You are on page 1of 16

Combustion and Flame 235 (2022) 111707

Contents lists available at ScienceDirect

Combustion and Flame


journal homepage: www.elsevier.com/locate/combustflame

Detailed assessment of the thermochemistry in a side-wall quenching


burner by simultaneous quantitative measurement of CO2 , CO and
temperature using laser diagnostics
Florian Zentgraf a,∗, Pascal Johe a, Matthias Steinhausen b, Christian Hasse b,
Max Greifenstein a, Andrew D. Cutler c, Robert S. Barlow d, Andreas Dreizler a
a
Technical University of Darmstadt, Department of Mechanical Engineering, Reactive Flows and Diagnostics, Otto-Berndt-Str. 3, Darmstadt 64287, Germany
b
Technical University of Darmstadt, Department of Mechanical Engineering, Simulation of Reactive Thermo-Fluid Systems, Otto-Berndt-Str. 2, Darmstadt
64287, Germany
c
The George Washington University, Washington DC, USA
d
Barlow Combustion Research, Livermore, USA

a r t i c l e i n f o a b s t r a c t

Article history: This study focuses on exploring the thermochemistry in flame-wall interaction (FWI) for fully pre-
Received 18 March 2021 mixed side-wall quenching of a laminar, atmospheric-pressure dimethyl ether flame at equivalence ratio
Revised 17 August 2021
 = 0.83 by simultaneous measurement of CO2 and CO mole fractions and gas phase temperature T . The
Accepted 18 August 2021
applied laser diagnostics are dual-pump coherent anti-Stokes Raman spectroscopy (DP-CARS) targeting
Available online 5 September 2021
N2 and CO2 , laser-induced fluorescence of CO and OH, as well as thermographic phosphor thermometry.
Keywords: The extension to DP-CARS to study FWI processes is the first of its kind, previous studies only provided
Flame-wall interaction (CO,T ) measurements. The laser diagnostics are benchmarked and calibrated to an adiabatic test case
Dimethyl ether and assessed in accuracy and precision. Subsequently, the approach is used to measure the thermochem-
Dual-pump CARS istry close to a quenching wall. The nominal flame-to-wall distances from the experiment well match
LIF the numerical simulation data with a marginal offset of 20 μm. Conditioning the thermochemical data
Thermochemistry
with respect to the instantaneous quenching point, named quenching-point conditioning, enables a novel
Differential diffusion effects
tracing of the wall-parallel chemistry evolution across the quenching location. The study provides the
first comparison of experimental three-scalar measurements (CO2 ,CO,T ) with two-dimensional (2D) fully-
resolved chemistry and transport (FCT) simulations. The validation of numerical simulations can now rely
on the three scalars (CO2 ,CO,T ) instead of the two scalars (CO,T ) in past studies. The evaluation reveals
that this novel three-scalar measurement allows highly sensitive probing of the thermochemical states
and is clearly superior to the previously applied two-scalar approach. CO2 is less affected by the quench-
ing wall compared to CO. Differential diffusion effects are experimentally confirmed by comparison to
2D-FCT, with the (CO2 , T ) state space being more sensitive than (CO,T ). As the experimental methodology
proved feasible for laminar operation, a transfer to turbulent cases, where the numerical analysis using
direct numerical simulations (DNS) including FCT is limited, appears promising.
© 2021 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction and motivation to depend on the use of high-energy-density chemical energy car-
riers. These include aviation and ground freight using diesel en-
The emerging consequences of human-induced climate change gines, where sustainable synthetic chemical energy carriers, either
and stricter legal limits for exhaust emissions demand new ap- e-fuels or fuels derived from biomass, will play an increasing role
proaches to the energy conversion of carbonaceous fuels by com- in the future [1,2]. Consequently, there is a major fundamental re-
bustion. While it is possible in some areas of society to replace the search interest in their combustion characteristics and related pol-
use of carbonaceous fuels with other concepts, such as the elec- lutants, such as carbon oxides (COx ), nitrogen oxides (NOx ), un-
trification of urban individual transport, other areas will continue burned hydrocarbons, aldehydes, and soot. For example, dimethyl
ether (DME, CH3 –O–CH3 ) and oxymethylene ethers (OMEs, CH3 -
O-[CH2 -O]n -CH3 ) are considered to be promising diesel fuel surro-

Corresponding author. gates featuring a reduced soot precursor formation [2–4] and a sig-
E-mail address: zentgraf@rsm.tu-darmstadt.de (F. Zentgraf). nificant NOx reduction potential [2]. For its gaseous state at ambi-

https://doi.org/10.1016/j.combustflame.2021.111707
0010-2180/© 2021 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

ent conditions, DME is considered less advantageous compared to temperature and relative hydrocarbon number densities in SWQ.
conventional liquid fuels in the transport sector [2]. Nevertheless, Saffman [21] applied a comparable setup for temperature and
DME is a representative of the partially oxygenated hydrocarbons hydrocarbon mole fractions. However, temperature and species
that are promising e-fuel pioneers, and it is related to the OMEs in concentrations were not measured simultaneously. Mokhov et al.
its chemical structure. Due to its low molecular complexity, DME [22] assessed profiles of the gas phase temperatures (by a line re-
is well suited for fundamental research. versal method using a spectral lamp), OH (absorption and laser-
A further important topic in applied combustion devices that induced fluorescence, LIF) and CO (by two-photon LIF) number
is not yet fully explored and understood is the quenching of pre- densities as well as velocities (by laser Doppler anemometry, LDA).
mixed flames at cold solid surfaces, called flame-wall interaction They used a SWQ configuration, but with the quenching wall posi-
(FWI). FWI is broadly relevant to combustion technologies and en- tioned fully in the exhaust gas of the burner at equilibrium condi-
vironmental protection because FWI involves non-desirable pro- tions. Furthermore, it was not stated whether or not the data were
cesses, like wall heat loss or incomplete combustion causing pollu- measured simultaneously. Fuyuto et al. [23] performed a compre-
tants and soot. Especially, for more complex high-molecular-weight hensive study in a SWQ device measuring two-dimensional distri-
fuels, it is well known that the molecular structure modifies FWI butions of the gas phase temperature (by multi-line NO-LIF ther-
and particularly the evolution of intermediate hydrocarbons, as for mometry), OH and CH2 O (both by LIF), as well as CO (by two-
example demonstrated early by Kiehne et al. [5] for propane or photon LIF). However, the data were not measured simultaneously,
Hasse et al. [6] for iso-octane. This results in the pressing need to and species distributions were qualitative or semi-quantitative.
explore FWI for novel e-fuels, where different intermediate hydro- More recent studies on thermochemistry in FWI simultaneously
carbons are formed and other chemical reaction pathways may be quantified the gas phase temperature T by coherent anti-Stokes
relevant. Raman spectroscopy (CARS) and the CO mole fraction XCO by two-
The major challenges in numerical studies on FWI are to predict photon LIF [16,24,25] in HOQ and SWQ configurations. This en-
the impact of the non-adiabaticity to (1) the chemical progress by abled a shot-to-shot correlation of T to the corresponding XCO .
reaction and (2) the convective and especially diffusive transport However, both parameters are influenced by the enthalpy loss at
in the vicinity of the quenching wall. In fully-resolved chemistry the quenching wall. An additional major reaction species that is
and transport (FCT) simulations,1 applied for example in Ganter less prone to the non-adiabaticity of the FWI process would be
et al. [7], Efimov et al. [8], Steinhausen et al. [9] for laminar flames desirable, e.g., a product species like CO2 or H2 O. So far and to
and in Gruber et al. [10], Ahmed et al. [11] for turbulent, generic the authors best knowledge, the work of Bohlin et al. [26] is the
FWI configurations, the transport equations of combustion-related only experimental multi-parameter study simultaneously probing
species is solved directly. This approach is very accurate, however, six major combustion species (N2 , O2 , H2 , CO, CO2 , and CH4 ) by
due to the high computational costs, FCT simulations of turbulent application of an ultrabroadband CARS approach for assessing the
flames are mostly suitable for generic configurations. Therefore, thermochemical state during FWI. Thus, for FWI experiments, there
it is necessary to reduce the combustion kinetics to enable the is a further need to explore the thermochemistry quantitatively
simulation of technical applications. Thereby, chemistry manifolds and simultaneously beyond (CO,T ) using a complementary diag-
combine the accuracy of the FCT simulations with low computa- nostical setup.
tional costs. For FWI, different chemistry manifolds have been de- Recently, the impact of DME fuel in FWI was addressed exper-
veloped including Flamelet-Generated Manifolds (FGM, applied in imentally by Luo and Liu [27] in a non-premixed HOQ configura-
Ganter et al. [7], Efimov et al. [8], Steinhausen et al. [9]), Quench- tion, focusing on soot formation, as well as by Kosaka et al. in pre-
ing Flamelet-Generated Manifolds (QFM, applied in Efimov et al. mixed SWQ with scope on thermochemistry by CO and gas phase
[8], Steinhausen et al. [9]) or Reaction-Diffusion Manifolds (REDIM, temperature [16] and the local heat release rate [28]. A numeri-
applied in Ganter et al. [7], Steinhausen et al. [9], Straßacker et al. cal study on premixed DME combustion in SWQ was presented by
[12,13]). To enable physical meaningful simulations of complex Steinhausen et al. [9]. The sparse treatment of DME as a generic e-
combustion devices, it is essential to validate the computational fuel in FWI processes in the literature calls for further fundamental
models employed, both in the DC simulations, as well as the re- studies to understand the phenomenological behavior.
duced chemistry approaches. Thereby, generic experiments play a The presented experimental approach expands beyond the pre-
crucial role by providing insights into the underlying physics of vious state-of-the-art on laser diagnostics in FWI. It aims to ad-
FWI. dress the open scientific requirements on FWI outlined in the
Experimental studies on two categories of FWI in premixed above paragraphs, namely to (1) extend previous diagnostic ap-
flames – head-on quenching (HOQ, flame propagates normal to proaches for measuring three scalars simultaneously, (2) gain fur-
wall) and side-wall quenching (SWQ, flame propagates parallel to ther insight into wall-bounded DME combustion and (3) provide
wall) – were recently reviewed by Dreizler and Böhm [14]. For an more comprehensive validation data for CFD. The specific scopes
experimental assessment of FWI, laser diagnostics are considered of this study are:
most promising to achieve spatially and temporally resolved data
close to a solid surface without altering the local flow and reaction • Gain more insights into the thermochemical states during FWI
conditions. Experimental studies on FWI can focus on various as- in an atmospheric, fully premixed SWQ configuration. The pre-
pects of heat transfer, fluid mechanics, or chemical reactions rele- vious approach of simultaneously measuring CO and T is ex-
vant to the quenching phenomenon. For example, investigations of tended by additionally measuring CO2 such that the following
quenching distance [15,16], wall heat flux [15,16], or the interaction parameters were captured simultaneously: (1) CO2 mole frac-
of quenching with the flow field [17–19] are prominent and com- tion XCO2 , (2) CO mole fraction XCO , (3) gas phase tempera-
monly reported. In contrast, the area of detailed near-wall thermo- ture T , (4) qualitative OH radical distribution. The post-flame
chemistry, i.e., the coupling of heat and temperature with chemical wall temperature is additionally assessed by laser diagnostics
reaction, has previously been covered only occasionally. for boundary condition control (for wall temperature reference
In the early 1980s, Clendening et al. [20] used a sponta- at higher wall temperatures). Here, CO2 is selected because, as
neous Raman-scattering setup to measure profiles of the gas phase a product species, it is expected to be less affected by the en-
thalpy loss through the quenching wall, and its mole fraction
should rise monotonically with progressing reaction. This also
1
also referred to as detailed chemistry (DC) in the cited literature. makes CO2 a suitable component of a progress variable defini-

