If this rotational level variation inthe quantum yield is ignored temperature errors of 10% are found in atmosphericpressure flames, according to simulations that model this rotational energy transfer.

In the work reported here, we measure excitation and dispersed laser inducedfluorescence spectra of predissociative and non-predissociative levels of CH B v'=1 inatmospheric pressure, Bunsen-type partially premixed, methane/air flames. Rotationallevels with N’>6 are predissociated, and the fluorescence lifetime decreases from 400 nsfor N’≤6 to less than 0.07 ns for N’>8.3 The existence of non-predissociative and highlypredissociative rotational levels in the same vibrational state provides a unique scenarioto test the effects of rotational relaxation in LIF measurements, due to the varying timescales. The vibrational level v'=1 has advantages over v'=0 for the predissociative LIFstrategy. The combination of the electronic transition moment variation and largeanharmonicity of the state results in an emission coefficient for the (1,0) band about thesame as that of the (1,1) band, so both efficient excitation and nonresonant detection arepossible. The predissociation onset is at N'=7, and the rotational levels most commonlyexcited for combustion diagnostics, N'=7-10, have collision free-lifetimes of 5, 0.3, 0.06and 0.015 ns, which are very attractive for laser-induced predissociative fluorescenceapplications. In B v'=0, the predissociation appears at N'>14, too high a rotational levelfor significant thermal population, and thus not practical for combustion diagnosticsapplications. In this study, we examine the feasibility of performing laser-inducedfluorescence in the predissociative levels of CH B v'=1 for combustion diagnostics atatmospheric pressure, and evaluate and model the effects of rotational relaxation (RET),vibrational relaxation (VET) and electronic energy transfer (EET) on the fluorescenceyield of CH B v'=1.

3

The burner system consists of a steel tube of 1.65 cm i.d. and 1.90 cm o.d.containing the rich premixed inner cone flame flow of methane and air (Φ=1.36 at 7.2slm), surrounded by a 20cm diameter air coflow with a matching velocity of 60 cm/s. TheCH B-X (1,0) transition at 365 nm is excited with a dye laser LambdaPhysik FL-2002(temporal width ~ 14 ns, spectral bandwidth ~ 0.2 cm-1

, vertically polarized), pumped bya XeCl excimer laser LambdaPhysik EMG 103 MSC. The path of the laser beam throughthe Bunsen flame perpendicular to the flame front is imaged, after passing adepolarization filter, onto a SPEX 270M spectrometer(f/4, 500 nm blaze angle) with agated intensified CCD camera (Princeton Instruments, ICCD-576G/RBT, 14 bitsdynamic range, 384x576 pixels, pixel size 23 µm). The spectrally dispersed signal isalong the horizontal direction of the array. The spectral resolution is 0.25 nm FWHM(about four camera pixels), determined from calibration measurements using the Hg lineat 404.6 nm. The vertical direction in the collected images corresponds to the horizontalposition along the path of the laser beam through the flame, which enables spatialresolution of the fluorescence images. One CCD pixel corresponds to 20 µm in theflame, but the imaging of a test card indicates the spatial resolution of our opticalcollection system is approximately 8 pixels or 150 µm. Appropriate background imagesof the dark current of the CCD camera and of the natural emission of the flame arerecorded and subtracted from the dispersed fluorescence images. The wavelength axes ofthe images are calibrated using known wavelengths of Raman signals from O2, N2, CH4,or optical flame emission bands. The system spectral response in this region is obtainedfrom the branching ratio B-X(1,1)/B-X(1,2). CH data are taken midway up the innercone, from the center peak region, at laser powers below saturation.

Wavelength (Å)

4020 4040 4060 4080 4100 4120 4140

ExperimentSimulation

Rotational number (J)

0.5 2.5 4.5 6.5 8.5 10.5

0.0

0.2

0.4

R

PQ1

Q2 2 4 6 8 10

2 4 6 8 10

Q1(8) Excitation

Fig. 1. Dispersed CH B-X (1,1) fluorescence (0.25 nm resolution and total time integrated) in anatmospheric pressure methane/air Bunsen flame after N'=8 excitation. Inset shows the timeintegrated rotational level fractional populations from the model calculation.

