Volume 51, Number 8, 1997 APPLIED SPECTROSCOPY 12290003-7028 / 97 / 5108-1229$2.00 / 0q 1997 Society for Applied Spectroscopy

Two-Dimensional Imaging of Flame Species UsingTwo-Photon Laser-Induced Fluorescence

NIKOLA GEORGIEV and MARCUS ALDEÂ N*Department of Combustion Physics, Lund Institute of Technology, P.O. Box 118, S-221 00 Lund, Sweden

The potential for two-dimensional visualization of combustion spe-cies by using two-photon laser-induced ¯ uorescence (LIF) has beeninvestigated. The technique was applied for two-dimensional (2D)imaging of carbon monoxide, ammonia, oxygen, and hydrogen at-oms in ¯ ames. Approaches for compensating the signal intensity forthe quadratic laser intensity dependence in two-photon imaging arediscussed. For the case of CO and H atom visualization, a potentialproblem is the interference from nonresonantly excited C2, whoseemission spectrally and spatially coincides with the ¯ uorescencefrom CO. Different strategies for elimination of the C2 emissionwere investigated. It was found out that the emissions from CO andC2 can be separated in time. For the case of the oxygen atoms, itwas observed that the relation between the intensities of the ¯ uo-rescence signals at 845 and 777 nm changes with the equivalenceratio of the investigated ¯ ame. An attempt to estimate the 2D de-tection limit for these species in ¯ ames is also made.

Index Headings: Laser-induced ¯ uorescence; Two-photon spectros-copy; Combustion diagnostics; Two-dimensional visualization.

INTRODUCTION

Pollution of the environment and limitations of theearth’s energy resources are two main targets in the at-tempts to gain a more detailed understanding of combus-tion processes, leading to more effective and environ-mentally friendly combustion. The application of ad-vanced laser-based techniques, with major features suchas high spatial and temporal resolution, and nonintru-siveness, as well as the ability to make species-selectivemeasurements, provides a primary means to ful® ll theseobjectives; see, for example, Ref. 1 and references there-in. Among these techniques, laser-induced ¯ uorescence(LIF) is the one that probably has received the most at-tention, mainly due to the possibility of achieving mul-tiple-point information through imaging measurements.2

Normally, LIF utilizes different electronic states of theinvestigated species. However, for many of the species ofinterest in combustion, the energy separation between theground state and the excited state is so large that laserradiation with a wavelength below 200 nm is requiredfor a conventional one-photon excitation process. Besidesthe dif® culties in creating laser radiation with suf® cientenergy in this wavelength region, the opacity of the at-mosphere and the ¯ ame gases for radiation below 200nm makes one-photon LIF almost nonapplicable in ¯ ameenvironments. In order to avoid these problems, two-pho-ton excitation schemes have been suggested and success-fully applied for detection of ¯ ame species.3,4 There areseveral features which make the two-photon LIF tech-nique quite different from the one-photon process:

Received 8 October 1996; accepted 14 January 1997.* Author to whom correspondence should be sent.

1. Different selection rules for dipole transitions allowstates that are not reachable by one-photon excitationto be coupled.

2. In the case where the two photons are coming fromtwo laser beams with k 5 2 k, Doppler-free measure-ments are possible.

3. The excitation rate is proportional to the square of thelaser intensity.

4. The temporal behavior of the laser beam is of greatimportance.

As stated, there are different signal dependencies onthe laser intensity along the laser beam for one- and two-photon processes:

2I I

S } ´A and S } ´Aone-photon two-photon1 2 1 2A A

where I is the laser-beam energy and A is the area of thelaser-beam cross section. It is obvious that, for a one-photon process, as long as the transition is not saturated,the focal length of the focusing lens is not of prime im-portance, because the decrease of the area close to thefocus is fully compensated by the increase of the laser-beam intensity. This is not the case for a two-photonprocess. Thus, the increase of the laser intensity by fo-cusing is not compensated by the decrease of the area.One way to minimize this problem is to assure that thearea does not change so much through the imaging sec-tion. This can be accomplished by using a lens with alonger focal length. However, the choice of the lens is atrade-off between the constant-area demand and the high-er power required in the two-photon process.