2
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 1. Schematic side view of the atmospheric SWQ burner facility. Zoomed views of the quenching region outline the relevant field-of-views (FOVs) and the near-wall
measuement locations. Numbers indicate dimensions in millimeters.

tion for reduced chemistry manifolds for both DME [9] and CH4 ated under laminar flow conditions. However, for laminar condi-
[7]. tions, a remaining unsteadiness of the flame existed that has been
• A general assessment of the feasibility, accuracy, and precision attributed to Helmholtz resonances [25]. The present measure-
of the combined approach using dual-pump CARS, targeting N2 ment cases exhibited fluctuation of the quenching point of about
and CO2 , with CO-LIF to simultaneously measure (CO2 ,CO,T ). ±310 μm (±1σ ) along the wall. These values are consistent with
Thus, no parameter variation is performed, and the focus is at past measurements [25]. Remaining fluctuations were corrected for
one selected DME/air mixture at  = 0.83, ∼ 60 ◦ C wall tem- using the instantaneous OH-LIF signal as described in Section 2.2.3.
perature with laminar inflow. After exiting the nozzle, a V-shaped flame was stabilized by a ce-
• Focus on DME as a relevant fuel from renewable sources, with ramic rod (1 mm diameter). One branch of this flame quenched
main emphasis on its differential diffusion effects in near-wall at a temperature-controlled, stainless steel wall. The wall surface
combustion, which are already much more pronounced than for is curved in the xy-plane (see top view in Fig. 2) with a radius
CH4 at atmospheric pressure. As the presented experiments are of 300 mm to allow the laser diagnostics to probe close to the
limited to atmospheric pressures, more complex features like wall. For orientation, this manuscript labels directions using the
low temperature heat release (LTHR) and negative temperature prefixes introduced by the coordinate system in Fig. 1. The y-axis
coefficients (NTC) are beyond the scope of this study. denotes the wall-normal and z the streamwise direction parallel to
the wall, while x gives the out of plane direction.
2. Experimental and numerical approach The wall was temperature controlled by either flushing
ambient-temperature water or heated thermal oil through the
In the following four subsections, the burner facility is de- cooling duct. An air co-flow shielded the central premixed main
scribed, the implemented laser diagnostics are outlined in detail flow. The burner was operated at atmospheric conditions. The
with a brief insight into data processing, the diagnostic methods Reynolds number of the main flow was maintained at approxi-
are benchmarked to an adiabatic reference case, and the numerical mately 5900. This value is based on the hydraulic diameter of the
simulation data used for comparison are documented. nozzle exit (approx. 40 × 40 mm2 ) and considers the physical con-
ditions at 20 ◦ C, relevant for the nozzle exit. As the burner was
2.1. Burner facility designed with a convergent (Morel) nozzle followed by a short
straight section, a nearly flat, laminar velocity profile with thin
Measurements were performed within an established side-wall boundary layers results that is not yet transitional at the burner
quenching (SWQ) burner facility [16,19,25,28] shown in Fig. 1. A exit. Three sheath thermocouples (type K, 1.5-mm diameter) on
fully premixed fuel-air mixture was side-fed to the inlet plenum the center line of the wall were used to monitor the temperature
of the burner and subsequently guided through a combination of of the quenching wall surface. The tip of each thermocouple was
coarse meshes, fine meshes, and a honeycomb structure to ho- placed approximately 200 μm to 400 μm below the solid surface.
mogenize the inflow. The mixture then passed through a converg- Calibrated mass flow controllers (Bronkhorst High-Tech B.V.) were
ing (Morel) nozzle, resulting in a nearly top-hat velocity profile used to define the flow conditions and equivalence ratios. For DME
at the nozzle exit. An optional turbulence generating grid could fuel at  = 0.83 (see operation condition in Table 1) the uncer-
be placed inside the nozzle. Without this grid, the burner oper- tainty in the equivalence ratio is δ  = 0.01 derived via error prop-

3
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 2. Schematic of the applied laser diagnostic setup subdivided in a side and front view for the laser beams as well as a top view for the detection arrangement. Numbers
on lenses indicate the focal length in millimeters.

Table 1
Operating conditions of the SWQ burner in this study. Twall expresses the minimum to maximum thermocouple
temperatures during operation, temporal average was 332 K with 2.5 K standard deviation.

fuel  (−) inflow Re (−) ujet (m/s) ucoflow (m/s) Ptherm (kW) Twall (K)

DME 0.83 laminar 5900 2.2 2.1 6.1 327–338

agation from the fuel and air flow rate deviations of the calibration. inal) discussed here. For the higher wall temperature cases, the
The entire burner facility was mounted on a computer-controlled thermocouple temperatures (mean ± standard deviation: 626 K ±
traversing system (x-, y-, z-axes). 6 K) agreed well with TPT measurements (670 K ± 0.7 K). In this
To perform the measurements, the entire burner assembly was higher wall temperature case, the average TPT result (or measure-
moved, while all diagnostics remained fixed. In the zoomed view ment) is 44 K higher (∼ 7% in absolute value) than the thermocou-
in Fig. 1, the fields-of-view (FOV) of the individual laser diagnos- ple values, which is physical as the TPT is measured at the sur-
tics and the domain of the numerical simulation are shown. The face, while the thermocouples are embedded just below the sur-
OH-LIF covered a FOV of (y, z ) = (24 mm, 32 mm ), while the face. However, the thermocouple measures do sufficiently express
simulation domain featured (y, z ) = (6 mm, 30 mm ). For DP- the wall surface temperature within 7% uncertainty and are ex-
CARS and CO-LIF, a 2D field is scanned point by point around the pected to accurately report the wall temperature also for the lower
quenching area as outlined by the sketch on measurement loca- wall temperature case discussed here as temperature gradients are
tions in Fig. 1. For the y-direction, 13 horizontal locations were lower compared to the higher wall temperature case.
acquired (0.1 to 1 mm with 0.1 mm increment and at 1.5, 2 and
4 mm wall-distance). This was done at five different z-locations 2.2. Laser diagnostics
centered around the quenching height (the z-location where the
flame quenches). Individual resolutions are given in Section 2.2. To study the SWQ phenomena, a comprehensive set of laser di-
Please note that the coordinate system embedded in Fig. 1 is for agnostics was applied. The main features of the laser optical setup
directional purpose only and does not show the origins of these are provided in Fig. 2. In total, four diagnostics were implemented
axes. For data analysis, a relative coordinate system was used with simultaneously: (1) dual-pump coherent anti-Stokes Raman spec-
z = 0 mm placed at the quenching height (individually by condi- troscopy targeting N2 and CO2 , (2) two-photon laser-induced fluo-
tioning for each recorded sample using simultaneous OH-LIF imag- rescence of carbon monoxide, (3) planar laser-induced fluorescence
ing) and y = 0.1 mm at the nominal wall-closest position. of the hydroxyl radical and (4) thermographic phosphor thermom-
A comprehensive parameter variation was done during the etry. The details of the individual approaches are outlined in the
measurements using different equivalence ratios, fuels, wall tem- following.
peratures, and both laminar and turbulent flow cases. For demon-
stration, this study focuses on evaluating the capabilities of the ex- 2.2.1. Dual-pump coherent anti-Stokes Raman spectroscopy
tended laser diagnostics for assessing the thermochemistry in SWQ (DP-CARS)
for only one set of operation conditions summarized in Table 1. The DP-CARS approach was applied to provide a quantitative
The wall temperature range in Table 1 resulted from the thermo- measure of both the gas phase temperature via N2 , as a very ro-
couples located underneath the wall surface downstream of the bust temperature indicator in combustion environments, and the
quenching height. A benchmark with the embedded thermographic CO2 mole fraction XCO2 . The concept of DP-CARS was introduced
phosphor thermometry (TPT) was done for the higher wall temper- by Lucht [29] to simultaneously measure the spectra of two species
ature cases but not for the cold wall case at ∼ 60 ◦ C (333 K nom- at a time with one detector. In this approach, a narrow-band dye

