1. Introduction

Due to the issue of combustion processes causing the formation of greenhouse gases such as CO$_2$, contributing to the global temperature increase, as well as potentially harmful pollutants, there is a strong demand to improve combustion efficiency and reduce pollutant emissions. For this purpose, it is essential to understand the underlying complex chemistry [1] but also turbulent flows and flow-flame interactions since combustion in practical devices, e.g. gas turbines, takes place under turbulent conditions that enhance mixing and mass consumption of reactants resulting in increased energy release rate and power [2–4].

Laser-based diagnostic techniques have been utilized for the analysis of the combustion processes due to their ability to provide valuable information on temperature and species concentrations non-intrusively with high spatial and temporal resolution [5,6]. Such methods are particularly valuable for studies of transient phenomena such as turbulent combustion. Laser-induced fluorescence is one of the most established techniques for sensitive detection of chemical species in combustion, in particular planar laser-induced fluorescence (PLIF), which is used to obtain two-dimensional (2D) spatial distributions of species in flames [7]. The technique is based on resonant absorption of laser photons which often requires a tuneable laser system, and many investigations are thus focused on studies of a singles species at a time. However, in turbulent flames, measurements of a single species provide limited information. Simultaneous measurements of more than one species enable investigations of interactions to obtain increased understanding of phenomena, which is essential for development of accurate numerical models for simulations and predictions of the process. To approach this problem, experimental setups for PLIF measurements of multiple species in turbulent flames have been arranged, see e.g. [6] and references therein. Specific examples of multiple species PLIF studies are the combined imaging of the hydroxyl (OH) radical and formaldehyde (CH$_2$O) for heat-release imaging presented by Paul and Naim [8], simultaneous visualization of CH and OH radicals by Kiefer et al. [9] and simultaneous imaging of distributions of temperature, OH and CH$_2$O by Gordon et al. [10]. Multi-species PLIF imaging in highly turbulent premixed flames operating in the distributed reaction zone regime has been realized by Zhou et al. [11]. Further diagnostic development has been made with high-speed lasers operating at kHz repetition rate and simultaneous PLIF imaging of OH and CH$_2$O has been implemented to investigate turbulent flames [12,13] also in combination with flow-field visualization [14].

Some light atomic species like H, N, and O have their absorption resonances at wavelengths around 200 nm approaching the vacuum-ultraviolet region, which makes these species inaccessible with single-photon excitation. Multiphoton excitation schemes have been proposed and during the last decade, diagnostics with nanosecond lasers have been increasingly supplemented with pico- (ps) and femtosecond (fs) lasers, which provide high peak powers beneficial for this purpose. In addition to providing efficient excitation for two-photon fluorescence (TPLIF), the use of ps/fs-lasers significantly reduces the occurrence of photolytic interferences in TPLIF measurements of atomic species [15–18]. The short laser pulses also enable time-resolved fluorescence lifetime measurements to determine collisional quenching, e.g. demonstrated for OH and CO by Jonsson et al. [19], and retrieve quantitative species concentrations from measured signals. For multi-species measurements, the broad spectral width of fs laser pulses gives an opportunity to simultaneously excite several species with one laser pulse if their absorption lines are within the bandwidth of the fs laser pulse. Simultaneous excitation of two species in flames using single fs laser pulses has been presented for OH and CO [20] as well as H and OH [21]. In the latter work, H-atoms were probed through a three-photon excitation process at laser wavelength 307.7 nm.

Atomic H and O are intermediate core species in the combustion chemistry of hydrocarbons [22] and also play roles in the formation of pollutants such as NOx [23]. The H-atom is a very reactive species with high diffusivity and its importance in a combustion process has, for example, been presented by Burluka et al. in a comparison between turbulent burning velocities obtained with hydrogen and its deuterium isotope [24]. Experimental data on instantaneous distributions of H and O radicals in flames will facilitate a deeper understanding of combustion phenomena, further development of computational models and assist in the future transition to efficient and sustainable energy supply.