An advantage of the two-photon process is that the¯ uorescence wavelength is normally in the visible/near-IR spectral region, widely separated from the exci-tation wavelength in the UV region, which makes thespectral separation of the signal more comfortable. Onthe other hand, two-photon LIF suffers from a small crosssection, which leads to the necessity of applying higherlaser powers with the potential problems of subsequentphoto-chemical effects.5

A list of species that are of great importance in com-bustion processes, requiring a two-photon excitation pro-cess, as well as their excitation and emission wave-lengths, is shown in Table I.

As was indicated above, the most important feature ofLIF might be the possibility of obtaining two-dimension-al recordings. Of the species requiring a two-photon pro-cess, there have to our knowledge been reports only ofsingle-shot 2D imaging in ¯ ames of H2

6 and H2O7 inaddition to CO visualization using a multiple-pass ar-rangement.8

1230 Volume 51, Number 8, 1997

TABLE I. Important combustion species requiring a two-photonexcitation process, along with their excitation and emission wave-lengths.

Species Excitation wavelength (nm) Emission wavelength (nm)

H 2 3 205 656O 2 3 226 845/777CO 2 3 230 451± 725N2H3 2 3 305 565, 720H2 2 3 193 750, 830H2O 2 3 248 400± 600N 2 3 211 868± 873, 822, 744

In the present paper we evaluate the potential for sin-gle-shot two-dimensional (2D) imaging in ¯ ames of CO,NH3, O, and H using a powerful Nd:YAG-based lasersystem. CO, being a major pollutant, motivates investi-gations with the aim of giving a more comprehensivepicture of the behavior of this species in combustion pro-cesses. Ammonia is of interest in combustion processesboth for studies of selective catalytic reduction (SCR)phenomena and for understanding of combustion of ni-trogen-bound fuels. Oxygen as well as hydrogen atomsare intermediate species involved in many elementary re-actions mechanisms that are directly related to the fuelconsumption rate, the burning velocity, and the pollutantformations. This consideration makes investigations oftheir distribution in ¯ ames of crucial importance.

SPECIES SPECTROSCOPY

CO Spectroscopy. Carbon monoxide has been thesubject of several laser spectroscopic studies with the useof multiphoton excitation steps involving the X1S 1

ground state and the A1 P and the B1S 1 excited states bothin room-temperature cells9± 11 and in ¯ ames.12,13 The ion-ization potential of CO is 14.0139 eV; therefore, whenthe B1S 1 state is excited, absorption of one additionalphoton is suf® cient to ionize the molecule. Thus, the ex-citation of the B1 S 1 can be detected either by a 21 1resonance-enhanced multiphoton ionization (REMPI)process or through the resulting ¯ uorescence in the B1S 1

® A1P AÊ ngstroÈ m system, occurring within the spectralrange of 451 nm (v² 5 0) and 725 nm (v² 5 6). In thisexperiment, CO molecules are excited in the Hop® eld±Birge system, from the ground state X1S 1 (v² 5 0) to theB1S 1 (v 9 5 0) state through absorption of two photons at230 nm. According to the two-photon absorption theo-ry,14 the polarization of the exciting laser beam is of cru-cial importance for the oscillator strength of the Q branchin this transition. The ratio of 200:1 was found13 in thetransition strength when excited with a linear polarizedlaser beam, compared with a circular polarized laserbeam. In this work, a linearly polarized laser beam, tunedto the spectrally very dense Q branch head, was used.

NH3 Spectroscopy. The ammonia molecule has a py-ramidal ground state and planar excited states;15 in theregion between 220 and 115 nm, the transitions havebeen classi® ed into four progressions.16 The strong pre-dissociation, characteristic of the ® rst excited state of am-monia, the A state, with absorption bands between 170and 215 nm, makes conventional one-photon LIF verydif® cult to apply. In the region of the second series be-tween 140 and 167.5 nm, transitions to the B and C states

are most pronounced. In the late 1970s, a new electronicstate, C 9 , was observed.17 This state is almost isoenergeticwith the C state. In this work, the B(6) and B(7) statesat ; 303 nm, as well the C 9 (2) state, at ; 305 nm, wereexcited by using two-photon excitation processes, and¯ uorescence from the B± A transition at 720 nm and theC 9 ± A transition at 565 nm was recorded. The ¯ uores-cence bands resulting from the C 9 ± A transitions are calledthe Schusters bands.