4
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

laser was used as one pump laser along with a frequency-doubled 0.5%, 10% transmission or no filter) in case the DP-CARS signal
Nd:YAG laser as the second pump laser and a broadband dye laser saturated the detector, e.g., at ambient temperatures. The sig-
as the Stokes laser [29]. For measuring N2 and CO2 using a 607 nal was focused to the slit of the spectrometer (SPEX Indus-
nm broad-band Stokes laser, N2 (Raman shift around 2330 cm−1 ) tries Inc., 1704, 1 m spectrometer, 2400 lines/mm grating) using
was pumped by 532 nm and probed around 560 nm, while CO2 a spherical lens ( f = +100 mm). The spectra were recorded us-
(Raman shift around 1380 cm−1 ) was pumped around 560 nm and ing a cooled, backside-illuminated CCD camera (Princeton Instru-
probed by 532 nm [30]. The major advantages of DP-CARS were ments, Pixis 400, 1 ms exposure time). In the horizontal wave-
that both species are probed exactly at the same time and loca- length direction, all 1340 pixels were used, while in the verti-
tion, as their CARS signals are resulting from the same three laser cal direction only 25 pixels were used and fully hardware binned
beams and the resulting spectra are nearly in the same wavelength prior to readout. All optics used in the setup had anti-reflective
range, allowing them to be imaged on one detector [31]. These coatings. For correction purpose, non-resonant DP-CARS spectra
advantages yielded an excellent correlation behavior of the two were recorded in pure argon gas. This was done several times
measured species, which is of great importance for proper thermo- during each measurement day, periodically between operation
chemistry studies. The experimental implementation in this study points.
followed the outlines of Lucht et al. [31]. The processing flow to derive gas phase temperature and CO2
Two custom-modified Q-switched, frequency-doubled Nd:YAG mole fraction from the DP-CARS raw spectra is outlined in Fig. 3.
lasers (Spectra-Physics PIV 400, 10 Hz, 5–8 ns pulse width, in Initially, the raw spectra were pre-processed and their intensity
which the PIV-overlap was removed and individual cavities were was corrected by a mean background subtraction as well as a
split, resulting in two independent lasers at 532 nm) were used normalization to a mean non-resonant signal. The square root of
as excitation sources. The first Nd:YAG laser was split into two the intensity-corrected spectra was taken, and a dispersion cali-
beam paths. One part of the beam served as first pump wave- bration was applied to assign the correct wavenumbers to the in-
length at 532 nm and was guided to the DP-CARS overlap optics tensities. For the dispersion calibration, experimental data recorded
directly. The second beam path was used to pump the custom- in an adiabatic flat-flame burner was fit to a simulated spectrum
built broad-band, modeless Stokes laser (design principle by Ewart under identical conditions by a least square scheme, optimizing
[32]). This laser was operated with a dye mixture of Rhodamine the wavenumber expansion and shift (a linear correction). Sub-
610 and 640 resulting in a spectral profile centered around 607 nm sequently, these spectra were processed by a spectral fitting al-
having ≈ 8 nm FWHM. The second Nd:YAG laser served as optical gorithm [33] with the gas phase temperature, N2 and CO2 mole
pumping source for a narrow-band dye laser (Radiant Dyes Laser fractions as well as the wavenumber shift as free fitting parame-
& Accessories GmbH, NarrowScan) operated with Rhodamine 590 ters. This procedure was applied to the single-shot spectra directly,
dye at 561.7 nm, serving as second pump wavelength. Note that yielding quantitative single-shot gas-phase temperature and CO2
561.7 nm is used here instead of 560 nm in Roy et al. [30] or mole fraction.
561 nm in Lucht et al. [31] to achieve a good spectral separa- Due to its lower concentration and quantum mechanical charac-
tion of the N2 and CO2 . The three individual beams (532, 561.7 teristics, the CO2 signal was weaker than the N2 signal at elevated
and 607 nm) were energy-controlled by combinations of half-wave temperatures approaching the burnt gas. To compensate for a bias
plates and polarizing beam splitter cubes and subsequently colli- originating from weak CO2 signals, e.g., caused by beam steering,
mated using telescopes with spherical lenses. The individual beams appropriate post-processing steps were performed. First, an em-
passed through delay lines and were overlapped according to their pirical thresholding procedure was applied to judge the CO2 sig-
phase matching conditions (α532 ≈ 2.0◦ , α561 ≈ −1.9◦ , α607 ≈ 2.2◦ ; nal quality from the most prominent CO2 feature at ≈ 2373 cm−1
angles with respect to the principal beam direction) using a planar (i.e., the 000 0–100 0 band of CO2 ). All data with a
spectral fit fea-
BOXCARS arrangement. Finally, they were guided to the burner fa- turing an absolute baseline to peak difference  ICO2 ≥ 0.6 were
cility and focused to the region of interest (ROI) using a spherical

considered for further analysis, while data with  ICO2 < 0.6 was
lens ( f = +300 mm). The average pulse energies in the ROI were excluded (see visualization in Fig. 3). The empirical thresholding
set to ≈ 25 mJ (532 nm, 1% coefficient of variation (CV)), ≈ 24 mJ criterion was validated, based on the (CO2 , T ) state space evo-
(561.7 nm, 3% CV) and ≈ 13 mJ (607 nm, 2% CV), respectively. These lution of data from the free branch of the laminar V-stabilized
average pulse energy values are reduced by ∼ 15% from those used flame and comparison with the corresponding numerical simula-
by Lucht et al. [31]. The CARS probe volume, i.e. the measurement tion. This is outlined in the final plot in the DP-CARS processing
resolution, spans about (y, z ) = (67 μm, 54 μm) (1/e2 value flow
of ICARS ∝ I532 · I561 · I607 intensities on beam monitor) in the cen-
 in Fig. 3, where green symbols visualize the excluded  data
( ICO2 < 0.6, 20% of samples) and black symbols ( ICO2 ≥ 0.6,
tral plane location at x = 0 mm, while the interaction length in x- 80% of samples) depict the retained data. From these observations
direction is estimated to x = 2.4 mm (1/e2 value of non-resonant it is evident that the criterion is effective in removing unphysi-
signal in a 100 μm thick glass plate). cal samples from the (CO2 , T ) state space. Note that for the near-
The coherent DP-CARS signal at around 496 nm was collected wall data presented in Section 3, only 10% of the samples were
using another spherical lens ( f = +300 mm) and was split from excluded by this procedure, indicating a high quality of the ac-
the overlapping 561.7 nm radiation by using three lowpass dichroic quired data. Second, an additional calibration was applied to the
mirrors (Thorlabs Inc., DMLP550L, 50% transmission/reflection at data after thresholding. Again, this is done using the known evo-
550 nm). While ≈ 90% of the 561.7 nm beam was transmitted lution of (CO2 , T ) in the adiabatic reference case. For this purpose,
on each mirror, the DP-CARS signal was reflected. Afterwards, the CO2 mole fractions of both the simulation and experiment are
an achromatic half-wave plate (Thorlabs Inc., AHWP10M-600, for curve-fitted over temperature. A calibration coefficient as a func-
40 0–80 0 nm) was used to optimize the polarization of the DP- tion of temperature can be derived from the fitted polynomials as
CARS signal with respect to the spectrometer. Two interference CCO2 ,calib (T ) = f itCO2 ,sim (T )/ f itCO2 ,exp (T ). This coefficient is multi-
filters were used to remove remaining spectral residuals super- plied to the experimental CO2 mole fractions using the correspond-
imposed on the DP-CARS signal: (1) a notch filter at 532 nm ing temperature for correction towards the simulation values. The
to remove scattered light, (2) a short pass filter (Thorlabs Inc., calibration further aided in improving the average accuracy of the
FESH0550, cutoff wavelength 550 nm) to remove the rest of the CO2 mole fractions from 9% (not calibrated, see final plot for DP-
transmitted 561.7 nm intensity. A filter wheel was used to op- CARS in Fig. 3) to 2% (calibrated, see Fig. 4(a) and (f)). This is fur-
tionally put neutral density filters into the signal path (options: ther discussed in Section 2.3.

5
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 3. Flowchart summarizing the processing methodology for DP-CARS, CO-LIF and OH-LIF.

2.2.2. Two-photon laser-induced fluorescence of carbon monoxide tal dye laser emission at 690.27 nm was frequency tripled to
(CO-LIF) 230.09 nm in the UV using a second harmonic and third harmonic
The CO mole fraction XCO was assessed using two-photon CO- generation stage (i.e., 345.135 nm was mixed with 690.27 nm to
LIF. The excitation of CO molecules was done within the Hopfield- 230.09 nm). The fundamental wavelength was continuously mea-
Birge bands at 115.05 nm using two-photons at 230.09 nm for the sured by a wavelength meter (HighFinesse GmbH, WS6) and used
transition B1  + ←← X 1  + (0, 0 ), while the emission of the transi- to control the grating position of the dye laser. This control loop
tion B1  + → A1 u (0, 1 ) in the Ångström bands was detected [34]. guaranteed a constant excitation wavelength at 230.09 nm within
This scheme is well established in combustion diagnostics and the 0.025–0.036 cm−1 (difference between maximum and minimum
implementation of this approach in this study follows the outlines average wavenumber measured over a measuring day). The UV
of [16,24,25,35] for previous FWI investigations. beam was guided to the burner facility using high reflectivity di-
A Q-switched, frequency-doubled Nd:YAG laser (Spectra-Physics electric mirrors. A combination of half-wave plate and Glan-laser
GCR 4, 10 Hz, 5–7 ns pulse width (FWHM)) was used to pump double escape polarizer (Lattice Electro Optics Inc., PCGL-a-BBO-
a narrow-band dye laser (Sirah Lasertechnik GmbH, PrecisionScan, 04) was used to control the energy. The usual average pulse en-
2400 lines/mm) operated with Pyridine 1 dye. The fundamen- ergy within the ROI was ≈ 0.7–0.9 mJ (5% CV). A fused silica glass