With the aforementioned objectives in mind, the current study shows a method based on fs-laser excitation with the ability of simultaneous TPLIF for two-dimensional visualization of H and O atoms in turbulent flames on a single-shot basis. The possibility to retrieve quantitative H and O atom distributions in a flame is demonstrated with calibration versus signals and simulations of laminar flames. Measurements have been made in flames of different equivalence ratios and the trend in H and O atom number density profiles are compared with predictions from simulations of corresponding laminar flame cases. The discussion of the paper includes the potential for extending the technique to simultaneously visualize more than two species, possible diagnostic experiment to handle the concern of the photolytic effects with fs-laser excitation, effects due to the temporal and spectral characteristics of the fs-laser pulses, and methods for more accurate quantification of the data.

2. Method

2.1 Laser-induced fluorescence measurements

Simultaneous laser-induced fluorescence imaging of H and O atoms was performed using a system providing pulses in the femtosecond (fs) regime. The system comprised a Nd:YAG pump laser (Innolas), a diode-pumped mode-locked Ti:Sapphire laser (Coherent, Vitesse 2W), a Ti:Sapphire chirped pulse amplification (CPA) system (Coherent, Hidra-50-F) and dual optical parametric amplifiers (Light Conversion, HE-TOPAS) with frequency mixing apparatus (NirUVis unit). The fs laser sends two 800 nm 125-fs 10 Hz beams through separate optical parametric amplifiers (OPAs). Afterward, the OPA beams are sent through separate NirUVis units, which convert the incoming wavelength into 205 nm and 226 nm laser pulses that correspond to the two-photon excitation of the $3^2S$ $\leftarrow$ $1^2S$ and $3^3P$ $\leftarrow$ $2^3P$ transition of atomic H and O, respectively. A drawing of the arrangement of the laser system units with a single OPA can be found in [25] while Fig. 1 shows a schematic of the experimental arrangement around the burner in this study. The UV laser beams pass through cylindrical lenses ($f^{'}$ = 150 mm and $f^{'}$ = 300 mm) to convert the circular beams to 5 mm sheets with the energy of about 35 $\mathrm{\mu}$J for the 205 nm laser beam and 45 $\mathrm{\mu}$J for the 226 nm laser beam. The laser sheet thicknesses are on the order of 100 $\mathrm{\mu}$m resulting in fluences of 7 and 9 mJ/cm$^2$ for H- and O-atom excitation, respectively. The proximity of the two UV-wavelength makes beam overlap by a dichroic mirror unfeasible and the laser beams were therefore arranged in separate paths that intersected at the angle < 1$^{\circ }$ in the probe volume located $\sim$ 1mm above the burner surface.

 

Fig. 1. Diagram of the experimental setup. The angle between the 205 nm beam and the 226 nm beam is exaggerated for clarity. Laser beams are terminated with beam dumps.

Download Full Size | PDF

The two laser sheets were focused at the center of a modified McKenna burner. Fluorescence of the H $2^2P$ $\leftarrow$ $3^2S$ and O $3^3S$ $\leftarrow$ $3^3P$ transition was imaged by intensified CCD cameras (iCCD PI-MAX IV 1024i and 1024f, Princeton Instruments). Each camera is located perpendicular to the laser beam propagation and fitted with a $f^{'}$ = 50 mm camera lens (Nikon Nikkor, $f\#$ = 1.2 ). The acquired images had a spatial scale of 20 $\mathrm{\mu}$m/pixel and a feature of three pixels representing the smallest resolvable structure would then be 60 $\mathrm{\mu}$m, comparable to the resolution achieved in other PLIF studies of turbulent flames in this type of burner [11]. The cameras are assigned to their respective emissions due to the sensitivities of their sensors. A 15-nm narrow band-pass filter centered at 655 nm (Semrock, FF01-655/15- 25) is fitted to the PI-MAX IV 1024i camera to suppress the background when capturing the hydrogen atom fluorescence, while a 15-nm narrow band-pass filter centered at 850 nm (Semrock, FF01-850/15- 25) is fitted to the PI-MAX IV 1024f camera to capture the oxygen fluorescence. Additional background interferences were suppressed by using 50 ns time gating. The cameras were externally triggered. To synchronize their acquisition times the initial trigger pulse is sent by the master laser (Coherent, Hidra-50) to the iCCD camera used to detect the oxygen fluorescence (PI-MAX IV 1024f). This camera then sends a pulse to a delay generator (Stanford), which sends a trigger pulse towards the iCCD camera used to detect the hydrogen fluorescence. The delay between the excitation pulses in the probe volume is estimated to be within the nanosecond scale, which is much less than the time scale of the turbulent chemistry of the flame. To validate the synchronization of the cameras the shot-to-shot variation of flame chemiluminescence was depicted by both cameras simultaneously on a single-shot basis.