O Spectroscopy. Two-photon excitation of oxygen at-oms in a ¯ ame was ® rst reported in Ref. 3. The excita-tion, through absorption of two photons at 226 nm, fromthe triplet ground state 2p3P to the triplet excited state3p3P, should normally result in nine ® ne-structure tran-sitions. The two-photon dipole selection rules, however,exclude the 0 ± 1 and 1± 0 transitions. Due to the smallenergy splittings in the excited state, these seven allowedtransitions are normally not resolved. Thus, in an exci-tation scan three different peaks originating from the ® ne-structure levels of the ground state to the allowed levelsin the excited state are usually recorded. The strongestpeak, 23P2± 33P2,1,0, consists of three unresolved compo-nents, while 23P1 ± 33P2,1 and the weaker one, 23P0 ± 33P2,0,each consists of two unresolved ® ne-structure compo-nents. The resulting ¯ uorescence at 845 nm originatesfrom the direct decay from the 3p3P state to the 3s3S state,whereas the ¯ uorescence at 777 nm is due to the decayfrom the 3p5P state to the 3s5S state, where the 3p5P stateis populated through collisionally induced energy trans-fer.

H Spectroscopy. Two-photon laser-induced ¯ uores-cence has been applied for detection of H atoms in¯ ames.4,5 Through absorption of two photons at 205 nm,the hydrogen atoms are excited from the ground 1s2Selectronic level to the 3d2D excited level. Simultaneously,the 3s2S electronic level is also populated. For the caseof excitation with a linearly polarized laser beam, exci-tation of the 3d dominates over excitation of the 3s byan order of magnitude.18 The resulting ¯ uorescence, n 53 ® n 5 2, occurs at 656 nm.

EXPERIMENTAL

The experimental setup used for 2D imaging usingtwo-photon LIF is shown in Fig. 1. A Nd:YAG laser(Continuum NY82-S) with a repetition rate of 10 Hz,producing ; 900 mJ in the frequency-doubled output at532 nm, was utilized. The laser pulse has a duration of; 8 ns and a bandwidth of 1 or 0.001 cm2 1, dependingon whether a seeder was used in the YAG oscillator. TheYAG laser was used to pump a dye laser (ContinuumND60), producing tuneable radiation in the requiredwavelength region. A prismatic beam expander inside theoscillator cavity of the dye laser yielded a linewidth of; 0.1 cm2 1. The output of the dye laser was wavelength-extended through frequency-doubling and -mixing innonlinear crystals, producing the different UV beams.

In the case of H, a laser beam at 205 nm was createdby ® rst frequency-doubling the dye laser beam at 615 nmin a KDP crystal, followed by frequency-mixing the fun-damental dye laser beam at 615 nm with the doubledbeam at 307 nm in a BBO crystal, yielding the thirdharmonic of the dye laser frequency. The linewidth of the

APPLIED SPECTROSCOPY 1231

FIG. 1. Experimental setup used for two-dimensional imaging usingtwo-photon laser-induced ¯ uorescence. CL, cylindrical lens; F, ® lter;UVX, UV extension box.

UV beam was 0.17 cm2 1, and laser energies between 3and 6 mJ were achieved.

In the case of oxygen atoms, the output of the dye laserat 574 nm was frequency-doubled, followed by frequen-cy-mixing of the doubled beam at 287 nm with the re-sidual IR beam at 1064 nm from the Nd:YAG laser tocreate the required laser beam at 226 nm. In the case ofCO molecules, the same wavelength extension technique,mixing after doubling, was used to produce 230-nm ra-diation with only the wavelength of the output from thedye laser being different. The ammonia molecules wereexcited by a frequency-doubled laser beam in the red/yellow spectral region. Since the linewidth of the UVbeam at respective frequencies for CO and O is depen-dent on the linewidth of the IR beam from the Nd:YAGlaser, the in¯ uence on the signal intensity when using aseeder in the YAG laser was also investigated. A maxi-mum pulse energy of about 15 mJ could be achieved at226 nm; however, most measurements were made withlaser energies between 5 and 10 mJ.