When a level beyond the predissociation limit is excited in the atmosphericpressure flame, much of the resulting fluorescence originates from non-predissociatinglevels populated by rotational energy transfer. This is illustrated in the wavelengthresolved fluorescence spectrum in Figure 1. Here the Q1 (8) line is excited, but only 40%

4

of the fluorescence originates from the N’=8 level (solid line). The rotational relaxationrates Rij are modeled with a statistical power gap law:4

Rij = (2Nj+1) exp(∆Eij/kT) a (∆Eij/B)-α ,

with parameters a = 4 x 107 /s and α = 1.2 from a low pressure flame study of RET in CHB-X v'=0.4 Individual rotational level populations are obtained from solving the coupleddifferential kinetic equations involving emission, quenching,5 predissociation,3 and RETrates. The resulting simulated spectrum from integrated emission rates is shown as adashed line. The relative time-integrated rotational populations are shown in the inset,and we see that all of the levels below the predissociation limit are populated. Even incases of low RET and fast predissociation rates, the higher fluorescence quantum yieldsof the lower collisionally populated levels can dominate due to relatively higherfluorescence quantum yields.

We can also estimate the fluorescence quantum yield as a function of rotationallevel, again using quenching data from low-pressure flames, 5 our rotational energytransfer model, 4 and theoretical predissociative lifetimes. 3 The results are shown inFigure 2. Note that RET to levels above the predissociation limit reduces the quantum

Rotational number (N')

0 2 4 6 8 10

Flu

ore

sce

nce

Qua

ntu

m Y

ield

(C

H B

v'=

1)

10-4

10-3

10-2

No RETRET

Rotational number (N')

0 2 4 6 8 10

10-4

10-3

10-2

No RETRET

T =1900 K T = 1200 K

Fig. 2. Fluorescence quantum yield for CH B-X v'=1 in an atmospheric pressure methane/airflame at two different temperatures, modeled with and without RET, versus initially excitedrotational level.

yield for levels below predissociation. Conversely, levels above the predissociation limithave a larger quantum yield because of RET to levels that do not predissociate. Theamount of transfer is rotational level dependent, producing a rotationally dependentquantum yield.

For N'=9, and 10 the predissocation lifetime are 0.06 ns, and 0.015 nsrespectively, but collisions still account for 20% and 6% of the removal rate from theserotational levels. Avoiding competition with energy transfer collisions at atmosphericpressure, forces the use of highly predissociated rotational levels with very low

5

fluorescence yields. In the present experiment, at atmospheric pressure and typical flameCH concentrations, B v'=1, N'=8 is readily detected, N'=9 signal level is low andunacceptable for most imaging applications, while N'=10 is barely detectable.

From a diagnostics point of view, it is important to assess the fluorescencequantum yield variation due to changes in collisional rates, which are usually functions oftemperature and composition. Rotational relaxation between predissociative and non-predissociative levels complicates the analysis. RET promotes lower pumped levelsabove the dissociation limit, decreasing the quantum yield, whereas the opposite is true ifpredissociative levels are pumped. The model suggests that pumping N'=8 at flametemperatures nearly results in a cancelation of the RET effects But it is obvious from thisstudy that predissociative LIF schemes using CH B v'=1 are not as effective as hoped forthe reduction of collisional dependences.

Vibrational and electronic energy transfer (VET and EET) are also importantcollisional removal pathways even for levels above the predissociation limit. Energytransfer from predissociating to non-predissociating levels of only a small fraction of thepopulation can account for a large fraction of the total fluorescence, and result in asubstantially larger effective quantum yield. The data in Figure 3 show that EET andVET together account for nearly half of the total fluorescence observed after exciting thepredissociative N’=8 level of v’=1 of the B-state. Thus broadband fluorescence detectionfor larger signal levels must take these processes into account. Fortunately, there appearsto be little dependence of the A-X and B-X (0,0) fractions on temperature or initiallypumped B-state level.