2.2 Burner and flame conditions

The McKenna burner is a hybrid porous-plug/jet burner with an inner central tube of 2.2 mm diameter and outer porous plug of 60 mm diameter. A CH$_4$/O$_2$ mixture with varying equivalence ratios from $\phi$ = 0.8 to $\phi$ = 1.3 was fed into the jet, while a CH$_4$/air mixture of equivalence ratio $\phi$ = 0.9 was used as a co-flow pilot flame. The jet speed of 61 m/s resulted in irregular motion of the 14 l/min flame flow, which can be described by the Reynolds number (Re). For $\phi$ = 0.8 Re = 8382, for $\phi$ = 0.9 Re = 8363, for $\phi$ = 1 Re = 8227, for $\phi$ = 1.1 Re = 8331, for $\phi$ = 1.2 Re = 8316 and for $\phi$ = 1.3 Re = 8302, which puts the experimental condition in the turbulent flow regime.

2.3 Calibration measurements

In addition to the fluorescence measurements under turbulent flame conditions and chemiluminescence measurements, fluorescence measurements were performed in a laminar CH$_4$/O$_2$ flame of equivalence ratio $\phi$ = 1.85. The acquired image data, obtained under stationary conditions, were used for the compensation of the laser beam spatial inhomogeneity and calibration of the species concentration values. Measurements in the laminar Bunsen-type flame with uniform flow and temperature distributions along the sides of the flame cone provided an image of the laser two-photon excitation profile in the vertical direction of both H and O atoms. In the post-processing, the jet flame PLIF images were divided by the laser beam profile cross-section received from the laminar flame fluorescence measurements. These measurements also provided calibration factors for converting the H and O PLIF signals into number densities, which were obtained from modelled profiles of the laminar flame.

2.4 Simulations/modelling of species concentrations

The modelling of species concentrations was performed using the GRI-Mech 3.0 chemical kinetic mechanism [22] in the CANTERA software [26]. For simulations the laminar Bunsen-type flame stabilized by the surrounding flat pilot flame was arranged as a counterflow simulation with opposing flows consisting of the room-temperature Bunsen-type flame reactants and the hot combustion products of the surrounding pilot flame. The flow of the reactants set in the simulation was determined from the experimental reactant input flows and the flame cone angle. Compared with this flow, going through the small jet orifice, the opposed flow of combustion products from the pilot flame was considered negligible. In addition to simulations of the laminar calibration flame, simulations were also made for conditions, equivalence ratios and flows, of the turbulent jet flames.

3. Results

3.1 2D single-shot imaging

Figure 2 presents single-shot H- and O-atom fluorescence images recorded simultaneously in CH$_4$/O$_2$ jet flames. The right- and left-hand-side column shows the images for H and O PLIF, respectively. Different rows correspond to equivalence ratios ranging from $\phi$ = 0.8 to $\phi$ = 1.3. In the final image post-processing, two major steps were performed. First, the PLIF signals were compensated for the laser sheet inhomogeneous profile, and second, the compensated signals were converted to the number density (cm$^{-3}$) by multiplication with factors retrieved from the calibration measurements. The image colorbars (cm$^{-3}$) show the evaluated number density for both O and H data. The PLIF images were measured at a distance of 1 mm above the burner surface, and the measurement region extends to 4 mm in the vertical direction. It is evident from the images that close to the burner surface, the flame does not exhibit strong turbulence, and the layers for both O and H PLIF show symmetric shapes. The turbulent motion starts to be observed at $\sim$ 2 mm above the burner surface showing continuous wrinkled layers. As can be seen from the figure, the maximum distribution of both H and O atoms is located at the bending curves of the turbulent layers. This is more evident from Fig. 2 for H-atom for $\phi$ = 0.8, $\phi$ = 1 and $\phi$ = 1.1.