A Pellin± Broca prism, placed after the frequency-dou-bling or -mixing crystals, was used in order to spatiallyseparate the different harmonics from the laser. The re-quired UV beam was formed into a sheet by using cylin-drical lenses. The focal length of the lenses, normallybetween 200 and 500 mm, was chosen to minimize pho-tochemical effects. The beam waists were estimated to bebetween 100 and 250 m m. A Nikon lens (f 5 105 mm;f/1.2), positioned perpendicular to the laser beam, col-lected part of the ¯ uorescence and imaged it onto thephotocathode (S25) of the image-intensi® er of a charge-coupled device (CCD) camera (Princeton InstrumentsICCD-576S/RB-T), which was used as a detector. ThisCCD camera was chosen since the near-IR response wassubstantially higher than that of a conventional S20 cath-ode (about a factor of 10 at 845 nm). Appropriate optical® lters were chosen to spectrally isolate the desired ¯ uo-rescence emission.

In order to obtain spectral imaging, i.e., spatially re-solved spectral measurements, a spectrograph (JarrellAsh, Monospec 27) equipped with three interchangeablegratings, yielding dispersions of 1.5, 6.0, or 24 nm/mm,could be placed in front of the CCDcamera. The entranceslit of the spectrograph, normally 100 m m wide, was ori-ented along the laser beam, providing the spatial axis onthe recorded images parallel to the slit and the spectralaxis in the perpendicular direction. In the case when pow-er dependence was recorded, a variable beam attenuatorwas used in order to change the laser power without af-

fecting the beam properties, and the detector was run ina mode allowing changes in the signal intensity to berecorded as a function of time.

Several hydrocarbon ¯ amesÐ e.g., C3H8/air and CH4/air, producing unstable ¯ ames with a Bunsen burner, andlaminar CH4/O2/N2± NH3/ air ¯ ames, as well as H2/O2

¯ amesÐ were investigated.

RESULTS AND DISCUSSION

CO. Since CO is one of the few combustion speciesexcited with a two-photon process that can be containedin a cell (or ¯ ow), this species was also used for studiesof the general characteristics of two-photon excitationprocesses. As mentioned before, in a two-photon processthe signal depends on the laser intensity squared. Thus,inhomogeneites in the ¯ uorescence image, due to the in-homogeneous distribution of the laser power across thelaser beam, are more pronounced. Generally, a correctionfor the laser beam intensity pro® le is required. One wayto do this is to divide the recorded ¯ ame images with theimage obtained from homogeneously distributed COmolecules on a pixel-to-pixel basis and thereby compen-sate for beam inhomogenieties. In this experiment, an im-age from a cold CO ¯ ow was used; see Fig. 2A. In Fig.2B, a two-dimensional CO distribution from a rich CH4/O2/N2 ¯ ame is shown. The result obtained after laserbeam compensation is shown in Fig. 2C. The advantageof this method is that the correction is made with the reallaser beam pro® le, but disadvantages are that the lasermode might change during the time between the mea-surements and that this approach in practice can be ap-plied only with stable species (e.g., CO, H2, or NH3).Another way to accomplish the correction for an inhom-ogenous intensity beam pro® le is to perform a Fouriertransformation of the image shown in Fig. 2B and removethe vertical frequencies corresponding to the laser beaminhomogeneities. An inverse Fourier transform is then ap-plied in order to obtain the corrected image, shown inFig. 2D. Variations in the intensity of the laser beam arenot strictly periodic, which prevents them from beingcompletely removed. The advantage of this method isthat there is no need for a reference image; thus it maybe used for image corrections with the use of unstablespecies (e.g., H and O). A disadvantage is that some ofthe frequency components in the original image disap-pear, leading to a deterioration of the spatial resolutionin the vertical direction. In order to compare the resultsobtained with these two approaches, standard deviationsfrom an area inside the ¯ ame, containing about 2500 pix-els, were calculated. With the use of the standard devia-tion as a criterion, the two techniques gave results whichwere similar ( s within 20%).