6
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

plate (≈ 10 mm thickness) was placed in the beam path to gener- centered around the DP-CARS probe volume. Then, a Gaussian was
ate both an energy and beam profile reference. The front side re- fit in z-direction, i.e., orthogonal to the principal beam direction;
flection was deflected to an energy meter (Coherent Inc., EM USB the intensity beyond the 1/e2 -bounds was subtracted as baseline
J-10MT-10KHZ), where the pulse energy was measured for a shot- intensity IBL (i ), and the intensity within the 1/e2 -bounds of the
to-shot energy correction. The backside reflection of the glass plate fitted Gaussian was subsequently integrated, ending up in one in-
was guided to a beam monitor (DataRay Inc., WinCam D) equipped tegrated intensity value ICO (y, z, i ) for each raw image i. This in-
with a UV-converting beam profiler (Star Tech Instruments Inc., tegrated intensity was subsequently corrected using Eq. (1) for
BSF23). Beam profiles were only measured occasionally to check shot-to-shot energy fluctuations (I), local density (II), and geomet-
location and intensity distribution. Note that the beam profile of rical influences (III), as previously described by Jainski et al. [25],
the laser didn’t change during the measurements, but the overall and is finally turned into quantitative values using a calibration
pulse energy did slightly drop in its average by ∼10–18% over a polynomial recorded in the free flame branch (IV). Here, the non-
measuring day. The beam was focused to the ROI using a spherical linear energy correction (I) followed a laser energy (E) dependency
lens ( f = +300 mm) with a beam diameter of (y, z ) = (260 μm, ∝ E n (i ) with n = 1.68 for the case presented in this study (see
200 μm) (1/e2 value of Gauss-fitted profiles). Since the applied CO- Table 1). It was measured in-situ within the exhaust gas of the V-
LIF is a two-photon process, the probe volume size is not sim- shaped flame by varying the laser pulse energy. The value of the
ply given by this beam diameter but rather determined by the exponent n < 2 confirms the presence of photo-ionization of CO
area with laser intensity In , with the exponent n ≤ 2 considering molecules, that promotes a signal less dependent on the quench-
the laser fluence dependency, impacted by photo-ionization (here: ing environment, as discussed below. By decreasing the laser en-
n = 1.68). The effective probe volume diameter, i.e., dimensions in ergy, n was chosen above the value in previous studies, like n =
y- and z-direction, corresponding to the laser intensity I1.68 rele- 1.24 in Jainski et al. [25], to minimize photolysis-generated CO
vant for CO-LIF here is (y, z ) = (140 μm, 140 μm) (1/e2 value from CO2 that could bias the combustion-generated CO2 quan-
of Gauss-fitted profiles). In x-direction, the probe volume for data tification via DP-CARS. Exponent values n ≈ 2, as achieved using
analysis spans 2 mm in length to overlap well with the DP-CARS femtosecond lasers [36], are hardly realizable using the applied
probe volume. To finally combine and overlap the CO-LIF beam nanosecond approach. The instantaneous temperature values T (i )
with the DP-CARS, a small rectangular mirror (10 × 20 mm2 ) was from simultaneous DP-CARS divided by a reference temperature
placed between the upper (532 and 607 nm) and lower (561.7 nm) T0 = 300 K were used to correct for the local density (II). A geo-
DP-CARS beams directly after their transmission through the focus- metric correction (III) considered the influence of the quenching
ing lens. wall distance on the CO signal (due to reflections of the stainless-
The CO-LIF emission was detected by an intensified relay op- steel wall, the intensity is slightly rising when approaching the
tic (IRO, LaVision GmbH, low speed IRO) gated at 100 ns to sup- wall). The correction function W (y, z ) was generated by operat-
press background and flame luminosity and recorded using a CCD ing the burner with a fully homogeneous calibration gas mixture
camera (LaVision GmbH, Imager E-Lite 1.4M) with 10 ms exposure (1% CO, 99% N2 ). For wall-distances of y ≥ 5 mm, W (y, z ) = 1. For
time to fully capture the phosphor decay of the IRO and a 4 × 4 final quantification, a calibration polynomial (IV) was derived by
hardware binning. The IRO was operated in the linear regime. A fitting corrected experimental (CO,T ) data (on T = 50 K incre-
reduced ROI of 70 (horizontal) by 30 (vertical) bins centered at ments) recorded in the free flame branch to a non-adiabatic sim-
the CO focus was used, corresponding to 3.5 × 1.5 mm2 (bin size ulation (see Section 2.4). This was done in the same manner as
≈ 50 μm). The signal was collected with a customized imaging op- for (CO2 , T ) in DP-CARS above and resulting in a calibration coef-
tics consisting of an achromatic lens (Qioptiq Photonics GmbH & ficient CCO,calib (T (i )) = f itCO,sim (T (i ))/ f itCO,exp (T (i )) as function of
Co. KG, G3222670 0 0, f = +160 mm, 80 mm diam.), an interference the instantaneous temperature. Such a calibration procedure cov-
filter (Laser Components GmbH, LC-HBP 490/20-75, >95% trans- ers influences from varying Boltzmann fractions, quenching envi-
mission at 480–500 nm) and an objective lens (Canon, EF 135 mm ronment and ionization, as pointed out by Mann et al. [24]. The
f/2L USM, focused to infinity). The entire imaging system was pro- different chemical composition between the adiabatic and non-
tected from thermal radiation by a heat shield cooled with a N2 - adiabatic flame branch at a given temperature causes a change
flow. in the quenching rate which impacts the calibration. However, for
The CO-LIF raw signal was turned into quantitative mole frac- the given conditions, the quenching rate Q and the ionization rate
tions using a well-established calibration procedure [16,24,25]. R are in the order of 109 s−1 as estimated with an in-house CO-
However, the procedure was advanced for the application in the LIF simulation code based on [37] with quenching data from [38],
current setup and is outlined under CO-LIF in Fig. 3. The quantifi- an ionization cross-section of 1017 cm−2 [39] and ∼ 0.3 GW/cm2 as
cation to CO mole fractions XCO (y, z, i ) was done by computed from the pulse energy, pulse duration and area of the
focal spot. For this estimation, a CH4 flame at 0.83 equivalence ra-
1 T (i ) 1 f itCO,sim (T (i ))
XCO (y, z, i ) = ICO (y, z, i ) · · · · (1) tio was simulated as there is no literature data on the quenching
a + b · E n (i ) T0 W (y, z ) f itCO,exp (T (i ))
          cross section of DME. Effectively, as R and Q are in the same or-
(I ) (II ) (III ) (IV ) der of magnitude, a change in Q loses its impact on the fluores-
cence quantum yield by approximately a factor of two. This ap-
with respect to the probed y- and z-location as well as sample
proach thus allows transferring the calibration to different chemi-
number i. Roman numbering (I) to (IV) summarize individual cor-
cal environments as long as the temperature is measured simulta-
rection steps with their impact and parameters explained in the
neously. For even higher fluences, the LIF signal becomes nearly in-
following. Initially, a pre-processed CO-LIF intensity ICO (y, z, i ) was
dependent of quenching. This has been more intensively discussed
derived by
in Greifenstein et al. [35]. This argument has also been made in
 Voigt et al. [40] that in this regime of intensities, the LIF signal be-
ICO (y, z, i ) = Iraw (x, z, i ) − I¯BG (x, z ) − IBL (i ) (2) comes independent of the composition-dependent quenching even
z x
at higher pressures, with associated higher quenching rates. How-
from individual raw images Iraw (x, z, i ). An average background ever, this increases issues with photolytic production of CO by dis-
I¯BG (x, z ) was subtracted (recorded between consecutive laser sociation of CO2 [36,41]. The used fluence presents a carefully se-
pulses, flame on) and the resulting signal was integrated in x- lected compromise between these effects to obtain the best possi-
direction, i.e., principal beam direction, within a 2 mm region ble accuracy.

7
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

2.2.3. Planar laser-induced fluorescence of hydroxyl radical (OH-PLIF) ent sensitivity in the detection setup, this value may need to be
The hydroxyl radical OH was visualized qualitatively by a pla- adapted. Finally, all DP-CARS/CO-LIF data were conditioned to the
nar LIF (PLIF) setup to determine the flame front position and the instantaneous quenching point by this scheme, to correct for the
quenching location. In OH, transitions in the A2  − X 2  system remaining unsteadiness of the flame.
are used for laser excitation and the resulting detection [42]. Past
studies frequently applied the UV radiation of excimer lasers, for 2.2.4. Thermographic phosphor thermometry (TPT)
example at 248.45 nm [23] or 291.5 nm/313 nm/268.5 nm [43], Please note that the benchmark values from TPT were not used
and used various different transitions for exciting OH, while recent for the low wall temperature case at 60 ◦ C discussed here but for
studies on FWI excited the Q1 (6) transition around 283 nm using controlling the higher wall temperature cases only. The values are
the frequency doubled radiation of dye lasers [16,25], to only name not used for correlation with thermochemistry but are used to
a few. The latter scheme using dye lasers is applied here as the control the boundary conditions. Thus, the TPT diagnostic is an
presented study aims at simply visualizing the qualitative spatial essential part of the simultaneous thermochemistry measurement
distribution of OH. setup and is briefly described here for completeness.
A Q-switched, frequency-doubled Nd:YAG laser (Spectra-Physics A decay-time approach was applied using powder of the ther-
PIV 400, 10 Hz, 5–8 ns pulse width, only one cavity used) was used mographic phosphor Gd3 Ga5 O12 :Cr3+ ,Ce3+ (GGG:Cr,Ce), as in a
for optically pumping a narrow-band dye laser (Sirah Lasertechnik previous study [16]. The phosphor was airbrush-coated on the
GmbH, Double Dye, 2400 lines/mm, only one cavity used) operated quenching wall in the far post-flame region. For preparation, 1 g
with Rhodamine 590 dye. The fundamental wavelength was set to of GGG:Cr,Ce was mixed with 5 ml high-temperature binder (ZYP
567.86 nm (in air) and was frequency doubled to 283.93 nm. This Coatings Inc., HPC binder). For excitation of the thermographic
corresponds to an excitation of the Q1 (9) transitions of the OH rad- phosphor, the OH-PLIF setup described in the paragraph above
ical [44]. A half-wave plate was placed in front of the frequency was used. The signal was captured with an objective lens (Nikon,
conversion unit to control the final pulse energy in the UV, i.e., the Nikkor 50 mm, f/1.2) equipped with an interference filter (Edmund
conversion efficiency, while preserving the beam profile. Pulse en- Optics Inc., long-pass 650 nm, #84-759). The luminescence de-
ergies within the ROI were ≈ 0.3–0.4 mJ (10% CV). The UV beam cay was finally recorded by a combination of photomultiplier tube
was guided to the burner facility by high reflective dielectric mir- (PMT, Hamamatsu Photonics, type H11901-20) and digital oscillo-
rors. A combination of spherical and cylindrical lenses was used to scope (Tektronix Inc., TDS5034B).
form a laser light sheet. First, a Galilean telescope ( f = −100 mm As shown by Kosaka et al. [16], the temperature of the quench-
and f = +200 mm) expanded and collimated the laser beam. Sec- ing wall can be assumed constant in the post-flame region down-
ond, a cylindrical lens ( f = +500 mm) compressed the expanded stream of the quenching point. Thus, TPT measurements in the far
beam to the laser light sheet featuring 220 μm width in the ROI. post-flame region appear sufficient for control purpose. The de-
To cover both the quenching region and the thermographic phos- cay time was estimated iteratively using a non-linear least square
phor coating with this laser sheet, an additional cylindrical lens scheme, as described in Fuhrmann et al. [46]. The quantification in
( f = −300 mm) was used to further expand the height of the laser terms of temperature was done by an in-situ calibration with an
light sheet (approximately 60 mm height in ROI). Note that the ex- aluminum block equipped with heating cartridges and thermocou-
citing laser light sheet was introduced to the ROI inclined (angle of ples.
≈ 40◦ to horizontal) from above the CO-LIF detection system due
to the limited available space near the burner. 2.2.5. Synchronization and matching of the diagnostics
The OH-PLIF signal was collected using an objective lens (B. For simultaneous multi-parameter studies, both temporal and
Halle Nachfl. GmbH, 150 mm, f/2.5) equipped with an interfer- spatial matching of the systems is essential. All lasers and detec-
ence filter (Laser Components, BP300-325, 305–340 nm transmis- tion systems were synchronized using pulse generators (Quantum
sion). The emission was recorded using an IRO (LaVision GmbH, Composers Inc., 9520 Series). The individual diagnostics were sep-
low speed IRO) gated at 100 ns combined with a CCD camera (LaV- arated by 200 ns each to avoid crosstalk (temporal order: CO-LIF,
ision GmbH, Imager E-Lite 1.4 M) with 10 ms exposure time and a OH-PLIF, DP-CARS). All recording systems were operated at 20 Hz,
3 × 3 hardware binning (bin size ≈ 69 μm). Note that the entire while laser systems were running at 10 Hz only. Thus, every sec-
OH detection system was mounted inclined (angle of ≈ 15◦ to hor- ond image didn’t capture signal, but was used for the purpose of
izontal) above the DP-CARS excitation beams. background correction. For spatial matching of the diagnostics, a
The processing is outlined under OH-LIF in Fig. 3 and follows 100 μm reference pinhole was used to overlap both the DP-CARS
the procedure for flame front and quenching point detection de- probe volume and the UV beam for CO-LIF in the center of the
scribed in the work of Jainski [45]. Here, the procedure is refined ROIs (at x = 0 mm). The pinhole was further used as reference for
for the circumstances of the current study. OH-LIF raw images laser position in the two-dimensional OH-PLIF recordings. For spa-
were processed by an average background subtraction (recorded tial scaling of the CO-LIF and OH-PLIF data, a customized calibra-
between consecutive laser pulses, flame on), followed by two spa- tion target was used.
tial filtering steps: (1) a filtering in the Fourier-domain to selec- Further important aspects include (1) the procedure to deter-
tively remove a residual striped pattern on the laser light sheet mine the distance of the matched probe volumes from the wall
(most likely due to beam steering); (2) a 5 × 5 moving average fil- and (2) the in-situ monitoring of the probe volume matching of
ter to reduce spatial noise. Both filtering procedures enabled a ro- DP-CARS and CO-LIF during combustion operation. A detailed de-
bust calculation of the two-dimensional gradient magnitude, and scription of these aspects is given in Appendix A in the supplemen-
the maximum gradient magnitude value in y-direction for each z- tary material, as they are considered extremely helpful for quality
location was used to identify the flame front position. The quench- assurance in the laser diagnostic community.
ing point was then approximated from the flame front trace (gra-
dient magnitude normalized to [0; 1] on the trace) at the (y, z ) lo- 2.3. Experimental validation case
cation where the normalized gradient magnitude first exceeded a
value of 0.3, approaching from the post-flame direction. This value The experimental methodology for thermochemistry quantifica-
was chosen empirically to provide robust detection of the quench- tion was validated with an adiabatic reference measurement. Here,
ing location in the presented study (see results in Fig. 3 under OH- the free flame branch of the rod-stabilized V-flame was used. As
LIF). However, for reproduction in future studies, i.e., with a differ- the zone where the free flame branch was probed is 16 to 20.5 mm