 

Fig. 2. Arbitrary selected 2D single-shot fluorescence images of O and H atoms in CH$_4$/O$_2$ jet flame for different equivalence ratios, $\phi$ ($\phi$ = 0.8, $\phi$ = 0.9, $\phi$ = 1, $\phi$ = 1.1, $\phi$ = 1.2 and $\phi$ = 1.3). The right and left column shows H- and O-atom PLIF images, respectively.

Download Full Size | PDF

The wrinkled shapes of the fluorescence signal structures of O and H for different equivalence ratios agree quite well. This confirms that the single-shot TPLIF images in Fig. 2 show the same turbulent event for both O and H. For detailed analysis it is required to accurately determine the thicknesses of the laser sheets and to assess potential errors due to uncertainty in sheet overlap. The similarities in structures of single-shot images indicate that this error is sufficiently small for the current demonstration. Comparisons between image pairs show the propagation of H-atom distributions further into the unburned region of the jet flame compared with the O-atom distributions. This indicates that H-atoms diffuse back towards the reactant side and also explains the more vivid shapes of the turbulent events for the H-atom distributions.

The relative standard deviation of signal regions at HAB=1.8-2.6 mm for the single-shot images of the $\phi$ = 0.8 flame shown in Fig. 2 is 25% and 29% for the H and O-atom images, respectively. These values include image noise which can be further assessed from number density profiles measured in the horizontal direction across the images, as shown in Fig. 3.

 

Fig. 3. Single-shot O and H-atom number density profiles measured at height 2.2 mm in CH$_4$/O$_2$ jet flame for equivalence ratio $\phi$ = 0.8

Download Full Size | PDF

The profiles were evaluated across three pixel rows at the middle position of the images corresponding to height 2.2 mm above the burner. A comparison of the profiles in regions with and without signal indicates the signal-to-noise ratios of the H- and O-atom image data to be 20 and 7, respectively. Detection limits for all the investigated equivalence ratio $\phi$ conditions are presented in Table 1 below and correspond to concentrations on the order of a few percent at flame temperatures $\sim$ 2000K.

Table 1. Detection limit value for H and O calculated after all the post-processing for a single-shot image of a turbulent flame.

View Table

At a height $\sim$ 3.5 mm above the burner surface the O-atom distribution expands further out into the surrounding atmosphere compared with the H-atom distribution. Due to the low concentration of O atoms, it is hard to clearly observe the turbulent events and draw conclusions about the O-atom distributions for $\phi$ = 1.2 and $\phi$ = 1.3. Nonetheless, it can still be noticed that at the base of the burner surface H atom starts forming further out from the center compared with the O atom for rich flame conditions.

3.2 H and O fluorescence profiles

Figure 4 presents radial number density profiles of H and O for the investigated flames. Experimental profiles were retrieved from images at a height of 1 mm from the burner surface, where the turbulent motion is assumed to be low as the images show no large vortex structures in this region (see Fig. 2). Each experimental single-shot image was compensated for the spatial inhomogeneity of the laser beam profile, and the number density values were obtained through the calibration procedure. A low-pass filter has been applied to the experimental data to smooth the profiles.

 

Fig. 4. Experimental (solid lines) profiles averaged from 100 single-shot TPLIF measurements in the turbulent flames. Profiles obtained from chemistry simulations of laminar counterflow flames (dashed lines) are compared with experimental data for O (green-colored lines) and H (purple-colored lines) atoms in CH$_4$/O$_2$ jet flames for different equivalence ratios, $\phi$ a) $\phi$ = 0.8, b) $\phi$ = 0.9, c) $\phi$ = 1, d) $\phi$ = 1.1, e) $\phi$ = 1.2 and f) $\phi$ = 1.3.