Spectral interferences from nonresonantly excited C2

in hydrocarbon ¯ ames have previously been reportedwhen laser-induced ¯ uorescence was used for ¯ ame di-agnostics of CO.12 In this experiment, strong efforts weremade to investigate this effect in order to reduce the C2

emission, which, as is shown below, interferes both spec-trally and spatially with the emission from carbon mon-oxide. In Fig. 3, single-shot 2D images recorded on andoff CO resonance by using the Bunsen burner ¯ ame arepresented, showing CO and C2 in Fig. 3A and only C2 in

1232 Volume 51, Number 8, 1997

FIG. 2. Corrections for laser beam inhomogeneities in 2D two-photon LIF. (A) Laser beam pro® le obtained from a ¯ ow of cold CO; (B) 2D COdistribution from a CH4/O2/N2 ¯ ame, F 5 2; (C) image in Fig. 2B corrected for the laser beam pro® le from Fig. 2A; (D) 2D CO distribution fromCH4/O2/N2 ¯ ame, F 5 2, corrected through ® ltering in the Fourier plane.

FIG. 3. Single-shot 2D images of CO from an unstable propane/air Bunsen burner ¯ ame. (A) On CO resonance; (B) off CO resonance.

APPLIED SPECTROSCOPY 1233

FIG. 4. Two-dimensional images on CO resonance. The position of thegate of the detector was moved ; 10 ns between the exposures. Theimage where only CO is present was recorded ® rst. The sequence showsthat the CO and C2 signals do not arise simultaneously and can beseparated in time. (A) CO emission only; (B) overlapping of CO andC2 distributions; (C) C2 emission only.

FIG. 5. Spectra corresponding to the images in Fig. 4.

Fig. 3B. Fluorescence from C2 may be a severe problemfor ¯ ames with an equivalent ratio F . 1 ( F 5 1 is de-® ned as the fuel/oxygen ratio for complete combustionto carbondioxide and water). For example, the peak C2

emission intensity from a methane/air ¯ ame with F 51.5 is of the same order of magnitude as the peak emis-sion intensity from CO, whereas the CO emission isabout an order of magnitude stronger than the C2 emis-sion at F 5 0.8.

Different schemes have been suggested in order to ex-plain the C2 emission. One possible path is through sootablation and/or PAH photodecomposition.19,20 Also thepossibility of photodecomposition of acetylene has beenshown.21 According to the latter scheme, C2 moleculesare formed in the excited state (d3 P g) in a two-step pro-cess:

C H ® C H 1 H (1)2 2 2

3C H ® C (d P ) 1 H. (2)2 2 g

The threshold energy for the ® rst step is 5.38 eV. Thus,one photon at 230 nm, corresponding to 5.39 eV, is suf-® cient to overcome this barrier. Since most ¯ ames inves-tigated in this work were non-sooting, the photodecom-position mechanisms seem to be the most probable causefor the C2 emission. Two different experimental ap-proaches were investigated as means to suppress the C2

interferences. The ® rst approach was to investigate the

temporal behavior of the CO and C2 emissions. In Fig.4, the sequence of three images recorded with a separa-tion of ; 10 ns between the position of the gate (width; 20 ns) of the CCD camera is shown, revealing thatseparation of the CO and C2 is possible in time. Thecorresponding spectra are shown in Fig. 5. The potentialof using a narrow-band ® lter (e.g., at 483 nm) to spec-trally isolate the CO emission may be applicable in pointmeasurements but was not applicable for imaging be-cause of the signal loss.

The second approach was to investigate whether therewas a different response in the CO and C2 emission in-tensity with respect to changes in the laser power, whichwould make it possible to choose a laser power wherethe CO ¯ uorescence could be promoted. By moving thegate of the detector to a position where only one specieswas observed at a time, we measured the power depen-dencies (PDs) (i.e., the slope of a log± log plot of signalintensity vs. laser intensity) of CO and C2, for a spherical(f 5 200 mm) and a cylindrical (f 5 150 mm) focusinglens, using rich CH4 air ¯ ames. By using a spherical lens,we measured a PD of about 1 for both CO and C2 and,while using the cylindrical lens, measured a PD ; 1.5±1.8 for both species. Due to the two-photon process, anI2 dependence is normally expected, but due to the strongphotoionization of the B1 S 1 state, this dependence is de-creased to values between 1 and 2. When ionization com-pletely dominates over the quenching rate, a power de-pendence of 1 should be expected, and the signal can beevaluated in terms of absolute number density.22 Linearpower dependence of CO has been measured in previousinvestigations13 and is in good agreement with the resultspresented here for the case of spherical focusing lens.When the cylindrical lens was used, the laser power den-sities in the imaged area decreased by approximately twoorders of magnitude, resulting in an increase of the powerdependence. The ionization cross section for this transi-tion has been determined in Ref. 13 to be at least 7 3102 17 cm2 at 230 nm. With laser pulse energies between1 and 5 mJ, the ionization rate for the case of cylindricallens was calculated to be between 6.4 3 108 and 3.2 3109 s 2 1, which should be compared with an estimatedquenching rate of 2 3 109 s 2 1.23

