Within this study, dual-pump coherent anti-Stokes Raman spectroscopy (DP-CARS) is combined with planar laser-induced fluorescence of the hydroxyl radical (OH-PLIF) and nitric oxide (NO-PLIF) to measure the local temperature, \({{\mathrm CO}_2}\) mole fraction together with the instantaneous location of the flame front and the local equivalence ratio. The following subsections present details on the underlying measurement techniques.
2.3.1 NO-PLIF
To investigate mixing between the main flow and secondary fuel, NO calibration gas was seeded to the secondary fuel (10000
\({\textrm{ppm}_\textrm{V}}\)) and the main flow (150
\({\textrm{ppm}_\textrm{V}}\)). NO was chosen as a tracer due to its high thermal stability and slow reaction kinetics (Skalska et al.
2010), high signal-to-noise ratios (SNR) and well-known spectroscopic properties (Bessler et al.
2002; Sadanandan et al.
2012a; Paul
1997). The additional seeding of the main flow serves a twofold purpose: (a) it allows to extract the energy profile in the laser light sheet in areas with constant seeding density and temperature and (b) allows an assessment of the calibration uncertainty and absorption effects. The local equivalence ratio is derived based on a calibration, where various mixtures with equivalence ratios
\(\mathrm {\varPhi } \in [1,2]\) were established using calibrated thermal mass flow controllers. For calibration, the mixture was fed into a pipe (4 mm inner diameter) with a fine mesh at the outlet for a homogeneous outflow. Measurements within the core of the jet were used to calibrate the LIF signal. The calibration measurements were taken with the same settings and on the same day as the measurements at the SWQ burner.
Even though the measurement of local equivalence ratio is calibration based and thus the excitation scheme for NO-PLIF is in principle arbitrary, it is tremendously helpful to choose a transition for excitation, where the resulting LIF signal is nearly linear with the quantity of interest. This greatly reduces uncertainties in the evaluation. The line selection procedure is similar to the approach presented by Greifenstein and Dreizler (
2021a). In short, LIF excitation spectra were simulated using LifSIM (Bessler et al.
2003) with quenching data from Settersten et al. (
2009) for various theoretical mixtures that were also used for the calibration measurements. The simulation was performed in the wavelength range 225.5–226.8 nm and accounted for the measured laser linewidth (High Finesse, WSU-30) and the detection filters used in the experiment (Schott UG5, Semrock LP02-224R). From this simulation, it was found that the overlapping
\(\mathrm {Q_1(2.5)}+P_{21}(2.5)+Q_1(3.5)+P_{21}(3.5)\) transitions in the (0,0) vibrational band of the
\(A^2\mathrm {\Sigma ^+} \leftarrow X^2\mathrm {\Pi }\) electronic system near 44,197.5
\({cm^{-1}}\) yield a promising compromise between linearity to the local equivalence ratio and SNR. These transitions were excited with the frequency tripled output of a tunable dye laser (Sirah, PrecisionScan) operated with Pyridine I dissolved in ethanol, which was optically pumped using a frequency doubled pulsed Nd:YAG laser (Spectra Physics, GCR-4) running at 10 Hz repetition rate. The dye laser fundamental frequency and linewidth were measured with a wavemeter (HighFinesse, WSU-30) to ensure long-term stability and validity of the calibration. The maximum pulse energy near 226 nm was approximately 10 mJ at the laser exit.
The initially round beam was expanded in z-direction into a sheet using a cylindrical telescope (\(f={-50\,{\text {m}\text {m}}}\), \(f={200\,{\text {m}\text {m}}}\)) and focused in y-direction (\(f={500\,{\text {m}\text {m}}}\)) into the measurement volume. The lightsheet was \(\approx {8\,{\text {m}\text {m}}}\) in height and \(\approx {250\,{\upmu \text {m}}}\) in thickness at the beam waist (\(1/\textrm{e}^2\)). Pulse energies were reduced to 2.5 mJ to stay within the linear limits of LIF. The latter was verified by an energy scan in the calibration gas by varying the UV energy output using a half-wave retardation plate in front of the second harmonic generation crystal of the tunable dye laser. For this verification, a pyro-electric energy meter (Coherent, EnergyMax J-10-MT 10 kHz) was placed in a partial reflection of one of the lenses to provide a linear energy reference for the measured LIF signal during the energy scan. The lightsheet is guided into the field of view using high-reflective dielectric mirrors. The orientation was chosen parallel to the wall to avoid excessive scattering of radiation at 226 nm, see the next paragraph for details regarding the suppression of scattered laser radiation.