Download Full Size | PDF

The experimental profiles, obtained as averages of 100 single-shot measurements in an image region where the flame is rather steady (cf. Fig. 2), are presented together with profiles obtained from simulations of laminar counterflow flames. The modeling does not consider any effects of turbulent flow and does therefore not represent detailed simulations of the flames, which would include chemical kinetics combined with computational fluid dynamics. Instead, the simulations made here rather serve to check that the processed image data follow correct trends and provide quantitative number density values of realistic magnitude.

Qualitatively the results show good agreement as the experimental data reproduce the trends of the simulations for the dependence of peak number density on equivalence ratio and the relation between peak H and O number density. Going radially outwards from unburned jet flame reactants at position 0 mm, all profiles show a rapid increase in number density at $\sim$ 1 mm. Towards the combustion products on the outside of the flame brush, the experimental profiles show a similar decay for all the $\phi$-values, suggesting that the distributions are sufficiently wide to be influenced by the laser beam focusing, determined by the focusing lens changing the laser sheet thickness further away from the center of the burner. Such dependence on laser beam focusing is a characteristic of two-photon fluorescence measurements, see e.g. [27]. The simulated profiles instead show a plateau prior to a decrease in number density.

Quantitatively the simulated peak number density values are a factor of 2-3 higher than the experimental values for all equivalence ratios for both species. Considering that the simulations model laminar instead of turbulent flames, the agreement is satisfactory and demonstrates that the single-shot image data can be quantified using an appropriate calibration scheme. For example, concentrations of atomic species measured by two-photon fluorescence can also be retrieved by calibration versus noble gases, see e.g. [28–30]. The GRI-Mech 3.0 chemical mechanism employed here is able to predict oxygen atom concentrations measured in planar laminar flames [30] with an accuracy on the order of 10%. A further uncertainty is the difference in fluorescence quenching between the calibration flame and the turbulent flames. Collisional quenching rates have been reported for both the H- and O-atom [31]. Calculations of O-atom quenching cross sections for different flame equivalence ratios show differences of $\sim$ 10$\%$, suggesting the error due to quenching differences to be on this order of magnitude. However, to completely correct data for the impact of quenching would either require time-resolved fluorescence measurements or compensation of the measured data with quenching rates obtained from literature. The first alternative introduces significant experimental complexity, while the second requires knowledge on collisional partners, which then needs to be retrieved either experimentally or from flame simulations in an iterative process. There is also an uncertainty associated with the calibration procedure for the laser beam profile. The variation in the selected image pixel area used to form a laser beam profile cross-section used in the data evaluation gives a small uncertainty, $\sim$ 2$\%$. The laser pulse energy variation between measurements was $\sim$ 3$\%$. Combining these uncertainties as independent sources of error in a total experimental uncertainty of 15$\%$, which has been indicated with error bars in the experimental plots of Fig. 4.

The peak number density values for H and O atoms obtained for different equivalence ratios are shown in Fig. 5, with the 15$\%$ uncertainly included as error bars also in this graph. The values have been normalized to the values obtained for the $\phi =0.8$ flame, and it is evident from the graph that the experimental and simulated profiles overall show similar trends. The H-atom number density increases with equivalence ratio, while the O-atom concentration shows the opposite trend. An exception is observed for the hydrogen atoms measured in the fuel-rich $\phi =1.3$ flame for which the signal and evaluated number densities seem lower than expected from the simulation trend. Identifying the reason for this requires further investigations. Nevertheless, the experimental H- and O-atom number densities can be quantified and are qualitatively consistent with simulated profiles.

 

Fig. 5. Experimental (circle symbols) and simulated (dashed lines) number density data compared for O (green color) and H (purple color) atoms in CH$_4$/O$_2$ jet flames for different equivalence ratios, $\phi$. Flame number densities have been normalized versus the values for the flame of $\phi$ = 0.8

Download Full Size | PDF

4. Discussion

The dual-OPA setup allows for simultaneous measurements of species probed via two-photon excitation, as shown in this study. For imaging species distributions, measurements across a larger field of view are preferable. To accomplish this, the laser beam should have a fairly homogeneous profile, which was a limitation of the setup at the current period of time. The 800 nm fs-laser beam generated by the CPA system undergoes optical parametric amplification, sum-frequency and second-harmonic-generation processes in order to obtain a femtosecond UV beam. Slight misalignment of the original 800-nm laser beam results in multiple diffractions and nonlinear interactions in the OPA system, which consequently gives spatial variations in the laser beam profile and hence the laser sheet.

