Combustion and Flame 159 (2012) 2533–2562

Contents lists available at SciVerse ScienceDirect

Combustion and Flame

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e

Raman/Rayleigh scattering and CO-LIF measurements in laminar and turbulent

jet flames of dimethyl ether
Frederik Fuest a,⇑, Robert S. Barlow b, Jyh-Yuan Chen c, Andreas Dreizler a
a
FG Reactive Flows and Diagnostics, Center of Smart Interfaces, TU Darmstadt, Petersenstr. 32, 64287 Darmstadt, Germany
b
Sandia National Laboratories, 7011 East Avenue, Livermore, CA, USA
c
Department of Mechanical Engineering, University of California, Berkeley, CA 94720, USA

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

Article history: To reduce the impact of combustion of fossil fuels on air quality and climate change, dimethyl ether
Available online 28 November 2011 (DME) is a promising alternative diesel fuel candidate. Technical combustion processes, including forma-
tion of pollutants, are influenced by turbulence–chemistry interaction. Therefore, accurate prediction by
Keywords: computational combustion models of combustion systems burning DME must account for multiple
Dimethyl ether scalars and scalar gradients. The testing of such models requires detailed experiments. Here a study is
Raman/Rayleigh scattering presented on the feasibility of simultaneous species and temperature measurements in turbulent
Raman spectroscopy
dimethyl ether flames, using line-imaged Raman/Rayleigh scattering of the major species H2, O2, N2,
Combustion diagnostics
Hydrocarbon intermediates
CO, CO2, H2O, C2H6O and laser induced fluorescence of CO. The measurement system and data evaluation
Rayleigh cross-sections methods developed to investigate methane–air flames are extended to address dimethyl ether flames.
The Raman signal intensity and spectral shape of the Raman scattering from dimethyl ether over a range
of temperatures are presented, based on measurements in electrically heated flows and laminar jet
flames. These data are used to develop an iterative method for data evaluation that allows determination
of indispensable crosstalk correction terms for the concentration measurements of O2 and CO2. Issues of
fluorescence interferences, mainly from C2 radicals on the fuel-rich side of the reaction zone, and their
corrections are discussed. Laminar flame calculations are used to investigate the role of the intermediate
species (CH4, CH2O, C2H4, C2H2, C2H6, CH3) in the reaction zone. In particular, their effect on the mixture
fraction calculation and its relationship to the experimentally determined mixture fraction is examined.
The impact of the intermediate species on deviations in concentration and temperature profiles due to
species-specific Raman- and Rayleigh scattering cross-sections is demonstrated. Finally, species concen-
trations and temperature profiles from measurements in a turbulent piloted jet flame of dimethyl ether
are shown.
Ó 2011 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction measurements of the thermochemical state were performed. This

requires the simultaneous measurement of multiple scalars.
Turbulent combustion processes are of high practical relevance. Point-wise and more recently line-imaged instantaneous Raman/
A variety of complex physical–chemical interactions control pollu- Rayleigh scattering was developed and forms an important basis
tant formation, ignition, and flame stability, calling for a more de- for understanding turbulence-chemistry interactions in turbulent
tailed understanding. Significant progress has been made in flames. This paper reports on extending these investigations to
collaborative research within the TNF Workshops [1], addressing flames with more complex fuels than methane.
these phenomena by providing well-documented bench mark The emission of carbon dioxide (CO2) as a greenhouse gas
flames that also significantly pushed advances in numerical simu- enforces an increased use of renewable fuels. Among a variety of
lations. These collaborative efforts began by looking into diluted choices dimethyl ether (C2H6O) exhibits a number of interesting
hydrogen-fueled jet flames [2,3] and were subsequently extended properties. Chemically dimethyl ether (DME) is the simplest ether.
to hydrogen/methane/nitrogen jet flames [4]. As the next step in In atmospheric-pressure flames neither preheating of the fuel is
complexity, piloted partially-premixed methane/air flames were necessary nor does condensation occur in the fuel feeding pipes
studied [4–9]. In addition to velocity measurements, instantaneous due to sufficient vapor pressure at room temperature. DME is an
excellent alternative for Diesel fuel, with low NOx emission levels,
⇑ Corresponding author. Fax: +49 6151166555. low particulate emissions, and a high cetane number for good
E-mail address: [email protected] (F. Fuest). auto-ignition performance. In terms of Raman/Rayleigh scattering

0010-2180/$ - see front matter Ó 2011 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustflame.2011.11.001
2534 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

DME stands out due to relatively low C2 and soot precursor forma- formed in these flames constitute significantly higher mole frac-
tion that causes significant fluorescence interferences on different tions than in corresponding methane flames and their Raman
Raman bands. and Rayleigh scattering properties are, in some cases, significantly
Previous studies using Raman scattering in gaseous hydrocar- different than those of the parent fuel.
bon/air flames with fuels chemically more complex than methane In partially premixed methane flames it has been demonstrated
are rare. In an early study by Stårner et al. [10] simultaneous by Barlow et al. [19] that differences between the mass fraction of
Raman/Rayleigh/LIF measurements were made in piloted turbu- CH4 and the total mass fraction of all hydrocarbons are relatively
lent jet diffusion flames of diluted propane. It was shown that Ra- small. The Raman scattering signal corresponding to C-H bond
man measurements were feasible only when propane was diluted stretch in the hydrocarbon intermediates overlaps the spectrum
substantially by air or nitrogen, preventing overwhelming soot of CH4. Consequently, with appropriate calibration of the tempera-
precursor interferences. Pilot-stabilized non-premixed methanol ture dependent response of the ’CH4 channel’, the processed results
flames were investigated by Masri et al. [11] for stable conditions yield a good approximation of the total hydrocarbon mass fraction
as well as close to blowoff. Propane flames, partly in mixture with and good agreement with laminar flame calculations on profiles of
dodecane or Diesel, were investigated by Dreyer et al. [12]. The major species, temperature and mixture fraction.
speciality of this work was a Raman measurement in the vicinity The above fortuitous condition does not hold for DME flames,
of liquid fuel droplets. One important conclusion was that for their where laminar calculations show that the total mole fraction of
experimental conditions 355 nm excitation proved to be a better hydrocarbon intermediates can exceed 5% in the fuel-rich region
candidate wavelength than previously thought. Meier and Keck of a partially premixed DME/air flame. Intermediate hydrocarbon
[13] and Rabenstein and Leipertz [14] investigated premixed soot- species, such as methane (CH4), ethane (C2H6), or ethylene
ing and non-sooting C2H4/air and CH4/air flames, respectively. In (C2H4), exhibit rovibrational Raman bands that spectrally overlap
Meier and Keck’s comparative study the signal-to-background the DME Raman bands and cannot easily be separated, especially
ratio of Raman measurements was investigated for pulsed laser when on-chip binning is used in the spectral direction as a noise
radiation at 532, 489, 355, and 266 nm. Excitation wavelength of reduction strategy demonstrated by Miles [20] and for the current
532 nm proved to be most suitable, and limitations of concentra- experimental setup by Fuest et al. [21]. These intermediate species
tion measurements by laser Raman measurements were demon- contribute to measured Raman signal intensities on the same
strated. In contrast, for soot-volume concentrations nearly two detection channel as used for DME such that the Raman response
orders of magnitude higher, Egermann et al. [15] reported excita- of this channel depends strongly on the local hydrocarbon compo-
tion wavelengths of 266 nm to be beneficial compared to 355 nm sition as well as temperature. Due to huge differences in particular
in ethylene diffusion flames because of reduced spectral overlap Rayleigh cross-sections of important intermediate hydrocarbons
between Raman bands and LIF interferences. Nooren et al. [16] the effective Rayleigh cross-section also depends on the local
reported on Raman/Rayleigh/LIF measurements in Dutch natural hydrocarbon composition. Furthermore, rovibrational Raman tran-
gas jet diffusion flames. High levels of fluorescence interferences sitions of DME and hydrocarbon intermediates spectrally overlap
were subtracted from the Raman signals by empirical correlations with other species, including rovibrational lines of CO2, molecular
using, amongst others, an interference channel monitored at oxygen (O2), carbon monoxide (CO), and molecular nitrogen (N2),
615 nm. This approach was exploited by Dibble et al. [17] and it and also the spectral region used to monitor fluorescence interfer-
is followed similarly in the present study. Brockhinke et al. [18] ence. These crosstalk contributions need to be characterized in-
studied LIF of C2 following a UV-excitation at 248 nm. They identi- depth when quantifying species mole fractions. In short, Raman/
fied fluorescence in different spectral ranges and discussed possi- Rayleigh measurements cannot be interpreted in a quantitatively
ble interferences with Raman bands especially around 350 nm. useful way without accounting for the concentrations and scatter-
Removal of LIF interferences by separating and subtracting the sig- ing properties of the main hydrocarbon intermediates.
nals in two polarization directions was proposed as a possible The present paper presents a method for post-processing of
approach to measuring mole fractions by Raman scattering in fuel line-imaged Raman/Rayleigh scattering measurements in DME/
rich laminar premixed propylene/oxygen flames. Egermann et al. air flames. The method relies on species information derived from
[15] similarly proposed the possibility of Raman measurements laminar flame calculations, as well as detailed information on the
with excitation at 266 nm in sooting ethylene diffusion flames scattering properties of relevant molecules. The paper is structured
using horizontal and vertical polarization directions. None of these as follows: Section 2 discusses the experimental setup, briefly
studies, however, addressed in sufficient detail the role of interme- introduces the extended matrix inversion method [21], specifies
diate hydrocarbons or strategies to account for the resulting effects the laminar and turbulent DME/air flames, and outlines laminar
on Raman/Rayleigh scattering. flame calculations. In Section 3 the role of intermediates in DME/
The extension of quantitative line-imaged Raman/Rayleigh air flames is addressed. It is shown that particularly hydrocarbon
scattering to turbulent flames of hydrocarbon species more com- intermediates cannot be neglected. The impact of the intermediate
plex than methane is an important research priority. The present species upon Raman responses, crosstalks, effective Rayleigh cross-
work was initiated as an exploratory study to investigate the tem- sections and mixture fraction is examined. For the first time mod-
perature-dependent Raman scattering properties of several simple els to account for intermediate species impact on Raman and Ray-
hydrocarbon fuels (ethane, ethylene, propane, and DME) and to as- leigh responses are proposed and detailed instructions are given to
sess the prospects of obtaining turbulent flame measurements of systematically obtain the Raman response characteristics of the
the quality appropriate for combustion model validation. The focus fuel and crosstalk channels which rely upon a priori information
was on DME because of the practical relevance noted above and from laminar flame calculations. At appropriate stages, sensitivity
also because, as an oxygenated fuel, DME has the lowest propen- studies are used to show the influence of simplifying assumptions.
sity to form soot or soot precursors, which generate strong fluores- Following the conceptual explanations, Section 4 presents Raman
cence interference in Raman experiments. response and crosstalk curves specifically used in this study.
Raman scattering spectra were measured in heated flows and in Broadband and C2 interferences are discussed in detail followed
laminar jet flames. Early analysis made it clear that interpretation by results and discussions of laminar DME/air flames. The applica-
of Raman/Rayleigh signals from flames of DME (or any of the other bility of instantaneous line-imaged Raman/Rayleigh/LIF measure-
tested fuels) was significantly more challenging than for methane ments in piloted turbulent premixed and partially-premixed
flames. The main reason is that the hydrocarbon intermediates DME/air flames is demonstrated. Finally the most important
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2535

findings are summarized. A comprehensive Appendix presents ground luminosity around Raman shifts of 930 cm1 and
details that are addressed particularly to readers using combined 4300 cm1, respectively.
Raman/Rayleigh- or just Rayleigh- measurements to study any Data post-processing was based on the hybrid matrix inversion
complex hydrocarbon flames. method described in [21]. In this method the binned signals from
the Raman channels, fluorescence channel, and background chan-
nel comprise a signal vector S. Number densities of individual
2. Experimental and numerical approach
chemical species, N, are calculated from the signal vector S using
the matrix equation,
2.1. Experimental setup and data post-processing
S ¼ PðTÞN: ð1Þ
Line-imaged multi-scalar measurements were conducted at the
Combustion Research Facility of the Sandia National Laboratories. The matrix elements pij(T) depend on temperature and specific
The experimental setup has been described previously by Karpetis experimental conditions. The diagonal elements represent the
and Barlow [22,23] such that here only a brief description is pro- Raman detection system response for each species. For the non-
vided. Raman and Rayleigh scattering was excited by a cluster of hydrocarbon species (CO2, O2, CO, N2, H2O, and H2) the temperature
four sequentially fired frequency-doubled Nd:YAG lasers (Contin- dependence of these diagonal elements is determined from integra-
uum) operating at 532 nm. Three successive optical delay lines tion of spectral libraries over intervals corresponding to the binned
stretched the pulse to lower the probability of optical breakdown. Raman channels. The spectral libraries are based on quantum chem-
The combined laser energy at the probe volume location was up to ical calculations [21].
1.8 J/pulse with a beam width of 300 lm (1/e2). The one-dimen- For hydrocarbon species this approach is not feasible because of
sional probe volume length spanned 6 mm. Using a custom lacking reliable quantum chemical spectra simulations, particu-
designed achromatic lens (Linos Photonics, f1 = 300 mm, f/2, and larly for calculating reliable temperature dependencies of inte-
f/4) and spectral separation by a long pass beam splitter, Rayleigh grated Raman signals. Moreover, in DME/air flames, a variety of
and Raman scattering was measured by two CCD cameras. Ray- hydrocarbons contribute to the same channel and a special treat-
leigh images were recorded using a 2  2 hardware binning and ment is required as detailed in Section 3. In the following, diagonal
a pixel resolution of 20 lm. In front of the Rayleigh camera a matrix elements are termed as Raman responses. Off-diagonal
10 nm (FWHM) 532 nm band pass filter was positioned. Before elements represent crosstalk between different species due to
entering the Rayleigh camera, CO fluorescence was split off using spectral superposition of Raman signals. Additional off-diagonal
a dichroic beam splitter reflecting 484 nm but transmitting elements represent correlations between signals on the fluores-
532 nm. cence channel (or background channel) and fluorescence interfer-
Raman scattering passed a high-transmission thin-film pola- ence (or background luminosity) landing on the various species
rizer reducing the crosstalk of depolarized broadband and C2-fluo- channels. Again, for Raman scattering crosstalk between non-
rescence interferences by up to 50%. All measurements were hydrocarbon species, the temperature dependence of matrix
repeated with the polarizer turned by 90° to monitor the back- elements is determined by integrating theoretically based spectral
ground and minor contributions from depolarized Raman scatter- libraries. Temperature dependent Raman responses and crosstalk
ing. Polarization properties of Raman bands and background curves for the stable hydrocarbons of interest were determined
were used to gain insights into the spectroscopic nature of the ob- from measurements in electrically heated gas mixtures and various
served signals. The custom transmission spectrometer dispersing laminar flames. In Section 4.1 Raman response and crosstalk curves
Raman scattering provided high optical throughput [23]. Gating are discussed and Appendix C provides supplementary crosstalk
at 3.9 ls (FWHM) was achieved by a fast rotating shutter at the curves used in this work.
entrance to the spectrometer. Raman data were acquired in two Simultaneously to the line-imaged Raman/Rayleigh measure-
modes, one using on-chip (hardware) binning and the other using ments, CO was detected by two-photon laser-induced fluorescence
full spectral resolution (roughly 0.12 nm/pixel) and subsequent (LIF). For low CO number densities LIF has advantages over Raman
software binning. In both cases, on-chip binning by 10 pixels was scattering, yielding higher precision and accuracy, especially in
used in the spatial direction for a projected resolution of 102 lm flames with strong fluorescence interference. The CO-LIF setup
across the 6 mm probe length, resulting in 60 strips of spectral was described by Karpetis and Barlow [6] and Barlow et al. [23].
information. In wavelength direction spectral ranges for both hard- Uncertainties in these multi-scalar measurements are similar to
ware and software binning were selected to define seven Raman previous measurements [23], but slightly increased due to use of
detection channels to monitor seven major species as described the polarization filter. Representative values for precision and
in [21]. In addition, two spectrally binned regions were similarly accuracy in temperature and non-hydrocarbon species measure-
defined for monitoring of fluorescence interference and back- ments are detailed in Table 1. In this paper, measurements in