To detect the signal, a CCD camera (LaVision, Imager E-Lite 1.4M) equipped with a UV-sensitive image intensifier (LaVision, low speed IRO) in combination with a 100 mm UV-lens at
f2.8 (Sodern, Cerco 2178) was used. Choosing a detection scheme for near-wall NO-LIF measurements is challenging, as the separation between the (0,0) vibrational band for excitation and emission in the (0,1) band is only
\(\approx {10\,{\text {n}\text {m}}}\). It is desirable to choose a broadband detection of the NO-LIF signal between 235-270 nm for optimal SNR. However, this requires a very steep long pass filter to block the laser radiation, especially considering the finite collection angle of the used optics which one naturally aims to maximize. At 0
\({}^{\circ }\) angle of incidence and parallel light, the used filter (Semrock, LP02-224R) provides
\(>\textrm{OD4}\) at 226 nm,
\(>50\%\) transmission at 234 nm and
\(>80\%\) transmission for wavelength
\(>{244\,{\text {n}\text {m}}}\). However, due to the limited size (1" diameter), the filter was placed between the detection lens and the photocathode of the intensifier in order not to limit the effective aperture of the detection system. Accounting for the resulting finite cone half angle of
\(\approx {17\,\mathrm{{}^{\circ }}}\), blocking at laser radiation wavelength is reduced to
\(\approx \textrm{OD2}\), which proved to be insufficient for wall-normal measurements even including the additional Schott UG-5 filter. These filter characteristics have been simulated using Semrock’s MyLight filter modeling tool.
3
The detection systems were operated at double the repetition rate of the laser systems, allowing to use the ensemble average of every second frame for background correction. Additionally, the individual images were corrected for the inhomogeneous energy distribution in the laser light sheet by referencing to the mean signal acquired in the non-reacting seeded main flow. As this approach only allows to correct the energy profile for the temporal mean, a residual fluctuation of
\(\approx 7\%\) remains which limits the precision of the local equivalence ratio measurement. Absorption effects are accounted for by correcting the LIF signal with an approach similar to the one developed by Heinze et al. (
2011) for quantitative OH measurements. Details may also be found in Greifenstein and Dreizler (
2021a). In short, the local laser intensity can be extracted from the data itself—implicitly assuming a constant fluorescence quantum yield in the beam-wise direction—when the integral absorption is known. For this particular case, the absorption was not directly measured but can be indirectly inferred using the reference case (a) without additional inflow through the effusive wall inlet. For this case, the signal can be assumed to be spatially constant and a calibration factor for the correction was extracted which results in a horizontally flat profile. This calibration factor in units absorption per count is then used for all cases to correct for absorption effects. The preprocessed images are then converted to a local equivalence ratio by means of cubic interpolation to the calibration curve.
2.3.2 OH-PLIF
Qualitative planar laser-induced fluorescence of the OH radical was acquired simultaneously with the other techniques to infer the instantaneous location of the flame front and quenching height, and the impact on the mean OH distribution from partial premixing. The optical setup is similar to previous publications, see, e.g., Zentgraf et al. (
2022b). To generate the LIF signal, a frequency doubled narrowband dye laser (Sirah, PrecisionScan) operated with Rhodamine 6G dissolved in ethanol and optically pumped at 10 Hz repetition rate from a commercial pulsed and frequency doubled Nd:YAG laser (Spectra Physics, PIV400) was tuned to 35210.42
\({{\rm cm}^{-1}}\), corresponding to the
\(\mathrm {Q_2(7.5)+Q_1(9.5)}\) linepair in the
\(A^2\mathrm {\Sigma } \leftarrow X^2\mathrm {\Pi }(v'=1\leftarrow v''=0)\) system of OH. The choice of this linepair is motivated by the low dependency of the Boltzmann factor with respect to temperature, which amounts to
\(\approx 10\%\) between 1400 and 2500 K (Heinze et al.
2011). The spacing between the line centers is 0.44
\({cm^{-1}}\) (Luque and Crosley
1999), which overlap well when accounting for Doppler broadening (
\(\approx 0.26\) cm
\(^{-1}\) at 1800 K), collisional broadening (
\(\approx 0.06\) cm
\(^{-1}\) (Atakan et al.
1997)) and the linewidth of the UV beam (
\(\approx {0.15\,\mathrm{{cm^{-1}}}}\), calculated from fundamental linewidth measurement using wavemeter WSU-30, High Finesse) even at atmospheric pressure.
The UV beam was formed into a laser light sheet by first expanding it using a spherical telescope (\(f={-40\,{\text {m}\text {m}}}\), \(f={200\,{\text {m}\text {m}}}\)) to adjust the divergence and increase the diameter. Subsequently, the expanded beam was cropped to \(\approx {10\,{\text {m}\text {m}}}\) in diameter, using an adjustable iris to improve the spatial homogeneity of laser energy within the beam. The sheet was then focused into the measurement volume with a \(f={300\,{\text {m}\text {m}}}\) cylindrical lens and expanded in vertical direction with a \(f={-300\,{\text {m}\text {m}}}\) cylindrical convex lens to obtain a sheet with \(\approx {35\,{\text {m}\text {m}}}\) height and a waist of \(\approx {220\,{\upmu \text {m}}}\) (\(1/\textrm{e}^2\)). Due to the wall-parallel orientation of the NO measurement volume, the laser beam for OH excitation was guided above the detection system for NO-PLIF and guided into the FWI area at an angle of \(-\)40 \({}^{\circ }\) with respect to the horizontal axis. The pulse energy was adjusted to \(\approx {0.5\,{\text {m}\text {J}}}\) in the measurement volume by using an achromatic half-wave retardation plate in front of the frequency doubling crystal in the laser.