8
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 4. Evaluation of the diagnostical approach on the adiabatic flame branch as reference case. State spaces shown in (a)–(c), corresponding single-shot residua in (d) and
(e), relative accuracy and precision on 50 K temperature increments in (f) and (g). Error bars only displayed in (a) and (b). 4284 individual samples are considered. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

away from the quenching wall in y-direction and heat loss via light green filled circles). This high relative accuracy affirms a good
radiation is negligible, the assumption of adiabaticity is valid, implementation of the CO calibration polynomial to transform cor-
and the flame may be represented by the one-dimensional, adia- rected LIF signals to mole fractions. Above 1700 K, the relative ac-
batic freely-propagating flame simulation described in Section 2.4. curacy deteriorates to approximately 10%. This is due to the de-
The results are outlined in Fig. 4(a)–(c) in terms of the state creasing absolute CO concentrations in the progressing oxidation
space, while (d) and (e) show the respective single-shot residua process. The average residuum (light green line in Fig. 4(e)) is ap-
r between simulation and experiment, giving an absolute mea- proximately constant over the entire temperature range and is less
sure for accuracy (μ(r ), light green line) and precision (scat- than 0.003 in CO mole fraction above 1700 K. Thus, the absolute
ter in residuum). Figure 4(f) and (g) assesses a relative accuracy accuracy in terms of the average residuum appears constant, but
(μ(r )/Xi,sim ) and precision (σ (r )/Xi,sim ) within individual tempera- the relative accuracy deteriorates as the absolute CO concentration
ture increments of 50 K; μ expresses the average, σ the standard becomes small. The average relative precision of the CO mole frac-
deviation and Xi,sim the corresponding simulations mole fraction of tion is near 10% (Fig. 4(g) blue crosses) within the temperature
species i. According to Fig. 4(a)–(c), after post-processing and cal- range 750 K < T < 1500 K, where most of the relevant FWI pro-
ibration, the experimental data well reproduces the adiabatic sim- cesses occur (wall distance: y = 120–320 μm). However, CO pre-
ulation of CO2 , CO, and T in the state space. Please note that the cision degrades at higher temperature due to lower CO signal lev-
scatter is highest in the (CO,CO2 ) state space in Fig. 4(c), due to els and the higher state-space gradient across the 50 K averaging
the superimposed uncertainties in both CO2 and CO (compared to bins. The assessment of the gas phase temperature via DP-CARS re-
the lower uncertainty in the temperature). The small gap appar- sults in a relative accuracy of 3% and precision of 4% (not shown
ent in the experimental (CO,T ) data in Fig. 4(b) near 20 0 0 K is due in Fig. 4), estimated from measurements in the fully burned gas
to a skipped measurement location while scanning over the adi- of the free branch. It is considered a conservative assessment to
abatic flame branch and does not affect overall data quality. For only use the highest temperature samples for the evaluation of the
the deduction of CO2 mole fraction from DP-CARS, it is empha- temperature quantification, as the nitrogen CARS signal is weak-
sized that the values originating directly from the spectral fitting est there. Overall, the applied laser diagnostics seem to perform
algorithm already feature a good relative accuracy of ≈ 9% on av- well in capturing the (CO2 ,CO,T )-thermochemistry of the adiabatic
erage (Fig. 4(f) dark green hollow circles). As a function of tem- reference case with both a good accuracy and precision. Thus, it
perature, the relative accuracy of CO2 mole fraction is 5–10% up is assumed that this experimental procedure remains valid for as-
to 1600 K, increasing to 20% at the highest temperatures. To fur- sessing the thermochemical states in the non-adiabatic near-wall
ther improve the accuracy, the data were additionally calibrated to region during flame quenching.
the adiabatic simulation (see Section 2.2). This results in an aver-
age relative accuracy of about 2% that is nearly constant over the 2.4. Numerical simulation benchmark cases
full temperature range (Fig. 4(f) light green filled circles). The av-
erage relative precision of the CO2 mole fraction is roughly 16%, In this study, two numerical flame configurations are used to
decreasing somewhat with increasing temperature (Fig. 4(f) blue complement the experiments: a one-dimensional, adiabatic freely-
crosses). The relative accuracy of the CO mole fraction, after cali- propagating flame [47], in the following referred to as the adiabatic
bration, is ∼2% and approximately constant up to 1700 K (Fig. 4(g) simulation, and a FCT two-dimensional SWQ simulation referred

9
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

to as the non-adiabatic simulation. The general setup of the SWQ


simulation (solver, grid, numerics) is described in Steinhausen et al.
[9]. In the present work, a recently developed mechanism [48],
consisting of 31 species and 140 reactions, is used in combination
with a mixture-averaged approach for diffusion [49,50]. To assess
the importance of diffusive transport during FWI, an equal diffusiv-
ity model (unity Lewis number for all species) is shown for com-
parison in the non-adiabatic case, additionally.

3. Results and discussion

This section presents and discusses results of the near-wall


thermochemistry. As mentioned previously, the focus is on a thor-
ough evaluation of one operation condition with DME fuel at  =
0.83, a cold wall (∼ 60 ◦ C) and laminar inflow.

3.1. Evaluation of the nominal wall distance Fig. 5. Evaluation of the experimental data at y = 100 μm nominal wall distance,
to estimate an offset to the simulation data. 1983 individual samples are shown
in the scatter. The non-adiabatic simulation is provided with a mixture-averaged
The nominal wall distance (y-direction) of the measured ex-
transport approach. Both experiment and simulation contain values from stream-
perimental data is examined, as the near-wall thermochemistry is wise z-positions spanning ±2 mm relative to the quenching height. (For interpre-
highly sensitive to this parameter. The comparison of experimental tation of the references to color in this figure legend, the reader is referred to the
and numerical data in past studies on FWI suggested that there is web version of this article.)
an offset in y-direction between the nominal wall distance, where
the measurements are believed to have been taken, and the real
wall distance, where the measured result best agrees with phys- ployed in the following. It is pointed out that evaluating such an
ical reality. This offset is assumed to originate from multiple in- offset correction is only considered beneficial in the state space
fluences: (1) beam steering, (2) a spatial filtering of the laser di- rather than in the physical space, as the state space does not suffer
agnostics due to the finite beam width within the steep gradients deviations in assessing the quenching location. This issue is further
close to the wall, and (3) inaccuracy in defining the wall-closest discussed in the following section.
nominal position “y = 100 μm”. However, in the present study, the
major error is believed to be due to beam steering, while (2) and 3.2. Thermochemistry progress with relative streamwise distance to
(3) are considered negligible. As the wall closest y-position is set quenching height
during thermally steady FWI operation (see Appendix A in the sup-
plementary material for procedure), there should be only marginal This section discusses the evolution of the thermochemical
inaccuracy in defining the nominal “y = 100 μm” position. As it is states in z-direction for data measured at the wall-closest y-
hardly possible to quantify the instantaneous position error caused position of 120 μm. This is initially done for the physical space in
by beam steering from the experimental data, the effect can be Fig. 6, followed by the corresponding description in the state space
estimated by using the non-adiabatic simulation as benchmark. in Fig. 7. The single-shot experimental data within the physical
Please note that the non-adiabatic simulation is also based on space in Fig. 6 is described using both a smoothing spline fit (black
modeling assumptions, and the use as benchmark may also suf- line) and averages conditioned with respect to the instantaneous
fer inaccuracy. However, this approach appears to be a promising z-location of the quenching point (color-coded circles). This pro-
way to give an estimation of the potential positioning error made cedure is named “quenching-point conditioning” within this study
by beam steering in the experiment. The approach is outlined in and calculates corresponding statistics within 100 μm increments
Fig. 5 using the (CO2 , T ) state space with the experimental data in z-direction after conditioning to the instantaneous z-location of
at y = 100 μm nominal wall distance, but for different z-locations the quenching point that is set to z = 0 mm. Comparing the fitted
(heights). The variation of (CO2 , T ) relative to the model arises to to the quenching-point conditioned averaged experimental data, it
the diagnostics precision as well as beam steering. The simulation is directly evident that binning or filtering the data with 100 μm
data evaluated at y = 120 μm wall distance (Fig. 5, solid orange spatial increments in z-direction is not biasing the absolute val-
line) appears to fit best to the conditioned average values of the ues, as the color-coded circles exactly match the fitted black line.
experiment (Fig. 5, magenta triangles). The conditioned averages In Fig. 6(a), the average experimental CO2 evolution (black line and
are calculated on equidistant increments, each spanning 0.0025 on color-coded circles) across the quenching location closely matches
the CO2 mole fraction axis. That conditioning method within the the corresponding simulation trace (red line). In the CO mole frac-
state-space is named “state-space conditioning” in this manuscript. tions shown in Fig. 6(b) a minor deviation between the fitted ex-
This results in an apparent y-offset of 20 μm between experimental perimental data and simulation is observed in both the peak lo-
and numerical data, likely to originate from beam steering. In the cation (∼ 140 μm) and the peak value (∼0.007). Similarly, for the
following, this 20 μm y-offset is added to the nominal, experimen- experimental temperature evolution in Fig. 6(c), there are differ-
tal wall distance values. Furthermore, it appears that the majority ences in peak z-location (285 μm) and value (35 K). These minor
of experimental scatter is within a ±30 μm range in y-direction differences may occur due to (1) the remaining unsteadiness of the
compared to the numerical data (see red and yellow dashed lines experimental flame, (2) the assumed boundary conditions in the
in Fig. 5). This range is seen as a superimposition of beam steer- simulation (e.g., the inflow profiles). Consequently, visualizing ther-
ing fluctuation and temperature precision due to spectral noise and mochemical data in the physical space is usually considered much
beam profile fluctuation within the DP-CARS probe volume. How- more sensitive than in the state space. Keeping this in mind, the
ever, it is not possible to separate these influences here. In conclu- agreement between experimental results and numerical simulation
sion, it can be stated that experiment and simulation match very is considered excellent. The experiment thus provides an outstand-
well after applying a small y-offset correction of 20 μm, as em- ing validation of the simulation. It is emphasized that, apart from