In addition to measurements of different chemical species using separate laser units, femtosecond laser pulses can deliver further multiple species detection. The spectrally broad femtosecond pulses provide an enhanced two-photon excitation using a large number of photon pairs. There has been proposed a three-photon excitation scheme [21] and combination of two- and one-photon excitation schemes [32] for the H-atom. Following the idea presented by Jain et al. [21] it is possible, with the current configuration, to perform a simultaneous imaging of four species using two laser units for the excitation. Three-photon laser induced fluorescence can be achieved in H at 307.7 nm together with one-photon excitation of the OH radical at X$^{2}\Pi (v$"$=0)$ $\to$ A$^{2}\Sigma ^{+}(v$’$=0)$ band. Simultaneously, the two-photon laser induced fluorescence can be achieved in O at 226 nm and one-photon excitation in nitric oxide (NO) at X$^{2}\Pi (v$"$=0)$ $\to$ A$^{2}\Sigma ^{+}(v$’$=0)$ band [33]. This configuration can be simplified to simultaneous three species detection. In this case the laser excitation wavelengths can be the same as described in this study. The 205 nm wavelength can be used for H-atom excitation, and the 226 nm wavelength can be used for excitation of both O and NO.

However, it is worth mentioning that in the current study the two excitation laser beams centered at 226 nm and 205 nm were overlapped temporally within a nanosecond scale but not within a femtosecond scale. This means that while the measurements can be considered simultaneous on the time scales of the combustion, they are separate events on the time scale of the laser pulses. Overlapping two beams in time might give rise to additional nonlinear wave-mixing effects, or it might ionize the gas, in which case electrons and ions could produce new species. Such effects might result in interferences or loss mechanisms for the LIF measurements. Care thus needs to be taken when arranging and combining pulses for excitation of multiple species.

A big challenge with the two-photon LIF technique is the potential presence of photolytic interferences, i.e. photodissociation of molecules containing the atomic species of interest. For example, with the H atom, a significant contribution to the photolytic interferences may rise from the vibrationally excited H$_{2}$O in the product zone and vibrationally excited OH. A common approach to investigate this problem is to compare normalized spatial profiles of the fluorescence signal for different laser pulse energies [34]. If, at a certain pulse energy, the profile shows a deviation in shape, it can indicate that photolytic effects are taking place. However, in order to receive accurate quantitative H- or O-atom distribution in a flame zone, the comparison of normalized fluorescence profiles as presented in [17,35] might not be sufficient. Such a comparison can be uncertain if the distribution of the photolytic precursor molecule overlaps spatially with the species of interest. A potential experiment to investigate potential photolytic interference further can be to measure the time required for the photodissociation process to happen through a pump and probe study. This experimental idea is described in more detail in chapter 5 of the thesis by M. Ruchkina [35].

5. Conclusion

We would like to emphasize that in this study we have received simultaneous interference-free single-shot 2D images of H and O atoms in turbulent flames for the first time. The H- and O-atom fluorescence distributions measured in the reaction layers of the flames showed resembling patterns confirming that the same instantaneous turbulent event was recorded for both species. Imaging data were quantified through calibration measurements in laminar flames. Number density profiles for H- and O-atoms evaluated at position 1 mm above the burner surface show qualitative agreement with simulation data for laminar flames of corresponding equivalence ratio. Quantitatively, the experimental values are lower by a factor of $\sim$ 3, as the simulations only include combustion chemistry without considering effects of the flow. Altogether the study shows the potential for performing two-photon imaging measurements where more than two species can be visualized simultaneously on a single-shot basis. This represents a further step in bringing non-linear optical techniques based on the use of ultrafast lasers as valuable tools for studies of non-stationary reactive flows with atomic radical species, e.g. in turbulent combustion and plasma.