Table 1
Representative uncertainties of single-shot measurements at flame conditions in laminar premixed methane/air flat flames for software- (sb) and hardware-binned (hb) data-
acquisition. The increase in accuracy values for the turbulent flames is due to higher flame luminosity, fluorescence interference, uncertainties in Rayleigh cross-section and
Raman responses as discussed in detail in Sections 3 and 4. Turbulent data was acquired using software-binning.

Scalar Precision r (%) Accuracy (%) Laminar flame Accuracy (%)

hb/sb laminar flames condition turbulent flames
T 0.9/3 2 / = 0.97, T = 2171 K 3–8 (max. of 8% @1400 K, fuel-rich)
N2 0.8/4 2 / = 0.97, T = 2171 K 3
CO2 3/9 4 / = 0.97, T = 2171 K 6
H2O 2.3/6.5 3 / = 0.97, T = 2171 K 6
O2 ðX O2 ¼ 0:01Þ 35/150 2 / = 0.97, T = 2171 K 50 (strong LIF interference in T2)
/ or F 2.3/11 5 / = 0.97, T = 2171 K 10
CO 7.5/30 10 / = 1.28, T = 2029 K 20
CO-LIF 6.5/9 10 / = 1.28, T = 2029 K 15
H2 7.5/40 10 / = 1.28, T = 2029 K 15
2536 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

Table 2
Unburnt gas compositions of laminar premixed and partially premixed DME/air flames in mole fractions, with Ar and CO2 based on the composition of standard air with 5%
relative humidity. In the experiment the co-flowing air contained 35% relative humidity (0.007 mole fractions). The gas temperature was 290 K.

DME N2 O2 Ar H2O CO2 u (m/s) Re /

L1 0.114 0.6911 0.1854 0.0083 0.0009 0.0003 2.7 1730 1.85
L2 0.281 0.5608 0.1504 0.0067 0.0007 0.0003 2.3 1880 5.6

Table 3
Unburnt gas compositions of turbulent DME/air flames in mole fractions.

DME N2 O2 Ar H2O CO2 u (m/s) Re /

T1 0.114 0.6911 0.1854 0.0083 0.0009 0.0003 41.0 23,500 1.85
T2 0.197 0.6264 0.1680 0.0075 0.0008 0.0003 41.0 26,500 3.5

turbulent flames are evaluated using spectrally resolved Raman-

data in combination with software-binning, resulting in higher
cumulated readout-noise from the CCD. Note that using
software-binning shifts the overall precision-limit of the system
at flame temperatures from being shot-noise limited to readout-
noise limited.

2.2. Flame configurations

Fig. 1. Sketch of Tsuji burner (left) and opposed flow burner (right).
In this study laminar and turbulent rich premixed and rich par-
tially premixed DME/air jet flames were investigated. DME is
advantageous compared to other fuels, such as C2H4, C2H6, or jet flames, with most of the heat release occurring in a diffusion
C3H8 because of less fluorescence interference levels. The stoichi- controlled reaction zone at the stoichiometric condition. In
ometric point for DME in air is 6.5 vol.%. For the laminar jet flames contrast, flame T1 is a turbulent Bunsen flame, with much of the
a nozzle diameter of 8 mm was used. Bulk flow velocities of heat release occurring in a rich premixed reaction zone.
premixed and partially premixed jets were 2.7 and 2.3 m/s, respec-
tively. Corresponding Reynolds numbers are below 2000. Stoichi- 2.3. Numerical procedure
ometric values of the mixture fraction are Fst = 0.59 and 0.26 for
L1 and L2, respectively. The jet was shielded from the surrounding The analysis of DME/air flames was supported using 1D compu-
environment by a low-velocity (0.3 m/s) laminar coflow of air. All tations. Laminar flames are simulated using both the Tsuji and
electronic flow controllers (MKS or Tescom) were calibrated opposed jet geometries (Fig. 1). In the Tsuji geometry, the flame
against laminar flow elements. Line-imaged multi-scalar measure- is stabilized in the forward stagnation region of a porous cylinder
ments were conducted 20 mm downstream of the nozzle exit in ra- immersed in a uniform oxidizer flow. The imposed strain on the
dial direction. Two different mixture compositions were used that flame is calculated as
are termed L1 and L2. The gas compositions of these two flames are a ¼ 2U=R; ð2Þ
provided in Table 2. Flame L1 burned with a central premixed cone
(tip height 80 mm) surrounded by a stratified post-oxidation where U is the approaching velocity of the oxidizer and R is the
region (Bunsen type flame), where a methane-counterpart was radius of the cylinder. With a stagnation flow formulation, the Tsuji
investigated by Chou et al. [24]. Flame L2, having a richer jet flames were computed using the OPPFLOW code developed by Mill-
mixture, did not exhibit an inner premixed reaction zone and, er et al. [25].
therefore, provides a flame structure more typical of non-premixed In the opposed jet configuration fuel and oxidizer issue from
flames, with a single reaction zone near the stoichiometric two opposed nozzles and imping against each other. The flame is
condition. stabilized between the two jets near the stagnation plane. The glo-
Multi-scalar measurements were applied to piloted, turbulent bal strain rate of the flame is calculated as proposed by Seshadri
DME/air jet flames to investigate the feasibility of applying the and Williams [26]
present approach to flames of relevance to turbulent combustion  rffiffiffiffiffiffi
Uo Uf qf
model validation. For this purpose the well-known burner config- a¼ 1þ : ð3Þ
H U o qo
uration of the Sandia-Sydney piloted flame series A–F was used
[5,4,6–9]. Two DME/air flames were investigated, as detailed in Ta- Where Uo is the velocity and qo the density of the oxidizer stream
ble 3. The gas composition of the lean premixed pilot was CH4/H2/ and Uf, qF of the fuel stream, respectively. H is the distance between
air: 0.055/0.055/0.89. This pilot composition was used for simplic- nozzles. The computer code OPPDIF developed by Lutz et al. [27]
ity in this first study, knowing that it will be appropriate in future was used. In case of rich partially-premixed jets, two flame zones
work to match the enthalpy and atom balance of pilot to that of the form consisting of a rich premixed flame and a diffusion flame. Both
main fuel at the same equivalence ratio. Stoichiometric values of flame zones partly overlap resulting in a flame structure more
the mixture fraction are Fst = 0.59 and 0.36 for T1 and T2, respec- complex than pure premixed or diffusion flames.
tively. The total bulk velocity of the main jet was fixed at Two reaction mechanisms from Zhao et al. [28] (55 species) and
41 ms1, but the DME/air ratio was varied. Reynolds numbers are Kaiser et al. [29] (78 species) were compared. Planar premixed
23.500 and 26.500 for T1 and T2, respectively, recalling character- flame speeds were computed using the PREMIX code by Kee
istics of piloted CH4/air flame D. As will be shown, T2 has a scalar et al. [30] as one method of evaluating differences between the
flame structure analogous to the piloted partially premixed CH4/air mechanisms. Laminar flame speeds resulting from the two reaction
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2537

60 0.30
DME 2100
T
50 0.25
1800
Flame Speed (cm/s)

40 0.20

mole fractions
H2O 1500
O2

T (K)
30
0.15 CO 1200
20 inter-
Zhao et al. 0.10 mediates CO2 900
Kaiser et al. N2-0.5 H2
10
0.05 600
0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0.00 300
Equivalence Ratio (-) 0.0125
CH4 × 0.25 2100
Fig. 2. Laminar flame speed of DME/air vs. equivalence ratio comparing both T
mechanisms. Lines between calculated equivalence ratios are interpolated and only 0.0100 CH2O × 0.5 1800
Ar
included to guide eyes.
C2H2

mole fractions
1500
0.0075 OH
mechanisms are compared in Fig. 2. The Kaiser mechanism over-

T (K)
predicts laminar flame speeds, whereas good agreement between 1200
results obtained from the Zhao mechanism and experimental data 0.0050 C2H4
is found by Zhao et al. [28]. Based on this comparison, the Zhao 900
C2H6 H
mechanism was used primarily in this work. However, the differ-
0.0025 O
ences in species prediction between the two mechanisms are 600
other CH3
briefly considered in Section 3.1. Tsuji flames were simulated over
a range of strain rates from a = 100 s1 to near-extinction for each 0.0000 300
0 2 4 6 8 10
of the four DME/air mixtures (L1, L2, T1, and T2), using both multi-
x (mm)
component transport and equal diffusivities (mass diffusivity
equals thermal diffusivity), with the latter results expected to be Fig. 3. Species and temperature profiles from laminar flamelet calculation. Tsuji
more representative of turbulent flames where differential diffu- geometry, Zhao et al. mechanism, a = 50 s1, L2, multi-component transport. In
sion effects become less obvious. Opposed jet calculations with (top) the sum of the species from (bottom) except Ar is shown as ‘intermediates’.
These seventeen species always represent more than 0.998 mol fractions of all 55
multi-component transport were done just for L1 and T1 at strain species from the mechanism.
rates of a = 50 and 100 s1. Results are used in the analysis that
follows.
3.1. Identification of species relevant to Raman/Rayleigh scattering in
3. Analysis based on laminar flame calculation DME flames

In hydrocarbon flames, Raman and Rayleigh scattering origi- Based on laminar flame calculations described in Section 2.3
nates from educts, products (major species) and intermediates. In major and intermediate species are now identified that signifi-
CH4/air flames these intermediates so far have not been treated cantly contribute to Raman and Rayleigh scattering in DME/air
systematically because of comparatively low concentrations and flames. This analysis, however, is focused only on fuel/air mixture
relatively minor influence on effective scattering cross-sections. compositions introduced in Tables 2 and 3. General observations
As stated in the introduction, Raman/Rayleigh measurements from are reported using just case L2 followed by a discussion of the
flames of DME and other more complex fuels cannot be processed impact of mixture composition variations, reaction mechanisms
using the same simple methods from CH4 flames; initial attempts from Zhao et al. and Kaiser et al., strain rate variation, and flow
in the present study revealed significant inconsistencies. geometry (opposed jet and Tsuji). To select relevant species
In this section, results of the laminar DME/air flame calculations contributing significantly to Raman and Rayleigh scattering a min-
are used together with data on integrated Raman signal intensities imum mole fraction of 0.001 was used as threshold. Thereby the
and Rayleigh scattering cross-sections to analyze the influence of seven common major species (CO2, O2, CO, N2, DME, H2O, H2),
hydrocarbon intermediates on Raman and Rayleigh scattering and additionally ten minor species (CH4, CH2O, Ar, OH, C2H2, H,
measurements. Details of scattering properties needed for this C2H4, O, C2H6, CH3) were identified for further analysis.
analysis are provided in Appendices A and B. Following the identi- For case L2 (28.1% DME in air) the spatial profiles for mole frac-
fication of all relevant intermediates, the special circumstances tions and temperature are presented in Fig. 3. In the top of Fig. 3
resulting from the matrix inversion (MI) method commonly used the common seven major species are shown in comparison to
for evaluation of Raman/Rayleigh scattering are addressed. In the the total mole fraction of nine intermediates (excluding Ar from
present experimental setup, the spectral region for Raman shifts the ten minor species above). Intermediate species contribute up
located between 2775 and 3263 cm1 (hydrocarbon channel) is to a mole fraction of 0.078 at maximum. This peak is located at
used to detect the Raman scattering signal from DME and some the fuel-rich side of the flame, with a corresponding temperature
of the intermediate hydrocarbon species. Details of the superposed of 1239 K at x = 2.57 mm. The bottom of Fig. 3 shows the ten spe-
spectral signature of the hydrocarbon mix are lost due to pixel cies individually and all remaining low-concentration species are
binning. Consequently, a priori information on composition and summed up in ’other’. The following observations can be made
temperature must be used. In the approach presented, different from these profiles: First, CH4 and CH2O are the most abundant
models for the Raman response within the hydrocarbon channel, intermediates with maximum mole fractions of 0.048 and 0.02,
the effective Rayleigh cross-section, and the calculation of the mix- respectively. At x = 2.66 mm (1287 K), the corresponding DME
ture fraction are introduced. mole fraction has approached already 0.03, corresponding to less
2538 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

than half of the intermediates. Second, four intermediate hydrocar- 1.0

Raman response normalized to 295 K

bons (CH4, C2H2, C2H4, C2H6, CH3) add up to a mole fraction of 0.07. T2,Tsuji,eq.diff.
a=2500s-1
Just these four intermediates contribute to the hydrocarbon Raman 0.9 T2,Tsuji,multi-comp.
-1
channel in the present measurements, but all of them contribute to a=100s
the effective Rayleigh cross-section as shown in Sections 3.2 and T2,Tsuji,eq.diff.
0.8
3.3. Third, all other species excluded from the seventeen species a=100s-1
mentioned above never exceed a total mole fraction of 0.002. In
spite of typical experimental sensitivities, especially for single laser 0.7
L2,Tsuji,multi-comp.
-1
shot Raman scattering, such low concentrations are well justified a=50s
to be neglected. 0.6 L1,opp.jet,multi-comp.
To evaluate the sensitivity of these findings, a parametric vari- a=50s-1