The Q branch in the X1S 1 (v² 5 0) ® B1S 1 (v 9 5 0)

1234 Volume 51, Number 8, 1997

FIG. 6. Two-dimensional images of NH3 from a rich atmospheric pressure NH3/O2 ¯ ame. (A) On a welding torch; (B) on a sintered plug, abovethe reaction zone.

transition of CO is spectrally very dense since the rota-tional constants for the X and the B states are very sim-ilar, 1.923 and 1.948 cm2 1, respectively.24 Thus, when thelaser frequency is tuned to the band head of the transition,several rotational lines are simultaneously excited. Theintensity of the ¯ uorescence signal depends on the pop-ulation of the ground states, excited within the bandwidthof the laser. Since the ground-state number density is de-pendent on the temperature through the Boltzmann dis-tribution, a change in temperature will yield different sig-nal intensities, depending not only on the different num-ber density but also on the different population distribu-tion. This linewidth-dependent temperature correctionfactor and its in¯ uence in 2D imaging thus needs to beestimated. In order to make this estimation, numericalcalculations, simulating the excitation spectrum at differ-ent temperatures of the X± B transition, have been per-formed. The spectral line pro® le was assumed to be Voig-tion, consisting of a temperature-dependent Dopplerwidth and a Lorentzian width, the latter resulting fromthe sum of the collisional broadening and ionization-rate-dominated lifetime broadening. The Voigt pro® le wascalculated by using the approach described in Ref. 25.From these calculations, it was seen that when a laserlinewidth of 0.2 cm2 1 centered at 230.105 nm is used,the correction factor changes by ; 50% when the tem-perature is increased from 1000 to 2000 K.

An attempt to estimate the detection limit for single-shot 2D CO visualization was also made with a stoichi-ometric methane/air ¯ ame. Taking the ¯ uctuations of thebackground in the image of the stoichiometric methane/air ¯ ame to be about 100 counts and having a maximumsignal minus background of 16,000 counts, we estimatedthe signal-to-noise (S/N) ratio to be about 160 when us-ing 6.5 mJ of laser energy, a laser sheet with a height of

5 mm, 150-mm focusing lens, and an effective pixel sizeof ; 70 m m. Assuming the maximum mole fraction ofCO in this ¯ ame to be ; 4%,26 and using the experimentalparameters given above, leads to an estimated S/N ratioof 1 at a mole fraction of ; 250 ppm.

NH3 Measurements. Prior to the ¯ ame measurements,cell measurements were performed in order to investigatepotential in¯ uence of two-photon absorption in the im-ages. The ¯ uorescence from ammonia at different pres-sure (from several tens of mbar to ; 900 mbar) was re-corded. A short-pass ® lter at 600 nm and a Schott OG550® lter were used in these measurements. The laser poweron each particular pressure in the cell was varied fromseveral mJ to ; 30 mJ. The maximum absorption in thecell did not exceed 10%, in comparing the laser beamtransmission on and off ammonia resonance. It was thusconcluded that this effect would have a negligible in¯ u-ence on the images recorded.

Flame measurements were performed with a rich at-mospheric-pressure NH3/O2 ¯ ame on a welding torch andon a sintered plug (Figs. 6A and 6B). According to theresults described in Ref. 27, the NH3 concentration in anatmospheric-pressure ¯ ame decreased by an order ofmagnitude within a millimeter from the unburned gasmixture. Thus, the rapid decline in the NH3 concentration,in combination with the temperature rise above the re-action zone, explains the very low signal outside the¯ ame cone. On the other hand, the image in Fig. 6B wasrecorded by using a sintered plug high above the reactionzone. The image is an average of 10 laser shots, and themaximum signal intensity is about 16,000 counts on abackground of 1700 counts. Thus, there is a considerablepotential for performing single-shot measurements ofNH3 in the ¯ ame zone.