10
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 6. Evaluation of the experimental data at y = 120 μm wall distance in z-direction in the physical space. 1983 individual samples are shown in the scatter. The ex-
perimental data is described by a smoothing spline fit (black line) and quenching-point conditioned averages calculated for each 100 μm increment on the physical z-axis
(color-coded circles). The non-adiabatic simulation (red line) is provided with a mixture-averaged transport approach. (For interpretation of the references to color in this
figure legend, the reader is referred to the web version of this article.)

Fig. 7. Assessment of the thermochemistry progression in z-direction at y = 120 μm wall distance in the state space with quenching-point conditioning applied. 92 to 152
individual samples are shown in each scatter plot. Adiabatic and non-adiabatic simulations are shown, with both simulations using a mixture-averaged transport model. The
non-adiabatic simulation features values z = ±2 mm relative to the quenching location. The right column (e), (j) and (o) includes additional experimental data at a z = +2 mm
bin location (green scatter points), corresponding to the fully reacted state. (For interpretation of the references to color in this figure legend, the reader is referred to the
web version of this article.)

11
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

the y-offset correction in Section 3.1, the simulation has not been 3.3. Thermochemistry progress with wall distance
adjusted in any way to match the experimental data.
Figure 7 shows the evolution of the thermochemical states in The evolution of the thermochemical state in the individual
the individual (CO2 , T ), (CO,T ) and (CO,CO2 ) scatter plots across the (CO2 , T ), (CO,T ) and (CO,CO2 ) correlations for different wall-normal
quenching height in z-direction. Each column presents the samples y-distances is shown in Fig. 8. Each column presents a differ-
within a 200-μm bin in the wall-parallel z-direction centered at ent wall distance given in the column title. For comparison to
the location given at top of the column (e.g., the z = −400 μm bin the simulation, adiabatic and non-adiabatic simulations with a
in the first column ranges from z = −500 to −300 μm upstream of mixture-averaged transport model are presented. Additionally, a
the instantaneous quenching height). Negative values indicate loca- non-adiabatic case with unity Lewis number transport is depicted.
tions upstream and positive values downstream of the quenching For more detail related to the simulations, see Section 2.4. As
height, which is located at z = 0 μm. As described in Section 2.1, supplement, Fig. 9 summarizes the trend of both the previously
the flame quenching location fluctuates axially in z-direction. This introduced state-space and quenching-point conditioned averages
unsteadiness is accounted for by evaluating each sample individ- shown in Fig. 8 with increasing wall distance to enable a direct
ually against the instantaneous quenching height determined by comparison (For clarity, only distances y = 120, 320 and 720 μm
OH-LIF. Due to the unsteadiness, samples scatter across a range in are given). For state-space conditioning, the (CO2 , T ) and (CO,CO2 )
the thermochemical states because the spatial range for each sub- scatter view was conditioned by CO2 mole fraction increments
plot extends 200 μm in streamwise z-direction. The numerical ref- of 0.0025. The (CO,T ) was conditioned by CO mole fraction in-
erences are sampled over z = ±2 mm range relative to the quench- crements of 0.00125, and the CO-production and CO-oxidation
ing location. The evolution of thermochemical states is described branches were split and conditioned separately. The state-space
in the following. conditioned results are shown with magenta-colored triangles in
At (z, y ) = (−400 μm, 120 μm) shown in the first column of Fig. 8 (with only every third marker shown) and in Fig. 9(a)–(c).
Fig. 7, (CO2 , T ) and (CO,CO2 ) correlations in (a) and (k) are sim- This procedure performed well in the past when analyzing (CO,T )
ilarly reproduced by adiabatic and non-adiabatic simulations be- only, but it does not capture the (CO2 , T ) trend approaching com-
cause the differences are small and below the experimental uncer- pleted reactions. Therefore, this study additionally uses quenching-
tainty. However, the (CO,T ) correlation in Fig. 7(f) is more sensitive, based conditioning. As previously demonstrated for the physical
and non-adiabatic effects are more clearly visible. At (−200 μm, space data in Section 3.2, Fig. 6, the data for each wall distance
120 μm) (Fig. 7, second column), the non-adiabaticity becomes y is subdivided into 100 μm wide streamwise z-increments (40
apparent for the (CO2 , T ) scatter in (b), while the (CO,T ) scat- bins ranging from z = −2 to +2 mm) relative to the quenching
ter in (g) is clearly following the non-adiabatic simulation. At the height location at z = 0 μm. Quenching-point conditioned averages
quenching height (z, y ) = (0 μm, 120 μm) (Fig. 7, center column), of (CO2 ,CO,T ) are calculated for each 100- μm bin in z-direction.
both the (CO2 , T ) and (CO,T ) scatter plots in (c) and (h) coincide The results are shown with color-coded circles in Fig. 8 (every
with the non-adiabatic simulation. The CO mole fraction is peak- marker shown) and in Fig. 9(d)–(f). This procedure is considered
ing here near the transition from the CO-production branch to to capture the reaction progress of both CO and CO2 well as it
the CO-oxidation branch. Downstream of the quenching height at is directly related to the wall-parallel evolution, i.e., in z-direction,
the z = +200 and +400 μm locations (Fig. 7, fourth and last col- of the species during reaction. On a first glance it seems that the
umn), both the (CO2 , T ) and (CO,T ) scatter keep on following the quenching-point conditioning procedure underestimates the exper-
non-adiabatic simulation, while the (CO,T ) in Fig. 7(i) and (j) is imental CO peaks compared to the simulation in Fig. 8(f)–(j), but
fully on the CO-oxidation branch. At z = +20 0 0 μm distance down- as demonstrated in Section 3.2 this does not result from condition-
stream of the quenching height (Fig. 7, included in the last column ing the experimental data but originates from the overall lower ex-
with green scatter points), the (CO2 , T ), (CO,T ) and (CO,CO2 ) scatter perimental CO peak values in the experiment. Potential reasons for
plots indicate completed reactions. this deviation are given in Section 3.2. Regarding conditioning, it
From this evolution, a major finding is that upstream of the is rather noted that state-space conditioning (magenta-colored tri-
quenching location, CO mole fractions are more strongly influenced angles) leads to increased experimental peak values in the (CO,T ),
by diffusive transport towards the wall than CO2 . This can be ex- as observed in Fig. 8(f)–(i). Despite these values better match to
plained by the flame structure. In the reaction zone, CO is formed the simulation, their conditioned average is evidently biased, as
first and then followed by its oxidation to CO2 . Compared to CO, up it is exceeding the average experimental CO trace observed in the
to approximately z = 200 μm upstream the quenching point, CO2 physical space in Fig. 6(b). The reason for the biased averages from
thus senses the wall less and adiabatic and non-adiabatic (CO2 , T ) state-space conditioning is the separation of oxidation and produc-
correlations coincide up to approximately 750 K. This suggests that tion branch. By separating the data, the mean values are forced
(1) the CO2 is less affected by the quenching process than CO (as to outlier instead of following the true average. It is concluded
expected) and (2) for premixed near-wall combustion processes that conditioning in the physical space, i.e., the quenching-point
with DME fuel, CO2 is a suitable component of a progress vari- conditioning, is more robust for filtering errors and the preferred
able composed of different chemical species. The latter applicabil- method.
ity as progress variable component is facilitated by its monotonic Major findings from Figs. 8 and 9 are discussed in the following.
increase and small gradient over the reaction zone. Furthermore, Regarding feature (A), it is evident from the (CO2 , T ) thermochem-
it is clearly observed from the (CO,CO2 ) state space in Fig. 7(k)– istry in Fig. 8(a)–(e) that the non-adiabatic simulation with unity
(o) that the adiabatic and non-adiabatic behavior (as shown by Lewis number assumption is not capable of capturing the correct
the simulation data) are very close to each other across the en- near-wall flame structure prior to quenching. This behavior is not
tire state space. This expected behavior, which is experimentally apparent in the (CO,T ) scatter view in Fig. 8(f)–(j). Thus, it is con-
confirmed here within the measurement uncertainty, means the cluded that (1) differential diffusion effects need to be accounted
(CO,CO2 ) scatter plot is nearly independent from the wall heat for, e.g., by a mixture-averaged transport model for laminar DME
loss. It is concluded that in the vicinity of the quenching height flames; (2) (CO2 , T ) appears to be much more sensitive to differ-
(z = 0 μm) and y = 120 μm from the wall, the (CO,CO2 ) correlation ential diffusion effects than (CO,T ). Consequently, the CO2 aids sig-
is only influenced marginally by the quenching wall and the influ- nificantly in experimentally identifying the limitation of the unity
ence is below the experimental uncertainty. Lewis number assumption in FWI.

12
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 8. Assessment of the thermochemistry evolution for different wall distances. 1565 to 2047 individual samples are shown in each scatter plot. Both experiment and
simulation contain data z = ±2 mm (streamwise direction) relative to the quenching location. Labels (A) to (C) highlight features discussed in the text. (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of this article.)