ation of laminar flame calculations was conducted. Using a mixture

0.5
composition of 19.7% DME in air, the reaction intermediates sum 400 800 1200 1600 2000
up to a mole fraction of 0.052, compared to 0.07 in the 28.1% T (K)
DME in air flame, both a = 100 s1, multi-component transport.
Fig. 4. Sensitivity of the Raman response on the hydrocarbon channel to strain,
Switching to equal diffusivity transport, the maximum intermedi-
transport model, and geometry using Eq. (4) and constant signal intensity ratios
ates mole fraction decreases to 0.05. Increasing the strain rate to from Table B.6 (right column, 820 K).
just below the extinction limit (multi-component: a = 1750 s1
and equal diffusivity: a = 2500 s1) the intermediates mole fraction
reduces to 0.035 and 0.028, respectively. The same trends are on the hydrocarbon channel IRam,HC is described by a linear combi-
observed when replacing the Zhao et al. mechanism by the Kaiser nation of the five relative hydrocarbon mole fractions
P
et al. mechanism. However, using the Kaiser et al. mechanism, the X i ðTÞ= 5j¼1 X j ðTÞ weighted by their corresponding relative intensi-
mole fractions summed up from the same intermediate species in ties IRam,i/IRam,DME
total is reduced by approximately 10%. It is apparent from these P5 IRam;i
calculations that the total mole fraction of hydrocarbon intermedi- i¼1 X i ðTÞ IRam;DME
IRam;HC ðTÞ ¼ P5 : ð4Þ
ates decreases with decreasing DME fraction in the fuel/air
j¼1 X j ðTÞ
mixture, decreases with increasing strain rate, and decreases when
equal diffusivities are applied rather than multi-component trans- This temperature dependence solely comes from the decompo-
port. However, the intermediates mole fraction remains important sition and oxidation reactions of hydrocarbons at rising tempera-
in all cases and must be considered because scattering cross-sec- tures. Density effects are excluded from Eq. (4). Raman responses
tions, molar masses, and atomic constitutions differ significantly based on Eq. (4) are shown in Fig. 4.
from the parent fuel. Neglecting hydrocarbon intermediates in Obviously, the consumption of hydrocarbon species by chemi-
the evaluation procedure would result in systematic errors in as- cal reactions depends on boundary conditions. Therefore the
sumed scattering properties, as is detailed in the following Raman response in Eq. (4) has been derived using Xi(T) from
sections. laminar flame calculations, various strain rates, transport models,
flow geometries, and mixture compositions. For the T2 mixture
3.2. Analysis of Raman scattering response composition (19.7% DME in air), the Raman response differs up
to 20% at 1500 K when the strain rate increases from a = 100 s1
Temperature dependent distributions of DME and relevant to a = 2500 s1. Switching the transport model from multi-
intermediate hydrocarbon species from laminar flame calculations component to equal diffusivities yields small differences in the
are now used to quantify their impact on the Raman response of range of 1–3%. Similar variations are observed for the mechanism
the hydrocarbon channel. A strong Raman crosstalk on CO2 and by Kaiser et al. [29] (not shown in Fig. 4). Results of the L1 mixture
O2 is addressed in Section 4 and the Appendix B. In the following composition (11.4% DME in air) are shown for the opposed flow
treatment sensitivities of Raman response to strain, transport, configuration. Although this flame burns with an inner premixed
mixture composition, and geometry are investigated. For this sen- cone the Raman response is bounded by the results for the diffu-
sitivity study, integrated Raman signal intensities from DME and sion-flame-like cases. It is concluded from this analysis that
intermediate hydrocarbons are assumed not to vary with temper- the proposed model approach simplifies the physical–chemical
ature. Referring to Table B.6 in the Appendix, this simplification is processes, especially for turbulent flames with varying influence
well justified up to 820 K as measured in the heated gas flows. of molecular transport effects, strain, flow patterns, or even local
However, this restriction is not applied for the final data evaluation extinction. However, Section 4 will show its practicability by pre-
of the flames. Response curves in Section 4 differ accordingly and senting results which are based on different Raman responses,
linear extrapolations of measured temperature dependences to each corresponding to an appropriate laminar calculation.
higher temperatures are included (compare Fig. 12). Note also that Signal intensity ratios describing the crosstalk onto detection
the effects of optical bowing in the spectrometer as indicated in channels for the fluorescence interference channel, CO2, and O2
Fig. B.24 and resulting strip-dependence of the Raman response at lower Raman shifts are provided in Tables B.7,B.8,B.9 in the
are included in the final data analysis. This is not discussed here Appendix. At these lower Raman shifts differences in scattering
for simplicity of presenting the conceptual approach. intensities are much more pronounced among the individual
The analysis that follows is for the center of the bowed image. hydrocarbon molecules because their rovibrational bands are sep-
The hydrocarbon channel ranging on this center strip from arated. The crosstalks onto the O2, CO2, and fluorescence channels
2798 cm1 to 3263 cm1 covers a major part of the spontaneous are discussed in detail in Section 4.
Raman scattering originating from DME, CH4, C2H4, C2H6, and
CH3. Concurrently, remaining hydrocarbons like CH2O and C2H2 3.3. Effective Rayleigh cross-section
are not contributing to this spectral range because of their different
Raman shifts (compare spectra and text presented in Appendix B). The laminar flame calculations from Section 3.2 are also used to
As the integrated Raman scattering signal is linear with number understand the influence of the intermediate species on the Ray-
density, the temperature dependence of the collective response leigh temperature measurements. The Rayleigh signal intensity is
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2539

proportional to the effective Rayleigh cross-section. The effective 2.6

Rayleigh cross-section is determined from a linear combination σRay,eff,ref
2.4
of species-specific Rayleigh cross-sections rRay,i weighted by the σRay,eff,exp
respective mole fractions Xi 2.2 ratio
X

2
σRay,eff / σRay,N
rRay;eff ¼ X i rRay;i : ð5Þ 2
-1
i
a=100s
1.8
...400s-1
In the data evaluation of simultaneous Raman/Rayleigh measure-
1.6 ...2500s-1
ments, species mole fractions are determined from the Raman multi-component
responses on the different channels. These mole fractions are used 1.4
to calculate the Rayleigh cross-section of the mixture. Then the ma- increasing strain
1.2
trix Eq. (1) is solved iteratively since Raman responses depend on
temperature. In contrast, Rayleigh cross-sections are independent 1
or only weakly dependent on temperature. In practice, Rayleigh 2.6
cross-sections measured at room temperature are used throughout σRay,eff,ref
2.4
the whole temperature range in flames. This assumption of constant σRay,eff,exp
cross-sections may cause a bias towards low temperatures by up to 2.2 ratio

2
2% as remarked by Sutton et al. [31]. However, according to the

σRay,eff / σRay,N
2
authors’ experience for the present flames, this systematic deviation a=100s
-1

is below 1%. Hence, temperature dependencies of Rayleigh cross- 1.8 -1

...400s
sections are neglected in the following analysis. Rayleigh cross-sec- -1
1.6 ...2500s
tions for various species have been derived from the literature and equal diffusivity
experiments and are listed in the Appendix in Table A.5. 1.4
In this study Rayleigh cross-sections based on refraction indices increasing strain
1.2
from Gardiner et al. [32] for 532 nm excitation wavelength along
with a static value from Bacskay et al. [33] for formaldehyde are 1
400 800 1200 1600 2000
used. All values are normalized to N2. The seventeen relevant spe-
T (K)
cies identified in Section 3.1 are now further assessed considered
for their contributions to Rayleigh scattering. Contributions to Fig. 5. Effective Rayleigh cross-section for T2 configuration (19.7% DME) from
the Rayleigh signal from all other lower-concentrated species are laminar flame calculation in Tsuji geometry, mechanism from Zhao et al., using
not significant and are neglected. Contributions from Ar, O, H, (top) multi-component and thermal diffusivity transport and (bottom) equal
diffusivity. Solid lines show the cross-sections for different strain rates including
and OH are neglected as well (see Appendix B.2). Therefore, thir- the thirteen most relevant species deduced from Eq. (6). Dashed lines are derived
teen species remain (CO2, O2, CO, N2, DME, H2O, H2,CH4, CH2O, following Eq. (7b), where rRay, HC = rRay,DME is assumed for the five hydrocarbons.
C2H2, C2H4, C2H6, CH3), and the effective Rayleigh cross-section Dotted lines show the ratio rRay, eff,exp/rRay,eff,ref.
for DME/air flames reads, P11
contribution to the Rayleigh scattering signal, i¼7 X i ðT; aÞ
P13
r
i¼1 X i ðT; aÞ Ray;i
r Ray;HC ¼ X HC ðT; aÞrRay;HC . Prior to introducing the model for this
rRay;eff; ref ¼ P13 : ð6Þ term, the temperature dependent deviation between Eqs. (6) and
i¼1 X i ðT; aÞ
(7b) is considered with the Rayleigh cross-section of DME being
Mole fractions are denoted here as functions of both temperature T used for all five in Eq. (7b), i.e. rRay, HC = rRay,DME. This is effectively
and strain rate a, expressing their dependence upon the specific what is assumed in Raman/Rayleigh processing measurements from
laminar flame calculation. This Rayleigh cross-section is denoted methane flames, where the Rayleigh cross-section for CH4 is applied
as ’reference’ because all mole fractions Xi of the contributing spe- to all molecules contributing to the fuel Raman channel, without
cies are assumed to be known, whereas, in the experiment, this causing significant error.
assumption is true only for the major species (CO2, O2, CO, N2, Figure 5 provides temperature dependent effective cross-sec-
H2O, H2). Since both C2H2 and CH2O do not contribute to the signal tions normalized to nitrogen. Results for rRay,eff,ref and rRay,eff,exp
on the hydrocarbon channel, data from this channel represent con- (assuming rRay,HC = rRay,DME) are shown for both transport models
tributions from five remaining hydrocarbons (DME, CH4, C2H4, C2H6, and various strain rates. The deviations of rRay,eff,exp are quantified
and CH3). Therefore, information accessible from the experiment by the ratio rRay,eff,exp/rRay, eff,ref. With rising temperature in DME
reduces to only eleven species. Accordingly, the ’experimental’ flames the true value of rRay,HC becomes significantly smaller than
effective Rayleigh cross-section is composed only by eleven species rRay,DME due to the build-up of the intermediate hydrocarbons that
and can be described as exhibit smaller Rayleigh cross-sections. Significant differences
P6 P11 between rRay,eff,exp and rRay,eff,ref are observed. These differences
i¼1 X i ðT; aÞ rRay;i þ i¼7 X i ðT; aÞ rRay; HC would translate directly to systematic error in temperature
rRay;eff; exp ¼ P11 ð7aÞ
i¼1 X i ðT; aÞ measured in fuel rich conditions. For the lowest strain rate
P6 a = 100 s1 the maximum of rRay,eff, exp/rRay,eff,ref around 1435 K is
r r
i¼1 X i ðT; aÞ Ray;i þ X HC ðT; aÞ Ray;HC
¼ P6 : ð7bÞ slightly larger for equal diffusivity (18%) than for the multi-compo-
i¼1 X i ðT; aÞ þ X HC ðT; aÞ nent transport model (16%). Independent of the transport model
In Eq. (7b), contributions from the six major non-hydrocarbon spe- the ratio decreases with increasing strain rates due to diminishing
cies and the five hydrocarbon species (excluding C2H2 and CH2O) of variation in rRay,eff,exp for rising strain.
P
are grouped into two terms, 6i¼1 X i ðT; aÞrRay;i and XHC(T, a)rRay,HC.
In the latter, instead of the species-specific Rayleigh cross-sections 3.4. Temperature and strain rate dependent Rayleigh cross-section
an effective, mole-weighted cross-section rRay,HC is used for the model
hydrocarbons. Because mole fractions of the five hydrocarbon
species remain unknown from the experiment, a model has to be In order to avoid the unacceptably large systematic errors
implemented providing a sound estimation of the hydrocarbon illustrated by Fig. 5, a model must be used to account for species
2540 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

10 0.20 5
-1
a=1750s
4 multi-component
9
a=2500s-1
a=100s-1 3

Σ (X i) of 5 hydrocarbons
equal diffusivity
8 -1 0.15
a=400s 2
2

Δσ Ray,eff (%)
σRay,HC / σRay,N

-1 -1
7 a=1750s 1 a=400s
multi-component
0 a=100s-1
6 0.10 multi-component
-1
5 -2
a=100s-1
4 0.05 -3 equal diffusivity
multi-component -4
3
-5
400 800 1200 1600 2000
2 0.00
10 0.20
T (K)

Fig. 7. Emerging differences between effective Rayleigh cross-sections using just

9
one particular model (a = 400 s1, equal diffusivity) for distinct strain rates and
a=100s-1