O and H Atoms. The measurements of atomic oxygen

APPLIED SPECTROSCOPY 1235

FIG. 7. Single-shot 2D image of oxygen atom distribution in a stoichiometric CH4/air ¯ ame.

FIG. 8. Single-shot 2D image of hydrogen atom distribution in a stoichiometric CH4/O2 ¯ ame. No ® lters were used. Spectral measurements revealthe strong contribution of C2 in the region of the reaction zone.

was initiated by spectral imaging of the ¯ uorescencewavelengths at 845 and 777 nm. The ratios between theintensities of the ¯ uorescence signals were investigatedfor H2/O2, CH4/O2, and CH4/air ¯ ames at different F . Theresults showed that the ratio of the maximum intensitiesof the two lines increased by a factor of 2 when goingfrom a stoichiometric mixture to F 5 2. This differenceis of importance since it indicates a difference in quench-ing for these ¯ uorescing channels with respect to speciescomposition. This observation has very recently alsobeen noted in direct quenching measurements using pi-cosecond resolved measurements.28 The practical impli-cations for imaging is that images may look differentdepending on whether the 845- or 777-nm emission isdetected. Measurements on a diffusion methane ¯ ameshowed practically no emission at 777 nm.

Two-dimensional images on each of the three oxygenexcitation peaks around 226 nm were performed with andwithout the injection seeder of the Nd:YAG laserswitched on. Due to the high spectral density when using

the seeder, the signal in this case is more than a factor of3 larger. Figure 7 shows a single-shot two-dimensionalimage of oxygen atoms in a stoichiometric CH4/air ¯ ame.The detection limit for the 2D oxygen atom measure-ments was in this case estimated to be 20 ppm with theuse of a laser power of 10 mJ, a beam height of 5 mm,a focusing lens of 250 mm, an effective pixel size of ; 70m m, and an assumed number density of oxygen atoms of3000 ppm.26

In the case of hydrogen atom visualization, mainly aCH4/O2 ¯ ame was investigated. The ® rst measurementswere made by using only a colored-glass ® lter (SchottRG 630) in front of the detector. A single-shot 2D imagefrom this ¯ ame is shown in Fig. 8. The spectral imageshown in Fig. 9 reveals that, besides the spatially broad¯ uorescence signal from H atoms, the C2 ¯ uorescencewas distributed in a narrow region close to the reactionzone. This interference was also clearly visible in a fuellean ¯ ame, although with less intensity. For comparison,for F 5 2, the maximum intensity of the ¯ uorescence

1236 Volume 51, Number 8, 1997

FIG. 9. Spectral imaging of a hydrogen atom distribution in a stoichiometric CH4/O2 ¯ ame. Besides the ¯ uorescence signal from H atoms at 656nm, ¯ uorescence belonging to the Swan band from unresonantly excited C2 is shown. The C2 ¯ uorescence is spatially narrower than the H signaland is located in the region close to the reaction zone.

from the D n 5 0 band in C2 emission at 516 nm wasapproximately the same as the maximum intensity of the¯ uorescence from the hydrogen atoms, while for F 5 1the C2 emission was about three times weaker than thesignal from H atoms. With the use of a narrow interfer-ence ® lter with peak transmission at 656 nm in front ofthe detector, the C2 emission could be totally suppressed.

The 2D signal from H atoms in the stoichiometricmethane ¯ ame produced up to ; 5000 counts in a singlelaser shot with a pulse energy of 5 mJ with a height of5 mm and a pixel size of ; 70 m m. The H atom concen-tration in a free propagating methane/oxygen ¯ ame, F 51, is calculated to 5%.29

CONCLUSION

Our investigations, performed on different ¯ ameswith various equivalence ratios, have undoubtedlyshown the capability and applicability of the two-pho-ton, two-dimensional LIF technique. Recorded images

reveal the potential of the technique for qualitativemeasurements of the spatial distributions of species in¯ ames.

Different approaches for compensating for the nonlin-ear laser intensity dependence have been discussed andcompared, where the Fourier transformation techniquehas the main advantage that no reference image needs tobe recorded.