From feature (B) in Fig. 8(b)–(e) it is apparent that, for increas- in Fig. 8(i) and (j). There, for wall normal distances larger than
ing wall distances y, the conditioned traces in the (CO2 , T ) state y = 520 μm, the Le = 1 approximation is obviously not valid for
space develop a trend, moving away from the mixture-averaged the conditioned experimental data in the CO-oxidation branch and
simulation and toward the Le = 1 simulation. This behavior is ex- the data coincide still better with the mixture-averaged modeling
clusively observed downstream of the quenching height (z = 0 μm) (see feature (B) in Fig. 8(i)–(j)). However, for higher temperatures
when the progress is approaching the fully reacted state. Al- (>1600 K) it is observed that the differential diffusion effects in
ready at y = 220 μm wall distance in Fig. 8(b), there is a minor (CO,T ) are not fully covered by the applied mixture-averaged mod-
bend towards the Le = 1 simulation and back to the simulation eling. This is not consistent with the observation made for (CO2 , T ).
with mixture-averaged transport, when the reaction approaches Error bars of the CO2 mole fraction are included for several spa-
the equilibrium. This bend is more pronounced for y = 320 μm tial bins used for quenching-point conditioning in Fig. 8(e) and (o)
wall distance in Fig. 8(c), but the reaction turns back to the sim- to outline that there is a pronounced experimental uncertainty in
ulation considering differential diffusion effects. For y = 520 and the CO2 mole fractions when approaching chemical equilibrium.
720 μm wall distance in (d) and (e), the conditioned thermochem- While the quenching-point conditioned average follows the Le = 1
ical behavior stays on the Le = 1 case after bending. This effect is assumption, the lower values of the error bars can reach to the
also apparent in the (CO,CO2 ) state space in Fig. 8(l)–(o). It is em- simulation with mixture-averaged transport. Thus, an interpreta-
phasized that for y = 120 to 320 μm wall distance in Fig. 8(a)–(c) tion of the bending from the simulation considering differential
and (k)–(m), the post flame values approach the equilibrium state diffusion effects towards the Le = 1 simulation for increasing wall
of the non-adiabatic simulation with mixture-averaged transport distances y needs further investigation. One possible explanation
modeling. This behavior is considered physically consistent. Even- could be the influence of local transport phenomena (also three-
tually this might also be the case for y = 520 and 720 μm distance dimensional, as there is a slight curvature of the wall and flame).
in (d), (e) and (n), (o), if additional downstream post-flame data However, this does not confirm that the Le = 1 assumption is cor-
were available to fully capture the trend toward the fully reacted rect in these regions. It is rather supposed that a compensation of
state. The observations are different in the (CO,T ) thermochemistry different physical effects may lead to the apparent Le = 1 behavior.

13
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

Fig. 9. Summary of the conditioned averages for the wall distance variation shown in Fig. 8. (a)–(c): state-space conditioning on defined CO as well as CO2 increments;
(d)-(f): quenching-point conditioning. (symbols equal to top row). (For interpretation of the references to color in this figure legend, the reader is referred to the web version
of this article.)

It is strongly suggested that a transport effect occurs, but it is not abatic and non-adiabatic simulation are very close to each other
conclusively clarified how best to model transport in the present and the experimental data suffers the superimposed uncertainties
configuration. in both CO and CO2 . However, it is further emphasized here that
It is furthermore observed in Fig. 8(c)–(e) that the upper lim- this correlation features least impact on the experimental ther-
its in the CO2 mole fractions in the experimental scatter data mochemistry data with increasing wall distances y. This is out-
exceed the simulation results, especially for increased wall dis- lined in Fig. 9(c) and (f) for both the state-space conditioning on
tances (y > 220 μm). According to simulations performed in Can- defined CO as well as CO2 increments in (c) and the quenching-
tera 2.4.0 [51] (well-stirred reactor at both constant pressure and point conditioning in (f). In contrast to the (CO2 , T ) in Fig. 9(a), (d)
temperature, Stagni-mechanism [48]), the equilibrium mole frac- and (CO,T ) correlations in (b), (e), the conditioned averages in the
tions of CO2 should vary between 0.101 and 0.104 (between 300 (CO,CO2 ) state space do not possess clear trends with increasing
and 2100 K). Experimental data exceeding the possible equilibrium wall distance but do overlap within their experimental uncertainty
value first appears unphysical but could be due to local diffusion (see transparent colored areas in Fig. 9(c), where the CO error with
effects that were not covered in the simulation (e.g., 3D effects). respect to CO2 is shown for y = 120 and 720 μm wall distance).
These phenomena need further investigation using advanced diag-
nostics and 3D numerical simulations to gain a deeper insight into 4. Conclusion and outlook
potential transport phenomena.
In Fig. 9(a), (b) and (d), (e) it is evident that the measured This study has presented the first application of simultaneous
(CO2 , T ) and (CO,T ) thermochemistry develop clear trends with in- dual-pump CARS, CO-LIF and OH-LIF in a side-wall quenching en-
creasing wall distance y. It is furthermore emphasized in Fig. 9 that vironment to simultaneously measure the three scalars CO2 , CO,
for both the (CO2 , T ) and (CO,T ) thermochemistry the trends of and T as close as 120 μm to the surface of the quenching wall
the conditioned averages are located within the boundaries of the and to determine the instantaneous quenching height. A compre-
physically possible state space between the non-adiabatic simula- hensive and detailed description of the applied methodology was
tion at y = 120 μm wall distance (dotted blue line) and the adi- provided, including a thorough discussion of accuracy and preci-
abatic simulation (solid red line). This is true for all wall dis- sion of the diagnostics. The capabilities of the extended laser diag-
tances y and independent of the conditioning method. This fact nostics for assessing the thermochemistry in side-wall quenching
creates further confidence that the conditioned averages are accu- for one operation condition (DME fuel,  = 0.83, wall 60 ◦ C, lami-
rate. This also implies that the trend experimentally observed for nar) were demonstrated and validated against an adiabatic laminar
the (CO,CO2 ) correlation is indeed real. Therefore, deviations from flame reference. The nominal wall distances from the experiment
the simulation data (see feature (C) in Fig. 8(n) and (o)) underline matched very well to the numerical simulation data with a small
the possible explanation above that there are further relevant ef- offset of 20 μm. Conditioning the thermochemical data with re-
fects that require a 3D examination. spect to the instantaneous quenching point enabled a novel trac-
As already mentioned in the discussion of Fig. 7, it is difficult to ing of the wall parallel chemistry evolution across the quenching
derive detailed statements from the (CO,CO2 ) correlation, as adi- location. The study showed the first comparison of the experimen-

14
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

tal three-scalar data against 2D numerical simulations. The major TRR 150. Robert S. Barlow is grateful for a related DFG Mercator
findings of this study are: Fellowship. The generous financial support is gratefully acknowl-
edged. Andreas Dreizler is grateful for support through the Gott-
• The novel three-scalar measurement approach considering
fried Wilhelm Leibniz program. Calculations for this research were
(CO2 ,CO,T ) allows highly sensitive probing of the thermochemi-
conducted on the Lichtenberg high performance computer of the
cal states and is clearly superior to the previously applied two-
Technical University of Darmstadt.
scalar measurement approach measuring (CO,T ) only. Simulta-
neously recording relative OH distribution allows correction for Supplementary material
remaining unsteadiness of the quenching height.
• Conditioning the experimental near-wall thermochemistry data Supplementary material associated with this article can be
within 100 μm spaced increments derived from the physi- found, in the online version, at doi:10.1016/j.combustflame.2021.
cal space, so called “quenching-point conditioning”, appeared 111707
superior to previously conditioning in the state space only,
e.g., to avoid ambiguities on the production and oxidation References
branches, respectively. Thus, quenching-point conditioning is
[1] J. Scheelhaase, S. Maertens, W. Grimme, Synthetic fuels in aviation—Current
recommended for future experimental flame-wall interaction barriers and potential political measures, Trans. Res. Proc. 43 (2019) 21–30.
(FWI) studies. [2] B. Lumpp, D. Rothe, C. Pastötter, R. Lämmermann, E. Jacob, Oxymethylene
• The CO2 mole fraction is less affected by the quenching wall ethers as diesel fuel additives of the future, MTZ Worldw. 72 (2011) 34–38.
[3] M. Härtl, K. Gaukel, D. Pélerin, G. Wachtmeister, Oxymethylene ether as po-
compared to CO and should serve as a more robust compo- tentially CO2 -neutral fuel for clean diesel engines part 1: engine testing, MTZ
nent of a progress variable definition in near-wall environments Worldw. 78 (2017) 52–59.
than CO. Especially, the (CO,CO2 ) correlation is only influenced [4] F. Ferraro, C. Russo, R. Schmitz, C. Hasse, M. Sirignano, Experimental and nu-
merical study on the effect of oxymethylene ether-3 (OME3) on soot particle
marginally by the quenching wall and the influence is below formation, Fuel 286 (2021) 119353.
the experimental uncertainty. [5] T.M. Kiehne, R.D. Matthews, D.E. Wilson, The significance of intermediate hy-
• Differential diffusion effects and their importance for CFD drocarbons during wall quench of propane flames, Symp. (Int.) Combust. 21
(1988) 481–489.
are experimentally proven by comparison to 2D fully-resolved [6] C. Hasse, M. Bollig, N. Peters, H.A. Dwyer, Quenching of laminar iso-octane
chemistry and transport (FCT) simulations using a mixture- flames at cold walls, Combust. Flame 122 (20 0 0) 117–129.
averaged transport and unity Lewis number modeling approach. [7] S. Ganter, C. Straßacker, G. Kuenne, T. Meier, A. Heinrich, U. Maas, J. Janicka,
Laminar near-wall combustion: analysis of tabulated chemistry simulations by
The (CO2 , T ) correlation is much more sensitive to differential
means of detailed kinetics, Int. J. Heat Fluid Flow 70 (2018) 259–270.
diffusion effects than the (CO,T ), even though (CO,T ) increas- [8] D.V. Efimov, P. de Goey, J.A. van Oijen, QFM: quenching flamelet-generated
ingly showed differential diffusion impact at higher tempera- manifold for modelling of flame–wall interactions, Combust. Theor. Model. 24
(2020) 72–104.
tures and larger wall distances that needs to be considered.
[9] M. Steinhausen, Y. Luo, S. Popp, C. Straßacker, T. Zirwes, H. Kosaka, F. Zent-
Thus, the CO2 aids significantly in experimentally identifying graf, U. Maas, A. Sadiki, A. Dreizler, C. Hasse, Numerical investigation of local
the limitation of a unity Lewis number assumption. The over- heat-release rates and thermo-chemical states in side-wall quenching of lam-
all agreement between the 2D simulations assuming mixture- inar methane and dimethyl ether flames, Flow Turbul. Combust. 106 (2021)
681–700.
averaged transport and the experimental results is very good. [10] A. Gruber, R. Sankaran, E.R. Hawkes, J.H. Chen, Turbulent flame–wall interac-
However, when approaching chemical equilibrium some devia- tion: a direct numerical simulation study, J. Fluid Mech. 658 (2010) 5–32.
tions in (CO2 , T ) and (CO2 ,CO) remained and their origins need [11] U. Ahmed, N. Chakraborty, M. Klein, Scalar gradient and strain rate statistics in
oblique premixed flame–wall interaction within turbulent channel flows, Flow
further investigation. Turbul. Combust. 106 (2021) 701–732.
• The experimental methodology proved feasible for laminar op- [12] C. Straßacker, V. Bykov, U. Maas, REDIM reduced modeling of quenching at
eration and yielded valuable, novel insights. This is an im- a cold wall including heterogeneous wall reactions, Int. J. Heat Fluid Flow 69
(2018) 185–193.
portant prerequisite and appears promising for the subsequent [13] C. Straßacker, V. Bykov, U. Maas, Reduced modeling of flame-wall-interactions
transfer to turbulent cases where the numerical analysis using of premixed isooctane-air systems including detailed transport and surface re-
direct numerical simulations (DNS) including FCT is limited. actions, Proc. Combust. Inst. 38 (2021) 1063–1070.
[14] A. Dreizler, B. Böhm, Advanced laser diagnostics for an improved understand-
As an outlook, the analysis described in this study will be ing of premixed flame-wall interactions, Proc. Combust. Inst. 35 (2015) 37–64.
[15] B. Boust, J. Sotton, S.A. Labuda, M. Bellenoue, A thermal formulation for sin-
expanded to the experimental data measured for different fuels, gle-wall quenching of transient laminar flames, Combust. Flame 149 (2007)
equivalence ratios and wall temperatures to gain a deeper under- 286–294.
standing of the parameter dependency in FWI processes. Especially [16] H. Kosaka, F. Zentgraf, A. Scholtissek, L. Bischoff, T. Häber, R. Suntz, B. Albert,
C. Hasse, A. Dreizler, Wall heat fluxes and CO formation/oxidation during lami-
the analysis of the turbulent cases appears of major interest, as nu-
nar and turbulent side-wall quenching of methane and DME flames, Int. J. Heat
merical simulations can get very complex and expensive when ap- Fluid Flow 70 (2018) 181–192.
proaching turbulent conditions featuring higher Reynolds numbers. [17] F. Foucher, S. Burnel, C. Mounam-Rousselle, M. Boukhalfa, B. Renou, M. Trinité,
Furthermore, for a clarification of the supposed transport phenom- Flame wall interaction: effect of stretch, Exp. Therm. Fluid Sci. 27 (2003)
431–437.
ena, further investigations by advanced diagnostics, like sponta- [18] M. Rißmann, C. Jainski, M. Mann, A. Dreizler, Flame-flow interaction in pre-
neous Raman scattering, to measure more species should be per- mixed turbulent flames during transient head-on quenching, Flow Turbul.
formed as well as 3D numerical simulations for a deeper insight Combust. 98 (2017) 1025–1038.
[19] C. Jainski, M. Rißmann, S. Jakirlic, B. Böhm, A. Dreizler, Quenching of premixed
into potential transport phenomena. flames at cold walls: effects on the local flow field, Flow Turbul. Combust. 100
(2018) 177–196.
Declaration of Competing Interests [20] C.W. Clendening, W. Shackleford, R. Hilyard, Raman scattering measurements
in a side-wall quench layer, Symp. (Int.) Combust. 18 (1981) 1583–1590.
[21] M. Saffman, Parametric studies of a side wall quench layer, Combust. Flame 55
The authors declare that they have no known competing finan- (1984) 141–159.
cial interests or personal relationships that could have appeared to [22] A.V. Mokhov, A.P. Nefedov, B.V. Rogov, V.A. Sinel’shchikov, A.D. Usachev,
A.V. Zobnin, H.B. Levinsky, CO behavior in laminar boundary layer of combus-
influence the work reported in this paper.
tion product flow, Combust. Flame 119 (1999) 161–173.
[23] T. Fuyuto, H. Kronemayer, B. Lewerich, J. Brübach, T. Fujikawa, K. Akihama,
Acknowledgments T. Dreier, C. Schulz, Temperature and species measurement in a quenching
boundary layer on a flat-flame burner, Exp. Fluids 49 (2010) 783–795.
[24] M. Mann, C. Jainski, M. Euler, B. Böhm, A. Dreizler, Transient flame–wall inter-
This project is funded by the Deutsche Forschungsgemeinschaft actions: experimental analysis using spectroscopic temperature and CO con-
(DFG, German Research Foundation) Projektnummer 237267381 centration measurements, Combust. Flame 161 (2014) 2371–2386.