Σ (X i) of 5 hydrocarbons
8 -1 0.15 transport.
a=400s
2
σRay,HC / σRay,N

-1
7 a=2500s
rRay; eff;ref  rRay;eff; exp ðrRay;HC;a¼400 s1 Þ
6 0.10 DrRay;eff ¼  100%: ð9Þ
rRay;eff;ref
5
Results shown in Fig. 7 are based on the effective cross-section of
4 0.05 hydrocarbons, rRay, HC, calculated with an intermediate strain rate
equal diffusivity of a = 400 s1 and equal diffusivity. For the reference effective
3
cross-section, rRay,eff,ref, strain and transport model are varied. For
2 0.00 equal diffusivities deviations sum up to ±4.5% at 1400 K. For
400 800 1200 1600 2000
multi-component transport models DrRay,eff is reduced to approxi-
T (K)
mately 2.5% using the identical strain rate of a = 400 s1. In case of
Fig. 6. Effective Rayleigh cross-section of the combined hydrocarbons DME, CH4, turbulent flames with varying strain inaccuracies will depend on
C2H4, C2H6, and CH3, based on mole fractions from laminar calculations for the two instantaneous flow properties and can be as high as 4.5%. Note that
transport models and the T2 boundary conditions (same calculations as in Fig. 5). all these considerations affect the effective Rayleigh scattering
cross-section only. In the iterative Raman/Rayleigh data evaluation
procedure these model-based inaccuracies do have a continuative
specific Rayleigh cross-sections of the important hydrocarbon impact because of the mutual interaction of temperature and mole
intermediates that contribute signal to the Raman fuel fraction determination. Using one particular laminar flame calcula-
channel. The following expression is used to represent the tion for both the Raman response and the Rayleigh cross-section
weighted Rayleigh cross-section for hydrocarbons as it depends model, inaccuracies may partly compensate or amplify. In the
on temperature and strain rate, based on laminar flame end, final inaccuracies of the entire evaluation scheme can be
calculations. assessed best by benchmarking measurements against laminar
flame calculations. Another issue is how the time-average in phys-
P11 P6
rRay;eff; ref i¼1 X i ðT; aÞ  i¼1 X i ðT; aÞ r Ray;i ical space and the conditional average on mixture fraction in turbu-
rRay;HC ðT; aÞ ¼ P11 ð8Þ
lent flames are affected. This will be discussed in Section 4.3 on the
i¼7 X i ðT; aÞ
experimental results of T2 (compare Fig. 22).
This equation is inserted to Eq. (7b), which is used within the iter-
ative Raman/Rayleigh data evaluation. Based on the input of lami- 3.5. Mixture fraction space, mass fractions, and atom ratios
nar flame calculations, by nature the model varies with strain,
transport model etc. This dependence is shown in Fig. 6 for two The presence of intermediate hydrocarbon species in the DME
different transport models at various strain rates. In addition the flames impacts the calculation of mixture fraction. Thus in this
P section the determination of mixture fraction is examined, using
integrated mole fraction 11 i¼7 X i ðT; aÞ ¼ X HC ðT; aÞ is included to the
figure. Whereas rRay,HC shows a sensitivity with regard to strain laminar flame calculations as well as information about experi-
and transport model, XHC is much less dependent on strain. How- mentally accessible data. The impact of detection issues regarding
ever, slight differences especially at low temperatures are observed the hydrocarbons described in Sections 3.1,3.2,3.3,3.4 is discussed
between the two transport models. Significant values of XHC are in the context of molar masses and atom numbers. The mixture
present throughout the whole temperature range underlying the fraction is calculated following the method of Barlow and Frank
important role of intermediate hydrocarbons in Raman and [5], where oxygen is excluded from the expression of Bilger et al.
Rayleigh scattering. Considering the sensitivity on the strain rate, [34]:
deviations for a = 100 s1 and a = 2500 s1, respectively, are up to 2ðY C  Y C;2 Þ=wC þ ðY H  Y H;2 Þ=2wH
+60% relative to the result for a = 100 s1 at 1550 K. Differences F¼ : ð10Þ
2ðY C;1  Y C;2 Þ=wC þ ðY H;1  Y H;2 Þ=2wH
resulting from the two transport models at equal strain rate are
smaller. At 1300 K differences add up to 8% but increase for temper- Herein YC, YH are local elemental mass fractions, wC, wH are atomic
atures above 1700 K. Increased sensitivity of the model at these masses, and subscripts 1 and 2 refer to the inflow conditions of the
high temperatures is of minor influence because of the relatively main jet and co-flowing air, respectively. This expression has some
low value of XHC. The impact of the sensitivity of rRay,HC upon advantage in partially premixed flames, where oxygen mass frac-
rRay,eff, exp calculated from Eq. (7b) is now evaluated by using the tion boundary conditions are similar in the two streams, making
following expression: the full Bilger expression more sensitive to experimental noise.
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2541

Elemental mass fractions are derived from the species mass frac- 0.4
tions Y Cx Hy Oz HC
0.35
X
Y C ¼ wC xi ðY Cx Hy Oz Þi =ðwCx Hy Oz Þi ; ð11Þ 0.3 Yi
i

mass fractions
X 0.25 *
Y H ¼ wH yi ðY Cx Hy Yi
Oz Þi =ðwCx Hy Oz Þi ; ð12Þ
i 0.2
O2 CO CO2
where x, y and z refer to the corresponding number of C, H, O atoms 0.15
and wCx Hy Oz to the respective molar mass of the species i. Raman H2O
0.1
measurements provide species concentrations C Cx Hy Oz . Species mass N2-0.5
or mole fractions are derived via post-processing, using a molar 0.05
H2
mass weighted normalization or simple normalization, 0
respectively: 0 2 4 6 8 10
, x (mm)
X
Y Cx Hy Oz ¼ C Cx Hy Oz wCx Hy Oz ðC Cx Hy Oz wCx Hy Oz Þi ; ð13Þ Fig. 8. Effect on species mass fractions due to the approximation (16) of using the
i molar mass of DME for all hydrocarbons. This molar mass and the resulting mass
, fraction is generally too high: Y HC P Y HC . Thus, all other species mass fractions are
X too small due to the normalization: ðY i Þ nonHC 6 ðY i ÞnonHC . Based on a laminar
X Cx Hy Oz ¼ C Cx Hy Oz ðC Cx Hy Oz Þi : ð14Þ calculation for L2 conditions, Tsuji geometry with a = 50 s1, Zhao mechanism, and
i multi-component.
In practical Raman/Rayleigh measurements, hydrocarbon species
are cumulated into the hydrocarbon channel. Thus, individual
Minor species contributions to Y HnonHC are summarized in parenthe-
concentrations of fuel and intermediate hydrocarbons (HC) are
ses. Their total contribution is small as will be shown in Section 3.6
summarized in
X below, and they are neglected. The hydrocarbon part reads
C HC ¼ ðC Cx Hy Oz Þi ; x ^ y – 0: ð15Þ
i Y CHC ¼ 2Y HC wC =wDME ð23Þ
Accordingly information on C Cx Hy Oz w Cx Hy Oz for intermediate hydro-
carbons necessary for calculation of Eq. (13) is not available from and
the measurements. In the present approach the molar mass of
DME is used for representing all hydrocarbons cumulated into the Y HHC ¼ 6Y HC wH =wDME : ð24Þ
Raman channel:
ðwCx Hy Oz Þi ¼ wDME ; x ^ y – 0: ð16Þ Consistent with Eq. (16) numbers of C and H atoms are those for
This is a crude simplification. However, as will be shown, compari- DME. Accordingly, corresponding atom ratios are also affected and
sons between experiments and calculations can still be made on a denoted by C⁄/H⁄, C⁄/O⁄ and so forth. Species which contribute sig-
consistent basis, and an additional level of complexity is avoided nificantly in Eq. (17) to the cumulated mass fractions of hydrocar-
in the implementation of the overall method of Raman/Rayleigh bons, Y HC , within Raman measurements in DME/air flames are
data analysis for DME flames. The assumption (16) causes signifi- discussed below. Finally, the adapted mixture fraction F⁄ reads
cant deviations in calculated mass fractions and mixture fraction    
from normally defined values, so adapted mass and mixture 2 Y C  Y C;2 =wC þ Y H  Y H;2 =2wH
F ¼ ; ð25Þ
fraction definitions are introduced and denoted by Y⁄ and F⁄, respec- 2ðY C;1  Y C;2 Þ=wC þ ðY H;1  Y H;2 Þ=2wH
tively. The cumulated mass fractions of hydrocarbons and
non-hydrocarbons read Clearly, the use of the adapted mass fractions Y⁄ and the corre-
,( )
X sponding F⁄ must be justified. First, this is the approach that is used
Y HC ¼ C HC wDME C HC wDME þ ðC i wi ÞnonHC ð17Þ in analyzing Raman/Rayleigh data from CH4 flames, where hydro-
i carbon intermediates have a much smaller influence on results.
and Second, species mass fractions and mixture fraction are quantities
,( ) derived only in the post-processing; they are not used within the
  X iterative evaluation of Raman/Rayleigh data. Third, because of lack
Y i nonHC
¼ ðC i wi ÞnonHC C HC wDME þ ðC i wi ÞnonHC : ð18Þ
i
of knowledge of individual intermediate hydrocarbon species con-
centrations, simplifying assumptions must be made. In order to
Note, that normalization in Eqs. (17) and (18) affects mass fractions keep the approach as simple as possible, the assumption (16) for
of all species. This is shown in Fig. 8. To derive corresponding ele- the molar mass and the assumptions in Eqs. (23) and (24) for corre-
mental mass fractions, Eqs. (11) and (12) are split into two terms: sponding atom numbers are preferred to other conceivable ap-
Y C ¼ Y CnonHC þ Y CHC ; ð19Þ proaches, such as taking the concentration-weighted molar
masses from specific laminar flame calculations. Fourth, the differ-
Y H ¼ Y HnonHC þ Y HHC : ð20Þ ences D(F⁄  F) seen below in Fig. 9 can be minimized by selecting
an optimal number of intermediate hydrocarbon species within the
For species emerging from the reaction mechanism by Zhao et al.
entire evaluation procedure. This selection is presented in the next
the non-hydrocarbon part is provided by
section. Finally, by applying the same method of post processing to
Y CnonHC ¼ Y CCO2 þ Y C CO
ð21Þ both computational and experimental results, consistent quantita-
tive comparisons can be carried out, as will be demonstrated in
and Section 4.3. Note that especially for comparison of species mass
 
fractions this procedure is essential, as easily seen from the differ-
Y HnonHC ¼ Y HH þ Y HH þ Y HOH þ Y HH þ Y HHO þ Y H : ð22Þ
2 2O 2 H2 O2 ences shown in Fig. 8.
2542 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

0.1 (e) 7 species, omitting Y CH4 ; Y C2 H4 ; Y C2 H6 ; Y CH3 in 2d.

P
±σF (f) 7 species, same as 2e, but normalized to 7i¼1 Y i .
exp
0.05 Additionally, two cases 3a and 3b are defined to mimic the actual
experimental condition where fuel and intermediate hydrocarbons
Δ (Fx-Fall ) w/o O

0
are cumulated in a single detection channel. Both cases require
-0.05 simplifying assumptions regarding the molar masses and atom
13 numbers of intermediate hydrocarbons corresponding to the
-0.1
11 assumptions leading to the adapted mixture fraction F⁄ given by
7
-0.15 Eq. (25). They are defined on the basis of mole fractions, which
7norm
6+(7) here is equivalent to the Raman-measured concentrations in Eqs.
-0.2
6+(5) (17) and (18).
-0.25 3. (a) 6+(7) species, X CO2 ; X O2 , XCO, X N2 ; X H2 O ; X H2 ,
0 0.2 0.4 0.6 0.8 1
(XDME, X CH4 ; X CH2 O ; X C2 H2 , X C2 H4 ; X C2 H6 ; X CH3 ).
Fall w/o O (b) 6+(5) species, as above but omitting X CH2 O ; X C2 H2 .
Fig. 9. Deviation of mixture fraction for cases 2b, 2d, 2e, 2f, 3a, and 3b plotted vs.
mixture fraction of the reference case. Crosses represent experimental precision of Case 3b, containing 6 non-HC species and 5 HC species, corre-
the mixture fraction in actual L2 experiments using on-chip binning. sponds to the selection of species already introduced in Sections
3.2 and 3.3 above. Elemental mass fractions, Eqs. (19) and (20),
were calculated from the species mass fractions, Eqs. (17) and
3.6. Impact of intermediates on mixture fraction determination (18), using mole fractions Xi and employing the approximations
(16), (23), and (24), which assign the molar mass and atom num-
Laminar flame results are now used to investigate the impact of bers of DME to all considered hydrocarbon intermediates. This
different sets of hydrocarbon intermediates on the computed mix- simplification leads to differences between the corresponding
ture fraction. Mixture fraction deviations are compared to a refer- cases 2b M 3a and 2d M 3b, despite of identical species considered.
ence case, which is calculated from Eq. (10). Results are found Maximum differences of all cases relative to the reference case
similar for all laminar flame calculations used in this work and are summarized in Table 4 for selected measures as specified in the
are presented for the Tsuji flame geometry using 28.1% DME in table caption. From cases 2 it turns out that the separate detection
air (L2 configuration), multi-component transport, and a strain rate of twelve species of case 2c would suffice to allow mixture fraction
of a = 50 s1. The reference case is listed as case 1 in Table 4. The and temperature measurements in the order of current measure-
case ’Bilger’ in the second row refers to the original mixture frac- ment precisions. Further negligence of species leads to unaccept-
tion definition including oxygen [34] and all 55 chemical species able distortions, i.e. for case 2e with just seven species of up to
in the Zhao mechanism. Cases 2a–f refer to the mixture fraction DFmax = 0.2 causing a DTmax = 472 K. Unacceptable distortions
definition from Eq. (10), excluding oxygen and progressively omit- are also obtained for the cases 3a and 3b due to the loss of informa-
ting the less significant hydrocarbon species. Elemental mass frac- tion on particular hydrocarbons.
tions are calculated from Eqs. (11) and (12) before insertion into Figure 9 shows distortions of the mixture fraction with respect
Eq. (10). These cases 2 show the impact of particular species and to the reference case for the entire mixture fraction space. Note
from an experimental point of view they reflect a separate that mixture fraction Fall > 1 is accessed. This is a consequence of
measurement of correct species mass fractions. A conclusion can differential diffusion effects and is especially pronounced in the
be drawn regarding which species would be necessary to allow rich-premixed laminar configuration L1 (see Fig. 20). As expected,
accurate mixture fraction measurements in DME/air flames. missing species in Eq. (10) cause the mixture fraction to go down.
Deviations for 2e and 2f, using only 7 species, peak at 0.2 near
1. Reference case using all 55 species mass fractions. Fall = 0.65. The deviations are up to two orders of magnitude larger
2. (a) 16 species mass fractions, Y CO2 ; Y O2 , YCO, Y N2 ; Y H2 O ; Y H2 , than in comparable flames of methane and extend over a wide
YDME, Y CH4 ; Y CH2 O ; Y C2 H2 ; Y C2 H4 ; Y C2 H6 ; Y CH3 ; Y OH ; Y O ; Y H . range in fuel-rich conditions where significant intermediate hydro-
(b) 13 species, omitting YOH, YO, YH in 2a. carbons are present. Obviously, differences for all cases 2 decrease
(c) 12 species, omitting Y CH3 in 2b. with increasing number of species considered. In case of 13 species
(d) 11 species, omitting Y CH2 O ; Y C2 H2 in 2b. (case 2b) the deviations are even below present experimental

Table 4
Maximum deviations in mixture fraction DFmax (column 3) and resulting deviations in temperature DTmax (column 5) for various cases relative to the reference case 1 as detailed
in the text. Mixture fraction of reference case at respective maximum deviation DFmax (column 4) and DTmax (column 6) from all other cases, respectively. Columns 7, 8 provide
the minimum of sums of mass and columns 9, 10 minimum of mole fractions within the entire mixture fraction space.
P  P  P  P 
Case No. species DFmax Fall@DFmax DTmax (K) Fall, T@DTmax min iYi w Ar min iYi w/o Ar min i Xi w Ar min i Xi w/o Ar