For CO visualization, the problem with spectral in-terferences from C2 has been specially addressed, andit was shown that, by proper gating, this interferencecould be almost totally suppressed. A detection limitfor single-shot imaging of some hundreds of parts permillion was estimated under our experimental condi-tions. Also for H, C2 could constitute a spectral inter-ference if a proper ® lter was not used. In the case ofO, spectral imaging indicated that the ratio between the845- and 777-nm emission was F dependent. For bothO and H, single-shot visualization was demonstrated.

APPLIED SPECTROSCOPY 1237

ACKNOWLEDGMENTS

This work was ® nancially supported by the National Swedish Boardfor Industrial and Technical Development (NUTEK). The authors ac-knowledge the help of Dr. Fabian Mauss in calculating the ¯ amemolefractions.

1. A. C. Eckbreth, Laser Diagnostics for Combustion Temperatureand Species (Abacus Press, Cambridge, Massachusetts, 1987).

2. R. K. Hanson, in Proceedings of the Twenty-First Symposiom (In-ternational) on Combustion (The Combustion Institute, Pittsburgh,Pennsylvania, 1986), pp. 1677± 1691.

3. M. Alde n, H. Edner, P. GrafstroÈ m, and S. Svanberg, Opt. Commun.42, 244 (1982).

4. R. P. Lucht, J. T. Salmon, G. B. King, D. W. Sweeney, and N. M.Laurendeau, Opt. Lett. 8, 365 (1983).

5. J. E. M. Goldsmith, Appl. Opt. 26, 3566 (1987).6. W. Lempert, G. Dishkin, V. Kumar, I. Glesk, and R. Miles, Opt.

Lett. 16, 660 (1991).7. H. Neij and M. Alde n, Appl. Opt. 33, 6514 (1994).8. J. Haumann, J. M. Seitzmann, and R. K. Hanson, Opt. Lett. 11,

776 (1986).9. S. V. Filseth, R. Wallenstein, and H. Zacharias, Opt. Commun. 23,

231 (1977).10. J. H. Glownia and R. K. Sander, Appl. Phys. Lett. 40, 648 (1982).11. U. Westblom, S. Agrup, M. Alde n, H. M. Hertz, and J. E. M.

Goldsmith, Appl. Phys. B 50, 487 (1990).12. M. Alde n, S. Wallin, and W. Wendt, Appl. Phys. B 33, 205 (1984).13. P. J. H. Tjossem and K. C. Smyth, J. Chem. Phys. 91, 2041 (1989).

14. R. G. Bray and R. M. Hochstrasser, Mol. Phys. 31, 1199 (1976).15. G. Herzberg, Electronic Spectra and Electronic Structure of Poly-

atomic Molecules (Van Nostrand, Reinhold Company, New York,1966).

16. A. B. F. Duncan, Phys. Rev. 47, 822 (1935).17. G. C. Nieman and S. D. Colson, J. Chem. Phys. 68, 5656 (1978).18. J. Bokor, R. R. Freeman, J. C. White, and R. H. Storz, Phys. Rev

A 24, 612 (1981).19. D. A. Greenhalgh, Appl. Opt. 22, 1128 (1983).20. P. E. Bengtsson, and M. Alde n, Combust. Flame 80, 322 (1990).21. J. M. Seitzman, Quantitative Applications of Fluorescence Imaging

in Combustion, HTGL Report No. T-275, Stanford, California(1991).

22. D. A. Everest, C. R. Shaddix, and K. C. Smyth, `̀ Quantitative Two-Photon Laser-Induced Fluorescence Imaging of CO in FlickeringCH4/Air Diffusion Flames`̀ , paper submitted to the Twenty-sixthSymposium (International) on Combustion (1996).

23. S. Agrup and M. Alde n, Appl. Spectrosc. 48, 1118 (1994).24. G. Herzberg, Spectra of Diatomic Molecules (Van Nostrand, Rein-

hold Company, New York, 1950).25. J. Humlicek, J. Quant. Spec. Rad. Transfer 21, 309 (1979).26. G. Tsatsaronis, Combust. Flame 33, 217 (1978).27. D. I. Maclean and H. G. G. Wagner, in Proceedings of the Eleventh

Symposium (International) on Combustion (The Combustion Insti-tute, Pittsburgh, Pennsylvania 1967), p. 871.

28. F. Ossler, J. Larsson, and M. Alde n, Chem. Phys. Lett. 250, 287(1996).

29. F. Mauss, private communication.