Wavelength (Å)

4000 4200 4400 4600

B-X (0,0) B-X (1,1) B-X (1,2)CH A-X

CH B-X (1,0) Q1(8) Excitation

Fig. 3. Dispersed CH Fluorescence ( 0.25 nm resolution and total time integration) in theatmospheric pressure methane/air flame.

3. Formaldehyde MeasurementIn the same wavelength region as the CH B-X (1,0) P-branch excitation a dense butstructured LIF signal can be easily observed, mainly toward the interior unburned side ofthe CH in the inner cone Bunsen flame. Based on line features in the excitationspectrum, and especially on the familiar LIF fluorescence spectrum, we have identifiedthis feature as a hot band excitation of the CH2O A-X transition - specifically the 41

0 bandinvolving the 1167 cm-1 CH2 wag. (Some older sources identify this mode as υ6.) An

6

excitation spectrum obtained by scanning the laser wavelength, taken in the premixedflame cone just inside the observed CH, is shown in Figure 4. Identification asformaldehyde is confirmed by the match of known prominent line and K sub-band headassignments.6 Figure 5 shows a typical dispersed fluorescence spectrum, following aparticular excitation by the laser. The well recognized pattern of CH2O fluorescence isevident, and matches the expected wavelengths and assignments. Similar spectra wereseen at other excitation wavelengths, including that of the broad background. Thisunderlying unresolved excitation is typical of hot polyatomic species in flames, and maybe assigned to higher hot band excitations (41

0511, etc.) and high rotational levels.

Previous flame LIF detection of CH2O has used the strong ground vibration level40

1 and noted the large temperature dependence of this choice due to variations of theBoltzmann population factor. 7 Premixed flame model calculations indicate significantCH2O concentrations in regions from 700K to 1800K. Over this range the groundvibrational level population fraction varies by a factor of 5. The 41

0 excitation is thepreferred choice to minimize such bias, with a variation of only 1.55 over the range 700-1800 K. When also taking rotational level population into account, the choice of a high

Fig. 4. Formaldehyde LIF excitation spectrum (broadband fluorescence detection) in theatmospheric pressure methane-air Bunsen flame, with some 41

0 assignments. Lower panel showspart of the laser scan at higher resolution.

368 369 370 371 372 373

4 2

6 0K

369.3 369.4 369.5 369.6 369.7

Flu

ores

cenc

e Int

ensi

ty (

arb

units

)

K = 3 <- 2

R6 3

2

124Q

Excitation Wavelength (nm)

7

Fig. 5. Fluorescence spectrum of formaldehyde in the atmospheric pressure methane-air flamefollowing laser excitation of the R(6) K=3←2 line of the 41

0 band. The characteristic progressionsin υ4 and υ2 are evident.

K level is also beneficial in reducing temperature bias, although not as important as thevibrational level choice. Extra uncertainty and bias is introduced by the underlying hotband background expected to increase with temperature, so finding a distinct featureabove the background is perhaps more important in picking a CH2O detection strategythan optimizing rotational level. A sharp K(5←4) feature of the CH2O 41

0 band islocated just to the red of the P(8) line of the CH B-X (1,0) excitation, and other K sub-band heads are also available. Collisional effects such as the temperature dependence ofthe fluorescence quenching are not known. It is evident from observations offluorescence bands from v4'=1 shown in Figure 5 that vibrational energy transfer to thislow-lying level (125 cm-1) readily occurs.