1, Bilger w/o O All (55) – – – – 1 0.987 1 0.991

Bilger Alla 0.0056 0.238 20 0.145, 478 1 0.987 1 0.991
2a 16 0.0027 0.668 6 0.608, 1382 0.998 0.987 0.999 0.991
2b 13 0.0032 0.251 11 0.289, 1963 0.994 0.983 0.988 0.980
2c 12 0.0034 0.638 11 0.289, 1963 0.994 0.983 0.988 0.980
2d 11 0.0569 0.699 130 0.998, 516 0.971 0.962 0.974 0.968
2e 7 0.2097 0.668 472 0.492, 1854 0.935 0.926 0.920 0.914
2f 7norm 0.1741 0.638 394 0.611, 1776 1b 1b 0.920 0.914
3a 6+(7) 0.0997 0.668 170 1.037, 531 1c 1c 0.988 0.980
3b 6+(5) 0.0318 0.584 99 1.022, 461 1c 1c 0.974 0.968
a
Using the definition by Bilger et al. [34].
b
Sum of mass fractions is that from 7, but renormalized.
c
Mass fractions are calculated based on mole fractions, this implicates renormalization.
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2543

2200 fraction F⁄ was introduced. In consistence with conclusions from

2000 Sections 3.2 and 3.3, a case of 6+(5) species (CO2, O2, CO, N2,
H2O, H2) + (DME, CH4, C2H4, C2H6, and CH3) was investigated and
1800
found
 to yield  smaller deviations in mixture fraction
1600
D F all  F 6þð5Þspecies than the case including all significant hydro-
1400 carbons. However, remaining differences are still too high to be
Tx (K)

1200 ignored. Therefore, all plots in mixture fraction space shown in

all
1000 6+(7) Section 4 are based on F  ¼ F 6þð5Þspecies rather than Fall. This
6+(5) adapted mixture fraction is proposed as an appropriate basis for
800
11 comparison of Raman/Rayleigh measurements with both laminar
600 7 calculations and turbulent combustion modeling results. That is,
400 7norm
mixture fraction should be calculated from the modeled species re-
200 sults using the adapted F⁄ definition before comparison with
Δ (Tx-Tall ) (K)

experimental results in mixture fraction space.

0

-200
4. Results and discussion
-400
0 0.2 0.4 0.6 0.8 1 This part is composed of three sections. Section 4.1 provides the
Fx w/o O temperature dependent Raman response and crosstalk curves
essential for the data evaluation. In Section 4.2 important correc-
Fig. 10. Impact of differences in the mixture fraction calculation on scalars, tions of broadband and C2-LIF interferences are discussed. Finally,
exemplified here for temperature. Closed circles denote the reference case 1 using Section 4.3 shows results of laminar and turbulent DME/air flames
Eq. (10) and including all species. Blue curves shifted to the left relative to the
for the mixture compositions provided in Section 2.2.
reference case show the influence of neglecting species in Eq. (10). Red curves
represent the adapted mixture fraction calculation defined by Eq. (25). (For
interpretation of the references to colour in this figure legend, the reader is referred
4.1. Hydrocarbon Raman response and crosstalk curves
to the web version of this article.)

In conclusions from Section 3 the Raman scattering from DME/

precision denoted by rF,exp (crosses in Fig. 9). This is still satisfied air flames detected on the seven species-channels is treated as
for the case of 12 species (2c) but is not shown. Cases 3a and 3b,
being composed of 6+(5) species. The signal of the hydrocarbon
which represent the actual experimental situation, exhibit devia- channel is composed of contributions from DME, CH4, C2H4, C2H6,
tions from Fall exceeding experimental uncertainties. In contrast
and CH3. This is labeled DME + 4HCs in the following. In the matrix
to the cases 2, cases 3a and 3b show an overestimation of the mix- inversion method diagonal elements represent the temperature
ture fraction and decreasing deviation with decreasing number of
dependent Raman responses of the species CO2, O2, CO, N2, H2O,
species considered. Closer agreement between Fall and F 6þð5Þspecies H2, and DME + 4HCs. The Raman response from the DME + 4HCs
is favourable and even provides a benefit of the exclusion of C2H2
channel is calculated using Eq. (4) with information taken from
and CH2O from the processing of the Raman response. the laminar flame calculation and the relative Raman signal inten-
A distortion of the mixture fraction coordinate impacts corre-
sities from calibration measurements. Off-diagonal elements
spondingly the scalar profiles in mixture fraction space. This is represent the crosstalk caused by spectral overlap. Crosstalk is
shown for temperature in Fig. 10. The temperature of the reference
denoted here by ’X Y’. For example, CO2 crosstalk onto the O2
case 1 is plotted versus its mixture fraction Fall. Cases 2d, 2e, 2f, 3a, channel is denoted O2 CO2.
and 3b are compared to the reference case. As mentioned already,
Calculation of temperature-dependent rovibrational Raman
apparent deviations DT are due to a distortion of the mixture frac- spectra of hydrocarbon species such as DME, CH4, C2H4, or C2H6,
tion coordinate and can reach up to 500 K depending on case and
from first principles is not available or not sufficiently accurate
mixture fraction. The case 3b selected in previous sections shows to replace calibration measurements. For this reason relative
the smallest deviation.
Raman signal intensities within the hydrocarbon channel are
extracted from Raman measurements in electrically heated gas
3.7. Summary on sensitivity analysis of the mixture fraction mixtures. The temperature accessed is limited to the range be-
calculation tween 295 K and 820 K. For temperatures exceeding 820 K relative
Raman signal intensities are extrapolated linearly to 2500 K. Four
The starting point for the sensitivity analysis was that Raman different mixtures of 9% hydrocarbon in nitrogen, DME/N2, CH4/
signals of intermediate hydrocarbons and fuel spectrally overlap N2, C2H4/N2, C2H6/N2, were investigated and Raman signal intensi-
and are cumulated into a single detection channel in the context ties were measured for the entire spectral range (720–4600 cm1)
of the matrix inversion method. Consequently, the local composi- that is monitored in the experiment. Contributions from individual
tion of intermediate hydrocarbons cannot be determined directly hydrocarbon species onto the hydrocarbon channel and all other
from the experiments. Thus, using laminar flame calculations species channels (crosstalk) were thereby quantified.
two different approaches were conceived to judge on the impor- To assess the possibility of thermal decomposition of DME in
tance of particular hydrocarbon intermediates regarding the calcu- the heating process a computation of pyrolysis of the same DME/
lation of the mixture fraction. First, contributions of twelve species N2 mixture was conducted using the plug flow reactor code by Lutz
mass fractions are found to impact the mixture fraction signifi- et al. [35]. Results for the three highest temperatures are shown in
cantly. Second, hydrocarbon mole fractions are summarized in Fig. 11. For experimental-like residence times of mostly 0.15 s
XHC, reflecting the actual experimental conditions. Instead of pyrolysis is just starting at 900 K (loss in XDME < 0.001). To further
species-specific molar masses and atom numbers those of DME ensure that no pyrolysis locally occured near the heating elements,
were used for the following conversion from species mole fractions the Raman channels of two products of the pyrolysis, CO and H2,
to mixture fraction. These assumptions distort the calculated mix- were carefully monitored where no evidence of pyrolysis was
ture fraction. To account for the difference, the adapted mixture found either. However, small parts of methane and formaldehyde
2544 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

DME + 4HCs and the most important crosstalk curves, CO2 D-

0.09
ME + 4HCs, O2 DME + 4HCs, and F560 DME + 4HCs, derived
0.08 using on-chip binning. Smaller crosstalk contributions from
0.07 N2 DME + 4HCs and CO DME + 4HCs were found at low tem-
mole fractions

0.06 peratures and were corrected without temperature dependence.

0.05 Other hydrocarbon-specific crosstalk curves are of less importance
900 K
0.04 and taken from previous methane-air flame investigations (see
0.03 Appendix C). Note that for CO2 DME + 4HCs and O2 DME +
0.02 4HCs two sets of curves are shown. The upper ones were adjusted
0.01 for temperature exceeding 820 K and account for possible errors
0 introduced by the extrapolation of the crosstalk-response to higher
temperatures, additional broadband interferences, that are missed
in the F560-channel (see next section), or differences between the
0.09 species composition of the underlying 1D-calculation and those
0.08 from the measured jet flame. These corrections are justified by bet-
0.07 ter matching experimental results with laminar flame calculations
DME as shown in Appendix D. The high importance of these crosstalks is
mole fractions

0.06
CH4 demonstrated in Fig. 16. For example at the fuel-rich side of the
0.05 H2 1000 K flame the crosstalk from DME on O2 is as high as the signal from
0.04 CO O2 itself.
0.03 CH2O
The spectrometer dispersing the Raman bands has a short focal
0.02
length, which leads to optical bowing of the image of the entrance
0.01
slit and laser beam [21]. In consequence Raman bands shift relative
0
to rectangular regions of hardware binning (see Fig. B.24) causing
space-dependent variations in all response and crosstalk curves
pij(T) in Eq. (1). Whereas for diatomic species, CO2, and H2O, these
0.09
bowing effects can be accounted for by calculations [21], hydrocar-
0.08
bon species still must be treated empirically. First, the effect of
0.07 optical bowing is separated from the response and crosstalk
mole fractions

0.06 characteristics. This is expressed by

0.05
1100 K
0.04 ½pij ðTÞk ¼ ½pij ðTÞhardwarebinned
37  ½ðfbowing Þij ðTÞsoftwarebinned
k ; ð26Þ
0.03
0.02 where the index k embraces all strip numbers from 1 to 60. In order
0.01
to take advantage of lower readout noise from hardware-binned
data the temperature dependence of curve pij is determined from
0
measured Raman signal intensities from hardware-binned data at
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 the bowing center on strip 37. The same measurements on strip
residence time (sec) 37 from spectrally resolved data are used to determine the strip-
dependent variations in the measured Raman signal intensities for
Fig. 11. Computation of thermal decomposition of DME in a flow plug reactor. the corresponding shift on strip k. The resulting correction functions
fbowing were already implied in Fig. 12 and are shown explicitly in
cannot be distinguished from DME in the Raman spectrum due to Fig. 13.
spectral overlap. Hence, a small uncertainty of approximately 2% in
the measured Raman signal from DME at 820 K is left. 4.2. Broadband and C2 interferences
For all strips the relative Raman signal from CH3 is treated as
one quarter of the CH4 response ð0:25  I Ram;CH4 Þ following the A spectrally flat background was determined around 4300 cm1
findings in Hädrich et al. [36]. In contrast to the response curves (b3) and was subtracted from all Raman channels. In addition, at
shown in Fig. 4, linear extrapolation of Raman responses and an rich mixture composition in hydrocarbon-air flames, a variety of
accounting for the effect of optical bowing [21], illustrated in intermediate species and diatomic C2 are formed. These intermedi-
Fig. B.24, are applied for each species before summing up the ate species are excited by 532-nm radiation and give rise to
hydrocarbon contributions to the final response curves using Eq. fluorescence signals interfering with rovibrational Raman bands
(4). Apart from the bowing effect, the resulting difference between and rotational H2 Raman lines. These interferences are monitored
the curves in Fig. 4 (based on constant integrated Raman signal by an additional channel located near 560 nm (F560 channel span-
ratios) and those computed here using elaborated temperature ning from 730 to 1100 cm1 in the Raman spectrum). Tempera-
dependencies for each species, is rather small (<5%, maximum at ture and bowing dependencies of the corresponding crosstalks
1400 K). This small difference is a good indication of the hydrocar- were treated empirically.
bon response curve being dominated by the mixing process of spe- First, to demonstrate the significance of the LIF interferences,
cies and not by unknown Raman signal intensities at higher Fig. 14 shows the temperature dependent signal cumulated in
temperatures. Especially for temperatures above 1400 K, the the F560 channel for both laminar and turbulent flames. These
relative importance of this response curve is attenuated due to interferences impact the Raman responses over a wide tempera-
upcoming LIF-interferences. The amount of hydrocarbons, which ture range above 1000 K with a peak between 1600 and 1700 K
are finally measured, is then dominated by the accuracy of the dependent on the mixture composition. Small scatter for the lam-
LIF-interference correction. inar cases are self-evident, but the large scatter in both turbulent
Starting with the reference strip 37 (bowing center), Fig. 12 cases with a wide spread of interfering signal intensities for tem-
shows the Raman response curve for the hydrocarbon channel peratures above 1000 K indicates intense turbulence–chemistry
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2545

CO2 ← DME+4HCs crosstalk

1.6
1 1.4

DME+4HCs response
0.9 1.2
1 including
0.8 0.8 correction

0.7 0.6
Strip 1
Strip 37 0.4
0.6 Strip 60 0.2
0.5 0
1.8

F560 ← DME+4HCs crosstalk

1.2
O2 ← DME+4HCs crosstalk

1.5 1
1.2 including
correction 0.8
0.9 0.6

0.6 0.4

0.3 0.2

0 0
500 1000 1500 2000 2500 500 1000 1500 2000 2500
T (K) T (K)

Fig. 12. Response and crosstalk curves from measuring electrically heated gas mixtures of N2/DME, N2/CH4, N2/C2H4, N2/C2H6 and using Eq. (4) in connection with laminar
flame calculation for L2 configuration, multi-component transport, a = 50 s1. For the centrally located strip 37, curves are derived from hardware-binned data. All other strips
are corrected for bowing effects relative to strip 37 using spectrally resolved data.