The LIF signal was detected with a combination of a low-speed image intensifier (LaVision, IRO) and an uncooled CCD camera (LaVision, Imager e-Lite 1.4M) equipped with a 100 mm/f2.8 UV transmissive lens (Sodern, Cerco 2178). A
\(2\times 2\) binning was used to improve the signal-to-noise ratio without sacrificing spatial resolution, as the optical performance was limited by the intensifier. The resulting object-plane resolution of a bin was
\(\approx {80\,{\upmu \text {m}}}\). The entire field of view (FOV) spanned an area of
\(\approx {41\,{\text {m}\text {m}}} \times {57\,{\text {m}\text {m}}}\). A narrow bandpass filter was used to suppress radiation from the laser and to isolate the emission of the (1,1) vibrational band between 311-320 nm to reduce signal trapping (Sadanandan et al.
2012b).
As OH-PLIF is used in this study for a qualitative description, the processing was kept comparably simple. Raw OH data were background corrected using the same strategy as for NO-PLIF before dewarping and conversion to physical coordinates using a dot pattern target. The remaining dominant uncertainties result from fluctuation in the integral pulse energy and spatial inhomogeneity of the laser sheet.
2.3.3 Dual-Pump CARS
As the same Dual-Pump CARS setup from Zentgraf et al. (
2022b) was recreated for this study, the description is kept short and the reader is referred to the cited publication. The CARS setup was build up as a planar BOXCARS configuration using two narrowband lasers at 532 nm and 561.7 nm for the two pump beams and a custom built modeless broadband dye laser at 607 nm (
\(\approx {6\,{\text{nm}}}\) FWHM) to probe ro-vibrational transitions of
\(\mathrm {N_2}\) near 2330
\({{\rm cm}^{-1}}\) and
\(\mathrm {CO_2}\) near 1380
\({{\rm cm}^{-1}}\). All pulsed lasers were operated at 10 Hz. Each beam featured an independent control of pulse energy via a half-wave retardation plate and a polarizing beam splitter as well as Galilean telescopes to match the divergence to ensure coincident foci in the measurement volume (MV). The beams were parallelized on the optical table and focused into the MV with a
\(f={300\,{\text {m}\text {m}}}\) plano-convex lens. The resulting MV, as measured with a beam profiler, had a lateral extension of
\(\approx {60\,{\upmu \text {m}}}\) and a beam-wise extension of
\(\approx {2.4\,{\text {m}\text {m}}}\). The lasers were guided into the FWI area parallel to the wall, such that the influence of limited spatial resolution along the principal axis does not influence the data interpretation too severely, as the V-flame is nearly planar and length scales in this direction are assumed to be longer than the measurement volume.
A symmetrically (with respect to the MV) placed \(f={300\,{\text {m}\text {m}}}\) plano-convex lens collimated the CARS signal near 496 nm before the signal was guided to the spectrometer using dichroic long-pass mirrors (cut-off at 550 nm). In addition, a short-pass filter at 550 nm and a notch filter at 532 nm were used to eliminate residual reflections at the pump wavelengths. Depending on the temperature, neutral density filters were used to prevent saturation of the detection camera. Before focusing the CARS signal into the spectrometer (SPEX Industries Inc., 1704, 1m, 2400 lpmm) with a \(f={100\,{\text {m}\text {m}}}\) focusing lens, an achromatic half-wave retardation plate was used to rotate the polarization for optimal dispersion at the grating. A cooled back-light illuminated CCD (Princeton Instruments, Pixis 400) captured the dispersed signal. The readout area was limited to 25 px in height which were hardware binned to obtain a one-dimensional spectrum. The camera was operated at 20 Hz to allow capturing an intermediate background frame between laser shots to correct for dark current. The spectral profile of the Stokes laser was accounted for by normalizing the CARS spectra to a non-resonant signal acquired in Argon at regular intervals during the day.
After background correction and normalization to the non-resonant signal, the CARS spectra were processed using our in-house spectral fitting algorithm
4 to obtain the temperature and
\(\mathrm {CO_2}\) mole fraction from single-shot spectra. The algorithm is an extension of our previously published approach for single-pump
\(\hbox {N}_2\) CARS spectra based on loss-less compressed libraries (Greifenstein and Dreizler
2021b). The underlying model for
\(\hbox {N}_2\) is described therein, while the
\(\hbox {CO}_2\) model of CARSFT (Palmer
1989) was used which is based on the approach published by Hall and Stufflebeam (
1984). Based on measurements in the free flame branch, the
\(\hbox {CO}_2\) mole fraction deviates by
\(\approx 8\%\) with a precision of
\(\approx 15\%\). For the temperature evaluation,
\(\approx 6\%\) and
\(\approx 2.5\%\) are reported.