4. Flame MeasurementsWe have taken a series of flame images using PLIF for CH and CH2O in a set of

atmospheric pressure methane-air Bunsen flames, at various inner flow stoichiometriesand with a selection of heat-removing inserts. From the investigations described above,the protocol should excite the Q1(8) line in the CH B-X (1,0) band, excite a sharp featureof the CH2O 41

0 band, and select laser and fluorescence detection wavelengths tominimize interference. Strong overlapping excitations should definitely be avoided.(The CH Q lines are relatively free of CH2O overlap, unlike the P lines.) Although thetwo species distributions are largely separated spatially in the Bunsen flame inner cone,the broad formaldehyde excitation and fluorescence pattern always presents a potentialcontamination of the CH measurements. The 2-D image in Figure 6 shows the LIFpattern as a function of excitation wavelength vertically and position through the flame

350 400 450 500 550 600

4.0x104

8.0x104

1.2x105

1.6x105

2.0x105

las

er

Inte

nsi

ty (

a.u

.)

Fluorescence Wavelength (nm)

403

4 01

4 21

4 30

4 41

4 61

2 10 4 50 + 2

30 4 21

2 20 4 30

2 10 4 41

2 20 4 212 10 4 30

2 20 4 10

2 10 4 21

2 10 4 10

8

Fig. 6. LIF signals through the primary flame front of the atmospheric pressure methane-air flameas a function of laser excitation wavelength (vertical axis) and position (horizontal axis). CH andCH2O features are identified. Only minimal processing has been done to all images.

front horizontally. The contamination of the CH signal when broadband fluorescence isdetected is evident. Using a narrow filter to confine CH LIF detection to the B-X (1,1)band would minimize this problem (although some CH2O excitation and fluorescencewill still occur at these wavelengths), at the expense of signal level. An alternativeprocedure employs subtraction of a second image taken just off resonance in laserexcitation wavelength from the sharp CH or CH2O feature.

To obtain Bunsen flame CH and CH2O PLIF images, CH was excited on theQ1(8) line of the B-X (1,0) band, and CH2O was excited at the A-X 41

0 K=7←6maximum. Results for the Φ = 1.36 flame are shown in Figure 7, along with a line slicethrough the 2-D images. Fluorescence was detected through a filter at wavelengthslonger than 380nm. A second image was taken with the laser tuned just blue of the lines,and subtracted from the first to remove interference from non-resonant fluorescence,Raman scattering, flame emissions, and laser scatter. For the CH image, this procedurealso removes the interfering CH2O LIF. The images are also corrected for laser beam anddetector array inhomogeneities by normalization to a Rayleigh scattering image. Clearseparate images can be obtained. We also see the CH2O is almost entirely on the interiorof the CH distribution, and although this is earlier in the sequence of the reactionchemistry (CH3 + O → CH2O + H) the complete separation is somewhat surprising. Bothsignal levels are quite uniform in intensity and width along the flame cone, and are onlyassociated with the premixed flame front area.

-4.0 -3.5 -3.0 -2.5 -2.0 -1.5

368.9

368.8

368.7

368.6

x (mm)

λ exc

(nm

)

368.9

368.8

368.7

368.6

41

0

K=5<-4

P1(8)

P2(8)

CH2O

CH

-4.0 -3.5 -3.0 -2.5 -2.0 -1.5

CH2O

CH

9

Fig. 7. PLIF images of CH and CH2O LIF in an atmospheric pressure Φ = 1.36 methane-air Bunsen flame. Lower trace is a cut at 12 mm height.

From the CH LIF image, an estimate of the fluorescence quantum yield using themodel described above, and using the size of the methane Raman scattering to calibratethe optical system, we can also roughly determine the amount of peak CH in thepremixed flame. For Φ = 1.36 our preliminary value is 2.0 ± 0.8 ppm. It is interesting tonote that for this mix of predissociative and non-predissociative levels, the fluorescenceyield varies more with changes in the predissociation and RET rates than with changes inthe quenching rates.

Images were also taken at richer stoichiometries for the inner cone premixedportion of the flame, up to the Φ = 3.0. The CH2O LIF levels remain fairly constantthrough this range, whereas the CH signal decreases and becomes very difficult to imagecleanly at Φ > 2.