1.07
CO2 ← DME+4HCs bowing

1.00
1.06 Strip 1
0.98
DME+4HCs bowing

Strip 37
1.05 Strip 60 0.96
1.04 0.94
1.03 0.92
0.90
1.02
0.88
1.01
0.86
1.00 0.84
0.99 0.82
F560 ← DME+4HCs bowing

1.60
O2 ← DME+4HCs bowing

1.00
1.50
1.40 0.90
1.30
0.80
1.20
1.10 0.70

1.00
0.60
0.90
500 1000 1500 2000 2500 500 1000 1500 2000 2500
T (K) T (K)

Fig. 13. Bowing correction vs. temperature of DME + 4HCs response and its crosstalks relative to the center strip 37.

interactions. For comparison, a laminar jet with a mixture of 25/ by approximately 3450 e, which corresponds to almost half of
75 vol.% methane/air has a count level of 340 and a mixture of the total signal of 7000 e.
44/54 vol.% methane/air a level of 1000 (not shown). This is slightly The dependence of F560 response on temperature and the asso-
smaller than in corresponding DME/air mixtures, but still very ciated important crosstalk curves are plotted in Fig. 15 for central
close and one of the main reasons DME is considered a relatively strip 37 and the most exterior strips 1 and 60, respectively. The
Raman-friendly hydrocarbon fuel. However, the crosstalk of LIF bowing effect at the F560 response crosstalk curve is accounted
interferences at temperatures exceeding 1000 K is significant and for by using a similar approach as described in Section 4.1. Temper-
must be accounted for. The peak is located near 1650 K. The ature dependencies for particular channels were derived from
highest signals are found in the turbulent T2 case of up to 900, comparison with laminar calculations in physical, temperature,
where on average the mixture composition of 28.1% DME in air and mixture fraction space. To provide reasonable agreement of
(L2 configuration) generates the highest LIF interference signals the same measurements at inner and outer strips, linearly increas-
of about 650 at the F560 channel. This signal corresponds to ing (O2, CO) or decreasing (N2) bowing corrections were
31.000 photoelectrons e (rshotnoise = 176 e or 0.6%). The cross- introduced. Neither temperature- nor strip-dependence was found
talk onto the N2 channel for these specific conditions contributes for H2 at all.
2546 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

T2
F560 is found between the L1 and L2 cases. Hence, a correspond-
800 ingly smaller crosstalk is observed in the L1 case (11.4% DME in
C 2 Interf. at F560 (arb. u.)

air) at the thermo-chemical state with highest F560 impact. The

L2 CO channel is influenced by 50% LIF interferences or the
600
DME + 4HCs channel by up to 40%.

400 4.3. Results of laminar and turbulent DME/air flame measurements

T1
200
The entire procedure discussed so far is now shown in its appli-
cation by evaluating spatially resolved 1D Raman/Rayleigh data
L1 collected from the two laminar (L1, L2) and two turbulent DME/
0 air flames (T2, T1) at mixture compositions given in Tables 2 and
300 600 900 1200 1500 1800 2100
3. Overall good agreement and remaining deviations between
T (K)
measurements and laminar calculations are discussed below.
Fig. 14. LIF signal cumulated in the F560 channel from mostly diatomic C2 Figure 17 shows results for the case L2 in physical space. The radial
following excitation at 532 nm in different laminar and turbulent DME/air flames. profile spans 13.5 mm and is composed of three line-imaged Ra-
man/Rayleigh/CO-LIF measurement positions r = 0,4.5,7.5 mm
resulting in 1.5 and 3 mm spatial overlap between adjacent probe
Based on the temperature dependent Raman responses and volume locations. Generally, 100 shots were recorded at each posi-
crosstalk curves from Figs. 15 and 16 provide a survey of the tion, at the most interesting position r = 4.5 mm two sets of 100
relative crosstalk contributions (from 100-shot average at most shots were taken. For comparison in physical space the laminar
interesting position in the L2 flame, centered at r = 4.5 mm) for se- calculation at strain rate a = 50 s1 (multi-component transport)
lected channels in physical space marks, with non-linearly mapped yields best agreement in the width of the measured temperature
temperature and mixture fraction coordinates also displayed. profiles.
Examining the L2 configuration at the maximum signal level in Overall the agreement between laminar flame calculations and
the F560 channel, the crosstalk contributions from various species experimental data using the post-processing discussed above is
compared to the Raman signal from the corresponding molecule very good. This holds true as well for the hydrocarbons at the
are: (1) at the N2 channel 100%, (2) 30% at the CO2 channel, (3) DME + 4HCs channel. The measured peak temperature is lower
300% at the CO channel, (4) 400% at the DME channel, (5) 35% at than the computed peak by 60 K. About 45 K are caused by
the H2O channel, and (6) 100% at the H2 channel. Around 1850 K negligence of radiation in the simulations, another 10 K by not
almost no oxygen is left and the crosstalk dominates completely. accounting for 35% relative humidity (0.007 mole fractions) in
Since the crosstalk impact of the LIF interferences decreases with the co-flowing air-stream apparent in the experiments, and a mis-
decreasing fractions of DME in air (see Fig. 14), just small match of 10 K at the inflow boundaries (290 K measured but 300 K
deviations from a linear dependence on the signal measured at in the calculation), which is equivalent to 5 K in Tmax. Deviations

1.03 11 1.40
1.02 10 1.35
9 1.30
1.01 8
CO2 ← F560

1.25
O2 ← F560

1.00 7 1.20
F560

6
0.99 1.15
5
0.98 4 1.10
Strip 1 1.05
0.97 Strip 37 3
Strip 60 2 1.00
0.96 1 0.95
0.95 0 0.90
14 11 80
10 70
12
DME+4HCs ← F560

9
10 8 60
CO ← F560

N2 ← F560

7 50
8 6
40
6 5
4 30
4 3 20
2 2
1 10
0 0 0
11 500 1000 1500 2000 2500
10 1.4 T (K)
9
H 2O ← F560

8
H2 ← F560

1.2
7
6
1
5
4
3 0.8
2
1 0.6
0
500 1000 1500 2000 2500 500 1000 1500 2000 2500
T (K) T (K)

Fig. 15. Top left: Temperature dependent response curve at the F560 channel caused by diatomic C2 and other unspecified species LIF interferences. All other graphs:
Corresponding crosstalk curves accounting for the impact of LIF interference on the respective Raman channel.
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2547

Fig. 16. Variation of relative signal components on selected Raman channels illustrated in physical space and with non-linearly mapped coordinates of temperature and
mixture fraction for the L2 mixture composition. Colors and names in legends are chosen in descending order depending on the corresponding amount of signal for each
subplot; b3 denotes the background. The three different abscissa labels apply for the corresponding tick marks in all nine subplots. Vertical dotted lines mark the point of
maximum temperature (2030 K). Dashed lines mark the stoichiometric mixture fraction (Fst = 0.26).

in gradients at the rich side, particularly in the temperature, O2, where

DME + 4HCs, or H2O profiles apparent at r = 0  2 mm are due to
specifics of the calculation in Tsuji geometry, which are not pre- Y H  Y H;2 Y C  Y C;2
cisely comparable to those present in the jet flame. Deviations at F H ¼ and F C ¼ ð28Þ
Y H;1  Y H;2 Y C;1  Y C;2
the right lean boundary, i.e. N/O⁄ profile, are partly attributed to
the missing water in the calculation.
Maximum interferences measured on the F560 channel are at are local elemental mixture fractions based on hydrogen and carbon
r = 4.75 mm, and interference on the Raman channels is generally and subscripts 1 and 2 refer to the inflow conditions of the main jet
well compensated for in the resulting species mole fractions. Slight and co-flowing air stream, respectively.
deviations at the overlapping positions between the data from the Due to the large amount of molecular hydrogen (twice that in
three different measurement locations occur. This is attributed to methane/air flames), differential molecular diffusion is prominent
variations in the index of refraction field causing different out-of- in DME/air flames. Conditional average values are very close to
focus effects at both ends of the line segment at different positions. the prediction by the laminar calculation assuming the detection
This effect was accounted for in the throughput-normalizations of of five hydrocarbon species. This provides additional confidence
DME, O2 and N2 by a linear correction of up to 1.2%, +2.1%, and in the applied Rayleigh cross-section model as well as the correct-
+1.7%, respectively. Mole fractions of CO and CO2 as well as the ness of the temperature dependence of all Raman response- and
C⁄/H⁄ ratio are generally higher in the experiment. A similar trend crosstalk-curves. Note that conditional averages are omitted for
is observed in the L1 configuration as well (Figs. 19 and 20), which values of F⁄ greater than unity, which result from strong differen-
may also be due to the use of 1D opposed flow calculations to tial diffusion effects in these laminar DME/air flames. Good agree-
approximate the scalar structure of a 2D laminar jet flame. ment in the profile of the molecular diffusion parameter z⁄ is
Results from the same L2 measurements are shown in mixture observed over the entire mixture fraction space, but underpredict-
fraction space F⁄ in Fig. 18, with the addition of temperature calcu- ed significantly between F⁄ = 0.4 and 0.6, which is consistent with
lated from total number density and the measured laboratory pres- the deviation in the C⁄/H⁄ ratio.
sure, the ratio of the two measured temperatures, and the All plots include simulated profiles for 6+(7) species (blue line),
differential diffusion parameter from Barlow et al. [8]. Here z⁄ including C2H2 and CH2O and using the corresponding F⁄ for 6+(7)
adapted to F⁄, such that species following the procedure in Section 3.6. However, better
matching is obtained with the profiles comprising 6+(5) species.
z ¼ F H  F C ; ð27Þ The mismatch of the simulation considering 6+(7) species
2548 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

0.3 0.8
DME+4HCs 0.2
0.25 DME+6HCs 0.75
DME
0.15
xHy

0.2 0.7
XDME+C

2
XO

XN
0.15 0.1 0.65
0.1 0.6
0.05
0.05 0.55
0
0 0.5
0.12 0.2
0.08
0.1
0.15
0.08 0.06
2

2O
XCO

2
0.06 0.1

XH
XH
0.04
0.04
0.05
0.02 0.02
0 0 0
0.18 700

C2 Interf. at F560 (arb. u.)

CO-Raman
0.15 CO-LIF 2000 600
500
0.12
TRay (K)

1500 400
X CO

0.09
300
0.06 1000
200
0.03 100
500
0 0
4
1 0.5
3.5
0.8 0.4
N/O*
C*/H*

0.6 0.3 3
F*

0.4 0.2 2.5

0.2 0.1 2

0 0
0.6
1
0.5 1.5
0.8 1.2
0.4
C*/O*

H*/N
C*/N

0.6 0.9
0.3
0.4 0.2 0.6
0.2 0.1 0.3

0 0 0
-4 -2 0 2 4 6 8 10 -4 -2 0 2 4 6 8 10 -4 -2 0 2 4 6 8 10
r (mm) r (mm) r (mm)

Fig. 17. Temperature, species mole fractions, LIF interferences, adapted mixture fraction F⁄, and atom ratios in physical space for the L2 mixture compositions. Results from
simulations are shown by lines, Raman/Rayleigh results by closed circles and CO-LIF by open circles. The vertical dashed line marks the stoichiometric condition at F st ¼ 0:26.

exemplifies the concentration difference due to the additional dominated by a central premixed cone. Accordingly, gradients in
contributions from C2H2 and CH2O. the reaction zone located around r = 2.5 mm are much steeper than
LIF interferences in the L2 flame are most prominent for mix- for the L2 case. This poses high standards for precise Raman mea-
ture fractions in the range of 0.4 < F⁄ < 0.6. Large scatter is observed surement at very high spatial resolution. Figures 19 and 20 show
in this range, especially for O2, and is attributed to inherent photon experimental results in comparison to laminar flame calculations
shot-noise of the O2 F560 crosstalk. Obviously, all atom ratios in physical space and mixture fraction space F⁄, respectively. The
shown in the bottom line are less sensitive to any of the differences laminar flame calculation shown in this comparison is from the op-
described above. Nearly perfect agreement between calculation posed jet geometry, with a = 50 s1 using Zhao et al. mechanism,
and experiment is observed. The impact of the corrections of the and multi-component transport. Perfect agreement between
crosstalk functions CO2 DME + 4HCs and O2 DME + 4HCs de- experiments and calculations is not expected, due to differences
scribed in Section 4.1 was investigated and is discussed in the between jet and opposed-flow geometries and also because of
Appendix D. the expected high sensitivity of the premixed reaction zone loca-
Similarly, the premixed Bunsen configuration L1 is examined. tion to radiation as shown by Barlow et al. [19] for laminar
Data were acquired at positions r = 0,3,6 mm corresponding to methane/air flames, which is not included in the calculation. Here,
3 mm overlap of the 6-mm probe volumes. The flame structure is the strain rate of the calculation was selected to best match the
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2549

0.3 0.8
Tsuji,6+(5) 0.2
0.25 Tsuji,6+(7) 0.75
6+(5): XDME 0.15
x y

0.2 0.7
XDME+C H

2
XO

XN
0.15 0.1 0.65
0.1 0.6
0.05
0.05 0.55
0
0 0.5
0.12 0.2
0.08
0.1
0.15
0.08 0.06

XH O

2
0.06
2

2
0.1

XH
XCO

0.04
0.04
0.05
0.02 0.02
0 0 0
0.18 0.18 700

C2 Interf. at F560 (arb. u.)

0.15 0.15 600
500
0.12 0.12
XCO-LIF

400
XCO

0.09 0.09
300
0.06 0.06
200
0.03 0.03 100
0 0 0
1.1
2000 2000
1.05

TRay/TRam
TRam (K)
TRay (K)

1500 1500
1
1000 1000
0.95
500 500
0.9
4
0.5 0.1
3.5
z* = F*H - F*C

0.4
N/O*

0
C*/H*

0.3 3
-0.1
0.2 2.5

0.1 -0.2
2

0 -0.3
0.6
1
0.5 1.5
0.8 1.2
0.4
H*/N
C*/N

0.6 0.9
0.3
C*/O*

0.4 0.2 0.6

0.2 0.1 0.3

0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
F* F* F*

Fig. 18. Comparison of measured and calculated results for the L2 laminar flame, including temperature, species mole fractions, LIF interferences, atom ratios, and differential
diffusion parameter z⁄ plotted versus the adapted mixture fraction F⁄ with stoichiometric condition at F st ¼ 0:26 (dashed line). Simulated profiles are shown as lines vs. F 6þð5Þ
(red) and F 6þð7Þ (blue). Black dots are results from single-shot measurements. Filled circles are conditional averages. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)

separation between the premixed reaction zone and the stoichiom- attributed to radiation, a different inflow temperature in the calcu-
etric condition, which is the portion of the flame important for lation (290 K measured vs. 298 K in the calculation), and relative
evaluation of the hydrocarbon data evaluation scheme. humidity. LIF interferences at the F560 channel are condensed to
The maximum temperatures in experiment and simulation a much smaller region in physical space and are significantly lower
match within 30 K (Tmax,calc. = 2136 K). Again, the difference is (compare Fig. 14). As observed in the L2 configuration above CO2
2550 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

0.12 0.8
DME+4HCs 0.2
0.1 DME+6HCs 0.75
DME 0.15
xHy

0.08 0.7
XDME+C

2
0.06

XO

XN
0.1 0.65
0.04
0.05 0.6
0.02
0 0 0.55

0.12 0.2
0.08
0.1
0.15
0.08 0.06
2

2O
X CO

0.06 0.1

2
XH

XH
0.04
0.04
0.05
0.02 0.02
0 0 0
120

C2 Interf. at F560 (arb.u.)