15
F. Zentgraf, P. Johe, M. Steinhausen et al. Combustion and Flame 235 (2022) 111707

[25] C. Jainski, M. Rißmann, B. Böhm, J. Janicka, A. Dreizler, Sidewall quenching [38] T.B. Settersten, A. Dreizler, R.L. Farrow, Temperature- and species-dependent
of atmospheric laminar premixed flames studied by laser-based diagnostics, quenching of CO B probed by two-photon laser-induced fluorescence using a
Combust. Flame 183 (2017) 271–282. picosecond laser, J. Chem. Phys. 117 (2002) 3173–3179.
[26] A. Bohlin, C. Jainski, B.D. Patterson, A. Dreizler, C.J. Kliewer, Multiparameter [39] M.D. Di Rosa, R.L. Farrow, Cross sections of photoionization and ac Stark shift
spatio-thermochemical probing of flame–wall interactions advanced with co- measured from Doppler-free B ← X(0,0) excitation spectra of CO, J. Opt. Soc.
herent Raman imaging, Proc. Combust. Inst. 36 (2017) 4557–4564. Am. B 16 (1999) 861–870.
[27] M. Luo, D. Liu, Effects of dimethyl ether addition on soot formation, evolution [40] L. Voigt, J. Heinze, M. Korkmaz, K.P. Geigle, C. Willert, Planar measurements of
and characteristics in flame-wall interactions, Energy 164 (2018) 642–654. CO concentrations in flames at atmospheric and elevated pressure by laser-in-
[28] H. Kosaka, F. Zentgraf, A. Scholtissek, C. Hasse, A. Dreizler, Effect of flame-wall duced fluorescence, Appl. Phys. B 125 (2019) 505.
interaction on local heat release of methane and DME combustion in a [41] A.P. Nefedov, V.A. Sinel’shchikov, A.D. Usachev, A.V. Zobnin, Photochemical ef-
side-wall quenching geometry, Flow Turbul. Combust. 104 (2020) 1029–1046. fect in two-photon laser-induced fluorescence detection of carbon monoxide
[29] R.P. Lucht, Three-laser coherent anti-Stokes Raman scattering measurements of in hydrocarbon flames, Appl. Opt. 37 (1998) 7729–7736.
two species, Opt. Lett. 12 (1987) 78–80. [42] B. Atakan, J. Heinze, U.E. Meier, OH laser-induced fluorescence at high pres-
[30] S. Roy, T.R. Meyer, R.P. Lucht, V.M. Belovich, E. Corporan, J.R. Gord, Temper- sures: spectroscopic and two-dimensional measurements exciting the A-X (1,0)
ature and CO2 concentration measurements in the exhaust stream of a liq- transition, Appl. Phys. B 64 (1997) 585–591.
uid-fueled combustor using dual-pump coherent anti-Stokes Raman scattering [43] T. Dreier, A. Dreizler, J. Wolfrum, The application of a Raman-shifted tunable
(CARS) spectroscopy, Combust. Flame 138 (2004) 273–284. KrF excimer laser for laser-induced fluorescence combustion diagnostics, Appl.
[31] R.P. Lucht, V. Velur-Natarajan, C.D. Carter, K.D. Grinstead, J.R. Gord, Phys. B 55 (1992) 381–387.
P.M. Danehy, G.J. Fiechtner, R.L. Farrow, Dual-pump coherent anti-Stokes Ra- [44] G.H. Dieke, H.M. Crosswhite, The ultraviolet bands of OH fundamental data, J.
man scattering temperature and CO concentration measurements, AIAA J. 41 Quant. Spectrosc. Radiat. Transf. 2 (1962) 97–199.
(2003) 679–686. [45] C. Jainski, Experimentelle Untersuchung der Turbulenten Flamme-Wand-Inter-
[32] P. Ewart, A modeless, variable bandwidth, tunable laser, Opt. Commun. 55 aktion, Optimus Verlag, Göttingen, Germany, 2016.
(1985) 124–126. [46] N. Fuhrmann, J. Brübach, A. Dreizler, On the mono-exponential fitting of phos-
[33] A.D. Cutler, E.C.A. Gallo, L.M.L. Cantu, WIDECARS spectra fitting in a premixed phorescence decays, Appl. Phys. B 116 (2014) 359–369.
ethylene-air flame, J. Raman Spectrosc. 47 (2016) 416–424. [47] R.J. Kee, J.F. Grcar, M.D. Smooke, J.A. Miller, A Fortran program for modeling
[34] S. Linow, A. Dreizler, J. Janicka, E.P. Hassel, Comparison of two-photon excita- steady laminar one-dimensional premixed flames, Report No. SAND85-8240,
tion schemes for CO detection in flames, Appl. Phys. B 71 (20 0 0) 689–696. Sandia National Laboratories, Albuquerque, NM, USA, 1985.
[35] M. Greifenstein, A. Dreizler, Influence of effusion cooling air on the thermo- [48] A. Stagni, Y. Luo, M. Steinhausen, A. Dreizler, C. Hasse, Chemistry effects in the
chemical state of combustion in a pressurized model single sector gas turbine wall quenching of laminar premixed DME flames, Combust. Flame 232 (2021)
combustor, Combust. Flame 226 (2021) 455–466. 111529.
[36] K.A. Rahman, K.S. Patel, M.N. Slipchenko, T.R. Meyer, Z. Zhang, Y. Wu, J.R. Gord, [49] C.F. Curtiss, J.O. Hirschfelder, Transport properties of multicomponent gas mix-
S. Roy, Femtosecond, two-photon, laser-induced fluorescence (TP-LIF) measure- tures, J. Chem. Phys. 17 (1949) 550–555.
ment of CO in high-pressure flames, Appl. Opt. 57 (2018) 5666–5671. [50] T. Coffee, J. Heimerl, Transport algorithms for premixed, laminar steady-state
[37] O. Carrivain, Etude de la Spectroscopie LIF à Deux Photons de la Molécule CO flames, Combust. Flame 43 (1981) 273–289.
Pour des Mesures en Flammes à Haute Pression, Université Pierre et Marie [51] D.G. Goodwin, R.L. Speth, H.K. Moffat, B.W. Weber, Cantera: an object-oriented
Curie - Paris VI, 2015 Ph.D. Theses. software toolkit for chemical kinetics, thermodynamics, and transport pro-
cesses, Version 2.4.0, 2021, 10.5281/ZENODO.4527812

16

You might also like