We also investigated the influence of heat-removing metal inserts into or near theflame, by obtaining and comparing PLIF images of CH, CH2O and flame structure for theΦ = 1.36 flame. Two examples are shown in Figure 8. First, a rectangular metal plate isleaned against the side of the outer diffusion flame. The images perpendicular to theplate shows increased CH on the side near the plate, with little effect on CH2O. Thesecond insert is a screen placed above the flame tubes, to which the flame becomesattached as an efficient heat sink. Here the pictures show a diffuse, slower burningmethyl-formaldehyde chemistry with little outer CH prior to the diffusion flame region,before the CH appears to mark the final premixed flame front with much the similarappearance and levels as in the unperturbed flame. While interpretation of these results isnot complete, they serve to illustrate the type of useful information these PLIFdiagnostics might provide.

-10 -5 0 5 10

CH

CH2O

x (mm)

-8 -4 0 4 8

5

10

15

20

CH

-8 -4 0 4 8

5

10

15

20

ΦΦΦΦ = 1.36

CH2O

10

Fig. 8. PLIF images of CH and CH2O LIF in atmospheric pressure Φ = 1.36 methane-air Bunsenflames with metal inserts. Left: rectangular metal plate leans as shown against the outer diffusionflame, perpendicular to the laser beam. Right: metal screen inserted below the flame

5. ConclusionsPredissociative LIF diagnostics of the prompt NO precursor and flame front

marker CH using B-X (1,0) excitation have been investigated and applied in atmosphericpressure methane-air Bunsen flames. Energy transfer collisions play an important role ininterpreting the results, and in fact this system is a good choice for the study of suchprocesses. From this understanding, we conclude that excitation of N'=8 provides thebest imaging diagnostic. In addition, nearby LIF excitations of a hot band of CH2O cangive measurements of this intermediate and potential toxic that are relatively unbiasedwith respect to temperature variations. By subtracting off-resonant background images,distributions of each species can be obtained in Bunsen flames. The CH2O distinctlyappears before the CH; effects of inserts were also probed with these methods.

AcknowledgementsThis work was supported by the Basic Research Group of the Gas Research Institute.

-8 -4 0 4 8

Plate M esh

x (m m )

CH2O

CH

11

References1. E. W. Rothe and P. Andresen, “Application of tunable excimer lasers to combustion

diagnostics: a review,” Appl. Opt. 36 3971-4033 (1997).2. J. Luque and D. R. Crosley, “Radiative, collisional, and predissociative effects in CH

laser-induced fluorescence flame thermometry,” Appl. Opt. 38 1423-1433 (1999).3. N. Elander, M. Henenberger, and P. R. Bunker, “Theoretical studies related to time

resolved spectroscopy: the iterative Rydberg-Klein-Dunham method and Weyl theoryapplied to the predissociation in the B state of CH,” Phys. Scripta, 20 631-646 (1979).

4. J. Luque, R. J. H. Klein-Douwel, J. B. Jeffries, and D. R. Crosley, “CollisionalProcesses Near the CH B2Σ- v’=0,1 Predissociation Limit in Laser-InducedFluorescence Flame Diagnostics,” Applied Phys. B., submitted, September, 1999.

5. M. Tamura, P. A. Berg, J. E. Harrington, J. Luque, J. B. Jeffries, G. P. Smith, and D.R. Crosley, "Collisional Quenching of OH, CH, and NO in Low-Pressure Flames,"Combustion and Flame, 114 502-514 (1998).

6. J. H. Callomon and K. K. Innes, "Magnetic Dipole Transition in the ElectronicSpectrum of Formaldehyde," J. Molec. Spectr. 10 166-181 (1963).

7. J. E. Harrington and K. C. Smyth, "Laser-Induced Fluorescence Measurements ofFormaldehyde in a Methane/Air Diffusion Flame," Chem. Phys. Lett. 202 196-202(1992).