0.15 CO-Raman
CO-LIF
2000 100
0.12
80
TRay (K)

0.09 1500
X CO

60
0.06 1000 40
0.03 20
500
0 0

1.2 0.5 3.6

1
0.4 3.3
0.8
C*/H*

N/O*
F*

0.6 3
0.3
0.4 2.7
0.2
0.2 2.4
0 0.1
0.6 0.25 0.7

0.5 0.6
0.2
0.5
0.4
0.15
C*/O*

H*/N
C*/N

0.4
0.3
0.1 0.3
0.2
0.2
0.1 0.05
0.1
0 0 0
-4 -2 0 2 4 6 8 10 -4 -2 0 2 4 6 8 10 -4 -2 0 2 4 6 8 10
r (mm) r (mm) r (mm)

Fig. 19. Mean temperature, species mole fractions, LIF interferences, adapted mixture fraction F⁄, and atom ratios in physical space for the L1 flame. Results from simulations
are shown by lines, experiments by circles. The vertical dashed line marks stoichiometric condition at F st ¼ 0:59.

and CO are both higher in the experiment. For CO2 the deviation at air) were investigated at intermediate Reynolds-numbers corre-
the peak is +5%, for CO-LIF +9%, whereas for H2O 4% is observed. sponding roughly to flame D of the piloted CH4/air flame series
Accordingly, the resulting C⁄/H⁄ ratio is too high. In the L2 config- [5,4,6–9]. The probe volume was centered near the location of
uration, better matching was observed in mixture fraction space maximum fluorescence interference in each flame (r = 6 mm in
with the calculation using the F 6þð5Þ -definition. For the L1 configu- T2 and r = 4 mm in T1). Despite higher measurement-noise (see
ration some scalars are in between both definitions F 6þð5Þ and F 6þð7Þ . Table 1) the data were acquired with full spectral resolution, to al-
However, this question is very sensitive especially to the measure- low for spectroscopic analysis, and process by applying software
ment of XDME+CxHy which depends on the exact knowledge of the binning before matrix inversion. So far, the Raman response
intermediate hydrocarbon composition, its corresponding Raman regarding DME and intermediate hydrocarbon species was taken
response curve and Rayleigh cross-section model and points the same as for the laminar L2 configuration above. For the
towards limitations of the current approach. Rayleigh cross-section model, a laminar flame calculation with an
Measurements in the two turbulent piloted DME/air flames intermediate strain rate of a = 400 s1 and equal diffusivity trans-
were originally intended mainly for evaluation of levels of fluores- port in the Tsuji-geometry was used.
cence interference and they were limited to 200-shot files at a Figure 21 shows scatter plots of all measured scalars, atom ra-
single downstream location x/D = 15 (D = 7.2 mm). Mixture com- tios, and the molecular diffusion parameter z⁄ versus the adapted
positions T2 (19.7 vol.% DME in air) and T1 (11.4 vol.% DME in mixture fraction F 6þð5Þ for the case T2. Data at the single 1D probe
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2551

0.12 0.8
opp.jet,6+(5) 0.2
0.1 opp.jet,6+(7) 0.75
0.15
x y
XDME+C H

0.08 0.7

2
2

XN
0.06

XO
0.1 0.65
0.04
0.05 0.6
0.02
0 0 0.55
0.12 0.2
0.08
0.1
0.15
0.08 0.06

XH O
2

2
0.06
XCO

2
0.1

XH
0.04
0.04
0.05
0.02 0.02
0 0 0

C2 Interf. at F560 (arb.u.)

120
0.15 0.15
100
0.12 0.12
80
XCO-LIF

0.09 0.09
XCO

60
0.06 0.06 40

0.03 0.03 20
0
0 0
1.1
2000 2000
1.05

TRay/TRam
TRam (K)
TRay (K)

1500 1500
1
1000 1000
0.95
500 500
0.9
0.2
0.5 3.6 0.1
z* = F*H - F*C

0.4 3.3 0
N/O*
C*/H*

3 -0.1
0.3
2.7 -0.2
0.2
-0.3
2.4
0.1 -0.4
0.6 0.25 0.7

0.5 0.6
0.2
0.5
0.4
C*/O*

0.15
H*/N

0.4
C*/N

0.3
0.1 0.3
0.2
0.2
0.1 0.05
0.1
0 0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
F* F* F*

Fig. 20. Comparison of measured and calculated results for the L1 laminar flame, including temperature, species mole fractions, LIF interferences, atom ratios, and differential
diffusion parameter z⁄ in the adapted mixture fraction space F⁄ with stoichiometric condition at F st ¼ 0:59 (dashed line). Simulated profiles are shown as lines for 6+(5)
species (red line) and 6+(7) species (blue line). Black dots are results from single-shot measurements. Filled circles are conditional averages. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

volume location give rise to values for F⁄ primarily in the range of and for equal diffusivities a = 100, 1000, 2500 s1. The highest
0.15 < F⁄ < 0.9. Superimposed are profiles from various laminar strain rates correspond to the respective extinction limits.
flame calculations in the Tsuji-geometry, using either multi-com- The non-premixed flame structure is evident by the equally
ponent transport or equal diffusivities. For the multi-component distributed scatter and gradual gradients in all scalars. Within a
transport cases the strain is varied, with a = 100, 1000, 1750 s1, maximum deviation of 1.5% at F⁄ = 0.5, temperatures determined
2552 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

0.8
0.2 0.2
Tsuji,multi 0.75
Tsuji,equal
0.15 0.15 0.7
xHy
XDME+C

0.65

2
0.1

2
XO
0.1

XN
0.6
0.05 0.05 0.55

0 0.5
0
0.45
0.12
0.1 0.2 0.1
0.08 0.08
0.15
0.06 0.06
2

2
2O
XCO

XH
0.1

XH
0.04 0.04

0.02 0.05 0.02

0
0 0
-0.02
0.2 0.2 800
700
0.15 0.15 600

C2 Interf.
XCO-LIF

500
XCO

0.1 0.1 400

300
0.05 0.05 200
100
0 0 0

1.2
2000 2000
1.1
TRay/TRam
TRam (K)
TRay (K)

1500 1500
1

1000 1000 0.9

500 500 0.8

0.6 4.5 0.3

0.5 4 0.2
z* = F*H - F*C

0.4 3.5 0.1

C*/H*

N/O*

0.3 0
3
0.2 -0.1
2.5
0.1 -0.2
2
0 -0.3
0.8 0.35 1
0.7 0.3
0.8
0.6 0.25
0.5 0.6
C*/O*

C*/N

H*/N

0.2
0.4
0.15 0.4
0.3
0.2 0.1
0.2
0.1 0.05
0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
F* F* F*

Fig. 21. Scatter plots for the turbulent partially premixed flame T2 using the adapted mixture fraction coordinate F⁄ with stoichiometric condition at F st ¼ 0:36 (dashed line).
Filled symbols are conditional averages. Lines are from laminar flame calculations as outlined in the text. Data were taken using software-binning. This explains part of the
stronger scatter compared to the laminar cases.

via Rayleigh and Raman scattering agree very well (see TRay/TRam) C⁄/H⁄- and z⁄-profiles, particularly in the range 0.5 < F⁄ < 0.8, where
supporting consistency in the data processing. Considering tem- differential diffusion effects are apparent, and the conditional
perature and main species profiles, the conditional means are best mean profiles of these two quantities are clearly better matched
matched by laminar flame calculations using equal diffusivities at by the multi-component transport model. It is apparent that both
intermediate strain. Different conclusions are drawn from both the molecular diffusion and turbulent transport affect scalar transport
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2553

tion model seems to be the best compromise for the T2 configura-

1800
tion to minimize the uncertainty on the average values of the
Rayleigh temperature measurement. It is of course a significant
1500 sensitivity and must be kept in mind as a possible systematic influ-
ence parameter when evaluating experimental Raman/Rayleigh
TRay (K)

data from DME flames, especially close to extinction and when

1200
even deriving observations from single-shot measurements.
Additionally, in Fig. 22 the impact of the Rayleigh cross-section
900 model on the hydrocarbon measurement is shown to be rather
small. Interestingly, the amount of hydrocarbons is smaller for
higher Rayleigh temperatures measured. From the decrease of
600
the Raman response versus temperature (Fig. 4) the opposite
would be expected. The explanation of this apparent contradiction
is to be sought in the fact that the measured amount of hydrocar-
0.16 bons in physical space is actually increasing as expected, but at the
same time, its impact on the mixture fraction calculation is domi-
nating and causing a shift of the coordinate to larger values. Finally,
0.12 it is worth noting that the same comparison in physical space is
x y
X DME+C H

different. There, deviations in time-averages of temperature are

strongly attenuated due to intermittency and are found below 2%
0.08
between models of strain rates a = 100 s1 and a = 2500 s1. How-
ever, this of course is not true for evaluation of single-shots and
0.04 generally, a good agreement with the simplified analysis presented
at the end of Section 3.4 is found.
Data measured in the T1 configuration are shown in Fig. 23.
0 These data were processed using response curves and Rayleigh
0.5 0.6 0.7 0.8 0.9 1
cross-section model from the laminar calculation using the op-
F*
posed jet geometry, L1 composition, a = 100 s1, and multi-compo-
nent transport. This calculation and those from the Tsuji geometry
for both transport models at different strain rates (a = 100–
2250 s1) are superimposed. The opposed jet geometry clearly bet-
ter match premixed flames. The probe volume is centered at
r = 4 mm and comprises mostly mixture fractions F⁄ > 0.6.
For temperature and main species mole fractions a bimodal dis-
tribution appears that is typical for premixed flames: The reaction
can occur without further mixing of fuel and oxidizer at almost
identical values of mixture fraction. Spatially thin reaction zones
reduce the probability measuring temperatures, educts, and prod-
ucts at intermediate states. Broadband and C2-interferences are re-
Fig. 22. Various temperature dependent Rayleigh cross-section models applied to
duced significantly compared to the T2 flame (same relative units
the data evaluation in the turbulent T2 measurement. One calculation (solid line) is
plotted as a guide line for the experimental data shown as conditional average in F⁄ in Figs. 21 and 23 in the scatter plot showing the level and posi-
space. tions of interferences). Conditional means are plotted only up to
F⁄ = 0.9 to avoid the mixed influence from burned and unburned
samples. Differential diffusion effects are well observable in all
in this flame, which is qualitatively consistent with results on profiles. In addition to the C⁄/H⁄ and z⁄-profiles that are well repro-
transport effects in the CH4/air piloted flame at similar Reynolds duced by the multi-component opposed jet calculations, mixture
number reported by Barlow et al. [8]. fractions F⁄ > 1 are regularly measured and laminar-like behavior
As mentioned in Section 3.3, Raman and Rayleigh results are shows up for all species. A calculation with a higher strain in
interlinked due to the iterative data evaluation scheme. This was opposed jet geometry would probably better match the results.
neglected in the Rayleigh cross-section study in Section 3.4. Here, In contrast to T2, the agreement with characteristics of differential
the sensitivity of results from the T2 flame to the Rayleigh cross- diffusion is as well observed for the most sensitive H2-profile. To
section model is examined within the complete data processing exclude possible errors in the data processing as a reason for mea-
scheme. Figure 22 compares conditional average temperature re- suring F⁄ > 1, its sensitivity was investigated against different
sults within the range 0.5 < F⁄ < 1, based on Rayleigh cross-section hydrocarbon Raman responses, Rayleigh cross-section models,
models derived from different laminar calculations. The calculation and variations in the interference-corrections. But neither of these
giving the overall best match in Fig. 21 (a = 400 s1, equal diffusiv- was found to influence the results shown in Fig. 23 significantly.
ity) is shown as a guide line in Fig. 22. First, an unacceptable distor- The influence of differential diffusion, however, should be investi-
tion of the temperature profile is found for data evaluation without gated in more depth in future studies.
applying any Rayleigh cross-section model, but using the constant
Rayleigh cross-section of DME. This would lead to an over estima-
tion of the temperature of up to 20%. Still up to 10% difference is 5. Summary and conclusions
found between models composed of the lowest and highest strain
rates, where higher strain rates cause higher temperatures. A few The feasibility of measurements using line-imaged Raman/Ray-
percent higher temperatures are found for the models with under- leigh scattering and line-imaged CO-LIF in laminar and turbulent
lying multi-component transport calculations. Hence, choosing an DME/air flames was investigated. DME was chosen as the next
intermediate strain rate of a = 400 s1 for the Rayleigh cross-sec- more complex fuel-candidate beyond methane due to significant
2554 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

0.12 0.8
Tsuji,multi 0.2
0.1 0.75
Tsuji,equal
0.08 Oppjet,multi 0.15
xHy

0.7
XDME+C

0.06

2
0.1

XO

XN
0.65
0.04
0.05 0.6
0.02
0 0.55
0
-0.02 0.5
0.12 0.12
0.2
0.1 0.1
0.08 0.15 0.08

2
2O
0.06 0.06
2

XH
XCO

0.1

XH
0.04 0.04
0.02 0.05 0.02
0 0 0

0.2 0.2 350

300
0.15 0.15 250

C2 Interf.
XCO-LIF

200
XCO

0.1 0.1
150
0.05 0.05 100
50
0 0 0

1.2
2000 2000
1.1
TRay/TRam
TRam (K)
TRay (K)

1500 1500
1
1000 1000 0.9

500 500 0.8

0.6 4.5
0.4
0.5 4
z* = F*H - F*C

0.4 0.2
3.5
C*/H*

N/O*

0.3 0
3
0.2 -0.2
2.5
0.1
2 -0.4
0
0.6 0.25 0.7

0.5 0.6
0.2
0.5
0.4
0.15
C*/O*

H*/N

0.4
C*/N

0.3
0.1 0.3
0.2
0.2
0.1 0.05
0.1
0 0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
F* F* F*

Fig. 23. Scatter plots for the turbulent partially premixed flame T1 using the adapted mixture fraction coordinate F⁄ with stoichiometric condition at F st ¼ 0:59 (dashed line).
Lines are from laminar flame calculations as outlined in the text, filled symbols are conditional averages. Response curves and Rayleigh cross-section model are from the
opposed jet geometry, L1 composition, a = 100 s1. Data were taken using software-binning. This explains part of the stronger scatter.

lower interference levels than observed in flames of ethylene, eth- levels of intermediate hydrocarbons arise in the reaction zone than
ane, and propane. The objective was to extend well-established in similar methane/air flames. In the present measurements
methods applied to methane/air flames to allow for quantitative important intermediate hydrocarbons, i.e. methane, ethylene, and
measurements of major species concentrations and temperature ethane were not distinguished from DME due to very similar
in DME/air flames. Raman shifts. Formaldehyde and acetylene have different Raman
In DME/air flames a number of additional complications were shifts and were not detected at all. This impacted both, the
quantified by using 1D laminar flame computations. Much higher measurement of major species by Raman scattering and the
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2555

temperature measurement by Rayleigh scattering. In order to ac- Appendix A. Rayleigh cross-sections of major species and
count for this fact, distributions of hydrocarbons were taken from intermediate species (>0.001 mole fractions)
1D laminar flame calculations to provide additional information for
the data processing of the measured Raman and Rayleigh signals. In DME/air flames various intermediate species in addition to
For the processing of both signals, temperature dependent models educts and products are identified that contribute to Rayleigh scat-
were derived to account for very different scattering properties of tering (see Section 3.1). Referring to Eq. (5) the effective Rayleigh
DME and intermediate hydrocarbons. cross-section contains mole fraction weighted species-specific
Measurements were obtained in two laminar jet flames of DME/ cross-sections. These cross-sections are needed as a basis for the
air with stoichiometric values of mixture fraction Fst = 0.59 and discussions in Section 3.3.
Fst = 0.26. The proposed temperature dependent models were ap- In this study data from the literature were screened. For a few
plied to the processing of the Raman and Rayleigh signals. Gener- hydrocarbons additional measurements were performed.
ally, good agreement of species mole fractions and temperature Table A.5 summarizes Rayleigh cross-sections and depolarization
with 1D laminar flame calculations was observed for both flames ratios relative to N2 for different excitation wavelengths. Values
in physical space. Locally, major differences are found in values in bold type were used for the Rayleigh cross-section model
of CO2, CO, and the C/H atom ratio, which were all underpredicted discussed in Section 3.4. DME was substituted by 8.9, the average
by the calculation. The highest temperatures were overpredicted value measured in laminar jet flows. Some values for common spe-
due to negligence of radiation in the calculation. Some of the cies like C3H8, He, NO, C2H5OH, C2H4O, C3H6O were actually not
observed differences may be due to 2D-effects apparent in jet used in this study but derived and listed as well.
flames, which cannot be accounted for in 1D opposed-flow calcula- Miles et al. [37] reviewed methods for deriving Rayleigh cross-
tions. In particular, at the oxidizer side of the premixed Bunsen sections. Three components contribute to Rayleigh signals: Placzek
configuration (Fst = 0.59) all scalar gradients are more gradual in trace scattering, Q-branch rotational Raman scattering, and O- and
the measurement. The approach to select the strain rate in the S-branches rotational Raman scattering. For the case of monochro-
calculation to best match the separation between the premixed matic excitation polarized linearly and perpendicularly to the
reaction zone and the stoichiometric condition is not able to pre- plane spanned by laser beam and detection directions, and an
dict correct gradients at both sides of this flame. arrangement of the detection direction perpendicular to the laser
Another complication was identified for the conversion from beam propagation, Eq. (A.1) provides the (differential) Rayleigh
measured species mole fractions to species mass fractions and cor- cross-section in dependence on refractive index n, excitation wave-
responding mixture fraction. The difference between the molar length k0, number density N, and depolarization ratio q0 for unpo-
mass of DME and the molar masses of intermediate species is large larized (natural) incident light. This expression has been taken
and additional assumptions were introduced to provide consistent from Miles et al. [37, Eq. (5.2) from Table 5 and Eq. (1.1) from
comparison between calculations and experiments in an adapted Table 1] for the total components of the scattered light:
mixture fraction space. Thereafter, good agreement was also found
 
for the two laminar cases in this adapted mixture fraction space. @ rRay 4p2 ðn  1Þ2 6
Towards measurements in turbulent DME/air flames sensitivi- ¼ : ðA:1Þ
@X N2 k40 6  7q0
ties of the introduced models for the Raman and Rayleigh data pro-
cessing were investigated with respect to strain rate and transport In molecular hydrogen pure rotational Raman lines are spectrally
model. Significant differences were observed between models well separated from the Rayleigh line (S(0) 10 nm for 532 nm
based on lowest and highest strain rates, whereas smaller differ- excitation) and do not contribute because of the 10-nm bandpass
ences were found between multi-component and equal diffusivity filter used during all experiments in front of the Rayleigh camera.
transport models. For H2 therefore the following Eq. (A.2) was used excluding the pure
Data from first measurements obtained in turbulent DME/air jet rotational O- and S-branch Raman scattering:
flames with stoichiometric values of mixture fraction Fst = 0.59 and
Fst = 0.36 and Reynolds numbers of 25,000 were processed with   
@ rRay 4p2 ðn  1Þ2 6 8  7q0
the proposed approach and compared with results from laminar ¼ : ðA:2Þ
@X N2 k40 6  7q0 8
flame calculations in the adapted mixture fraction space. As for
previously studied methane/air jet flames at similar conditions,
This equation was derived from Miles et al. [37, Eq. (5.1) from
the turbulent DME/air flame results showed some areas of agree-
Table 5, Eq. (1.1) from Table 1, VK00 + 0 VK000 = 180 + 7 from Table 4,
ment with laminar calculations based on equal diffusivities and
and Eq. (51)].
some areas of agreement with calculations based on multi-compo-
In the literature the depolarization ratio for particular mole-
nent transport, indicating the importance of both turbulent trans-
cules is mostly listed for linearly polarized incident light. However,
port and molecular diffusion. A higher impact of differential
both can be expressed in terms of spatially-averaged polarizability
diffusion is generally observed in all DME/air flames due to much
properties of the scattering molecule and can be converted into
higher levels of molecular hydrogen.
their corresponding counterpart. For the specific detection geome-
try described above, the depolarization ratio for unpolarized inci-
dent light is given by
Acknowledgments
6Da2
F. Fuest and A. Dreizler kindly acknowledge financial support by q0 ¼ ðA:3Þ
45a
 2 þ 7Da2
Deutsche Forschungsgemeinschaft SFB 568 and EXC 259. Work
performed at Sandia was supported by the Division of Chemical and for linearly polarized incident light
Sciences, Geosciences and Biosciences, Office of Basic Energy
3Da2
Sciences, US Department of Energy. Sandia National Laboratories q¼ ; ðA:4Þ
is a multiprogram laboratory operated by Sandia Corporation, a 45a
 2 þ 4Da2
Lockheed Martin Company, for the United States Department of where a is the mean polarizability, Da is the anisotropy, both being
Energy under Contract DE-AC04-94-AL85000. The help of R. Har- invariants of the polarizabilty tensor of a molecule, see Long [38].
mon during the experiments is gratefully acknowledged. Using Eqs. (A.3) and (A.4), the following expression is found
2556 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

Table A.5
Differential Rayleigh cross-sections relative to N2 and depolarization ratios q for linear polarized incident light and specific detection geometry outlined in the text. Cross-sections
used in the present work are highlighted by bold types and are based on Eqs. (A.6) and (A.7).

Molecule 100qb,g,h,i,j,k ð@ r=@ XÞ=ð@ r=@ XÞN2

532 nmexp 488 nma, 632 nmb,c, staticd 532 nme,f

g a c
CO2 4.0798 2.32 , 2.3866 2.3907
O2 2.9434h 0.85a, 0.855b 0.8592
CO 0.5132g 1.25a, 1.237b, 1.2346c 1.2446
N2 1.0612h 1 1 1
CH4 0.02k 2.14 ± 0.3 2.16a 2.1337, 2.2016f
H2O 0.03i 0.71a 0.6946
H2 0.9044h 0.22a, 0.216b 0.2124
OH 1.4859
Ar 0 0.87a 0.8650
O 0 0.17a 0.1713
H 0 0.15a 0.1479
DME 0.3679g 8.9 ± 0.35 8.5841c 8.6473
C2H4 1.2411g 4.7 ± 0.7 5.85a, 5.776b, 5.7312c 5.8029
C2H6 0.1847g 6.22 ± 0.9 6.33b, 6.3558c 6.2984
C3H8 0.2061g 12.2 ± 1.7 12.6835c 12.7542
CH2O 0.95j 1.99d
C2H2 1.8834g 4.01b, 3.9658c 4.0096
CH3 1.5770f
He 0 0.013a 0.0132
NO 1.54b 0.9834
C2H5OH (ethanol) 8.0026
C2H4O (acetaldehyde) 0.3292g 6.7971
C3H6O (acetone) 0.5862g 13.2247
a
From Carter [39].
b
Exp. at 632.8 nm from Bridge and Buckingham [40].
c
Exp. at 632.8 nm from Bogaard et al. [42].
d
Static calc. from Bacskay et al. [33], referenced on calculated static value for N2 from Pecul and Rizzo [44], Table 1, d-aug-cc-pVQZ.
e
From exp. refractive indices at 532 nm from Gardiner et al. [32].
exp
Present measurements.
f
From refractive indices based on atomic and bond refractivities at 532 nm from Gardiner et al. [32].
g
Linear interpolation to 532 nm from exp. using 514.5 nm and 632.8 nm from Bogaard et al. [42].
h
Linear interpolation to 532 nm from exp. using 488 nm and 632.8 nm from Rowell et al. [41].
i
From exp. at 515.5 nm from Murphy [43].
j
Derived from static calc. by Bacskay et al. [33], q ¼ Da2 =ð15a 2 þ 4=3Da2 Þ.
k
From Sneep and Ubachs [45].

2q tions and sensitivities regarding depolarization ratios, treatments

q0 ¼ : ðA:5Þ
1þq of CH3 and CH2O and the experimental measurements performed
to cross-check the Rayleigh cross-sections of DME, CH4, C2H4, and
Insertion of Eq. (A.5) in (A.1) yields the expression for the Rayleigh C2H6 versus values listed in the rightmost column of Table A.5.
cross-section in terms of the depolarization ratio q for linear polar-
ized incident light with all three components included A.1. Impact of uncertainties in depolarization ratios upon Rayleigh
  2 cross-sections
@ rRay 4p ðn  1Þ 6 þ 6q
2
¼ ðA:6Þ
@X N2 k40 6  8q
The contribution of the depolarization ratio q on a species-
and insertion of Eq. (A.5) in (A.2), the expression excluding the pure specific cross-section constitute a few percent only. Thereby the
rotational O- and S-branch Raman scattering sensitivity of species-specific total Rayleigh cross-sections upon
the depolarization ratio generally is low. Uncertainties in q result-
  
@ rRay 4p2 ðn  1Þ2 6 þ 6q 4  3q ing from measurement uncertainties are of minor impact and neg-
¼ : ðA:7Þ ligible. For example CO2 exhibits the largest depolarization ratio.
@X N2 k40 6  8q 4 þ 4q
Even assuming an error of ±10% in qCO2 , which is five times larger
These two latter equations are the base to calculate values in the than given by Bogaard et al. [42] for the experimental uncertainty
rightmost column in Table A.5. All values are listed normalized to in q CO2 impacts rRay;CO2 by just ±1% using Eq. (A.6).
nitrogen. Following the procedure applied for k0 = 488 nm by Carter
[39] refractive indices were calculated from Gardiner et al. [32] – A.2. Linear interpolation of depolarization values
except for CH2O where no data on the refractive index are available
– using ðn  1Þ ¼ a=ðb  k2 0 Þ with k0 = 532 nm. Depolarization As mentioned values for depolarization ratios q were derived by
ratios for linear polarized incident light are listed in the second interpolation using values at 488 nm or 514.5 nm and 632.8 nm.
column of Table A.5 based on experimental values from Bridge Exemplified by N2, interpolation errors are negligible. Using the
and Buckingham [40], Rowell et al. [41], Bogaard et al. [42], Murphy dispersion relation for N2 calculated by Pecul and Rizzo [44] from
[43], and a static calculation by Bacskay et al. [33]. Most of them quantum mechanical ab initio methods, values for 488, 532 and
were derived from interpolation to 532 nm using values for 632.8 nm were derived. The linear interpolation using 488 and
488 nm and 632.8 nm or 514.5 nm and 632.8 nm. In the following 632.8 nm mismatches the value at 532 nm by less than 0.05%,
some more background information is provided detailing assump- which is below the smallest experimental uncertainties discussed
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2557

by Bogaard et al. [42] and does not impact the calculation of the N2 A.5. Experimental determination of Rayleigh cross-sections for DME
Rayleigh cross-section significantly. and selected hydrocarbons

Combined Raman/Rayleigh scattering measurements were per-

A.3. Treatment of CH3 formed in air, DME/N2 and DME/air mixtures issuing from the jet
nozzle. Using pure DME jets was not possible with the present
For the methyl radical CH3 no measurements of refractive index focusing of the laser beam because of optical breakdown, i.e.
and depolarization ratio are available. Therefore the refractive described by Raizer [46], even at the lowest laser pulse energy.
index was estimated using atomic and bond refractivities from Measurements of DME/air and DME/N2 were necessary to evalu-
Gardiner et al. [32]. Applied to CH4 indeed this estimation provides ate the unknown Raman crosstalk from DME on N2 (quantification
values quite close to the value derived from Eq. (A.6) based on a of this crosstalk was complicated by the fact that no data without
measured refractive index (rightmost column in Table A.5 for presence of N2 was available in the present investigations). This
CH4: cross-section derived by atomic and bond refractivities is was achieved by the following steps. First, the crosstalk of DME
2.2016 and deviates by less than 4% from the value based on mea- upon the O2-channel (O2 DME) was quantified from the DME/
sured refractive index 2.1337). This provides confidence that the N2 jet measurements. Second, the DME/air jet was examined.
calculated value 1.5770 of the relative cross-section for CH3 is Using the corrected O2 mole fractions based on the crosstalk
trustworthy. The depolarization ratio of CH3, however, is arbitrarily O2 DME, the crosstalk of DME upon the N2-channel was ad-
set to zero. This is justified by the generally low impact of the justed to recover the correct ratio of N2/O2 mole fractions in air.
depolarization ratio upon the cross-section that is negligible com- For each mixture composition and temperature setting 100 single
pared to the uncertainties in the refractive index. shots were recorded. The reference gas temperature was moni-
tored by a thermocouple. Inaccuracies of the flow controllers were
A.4. Treatment of CH2O avoided by measuring absolute Raman scattering from O2 before
recording the different DME/air mixtures. The DME/N2 measure-
The Rayleigh cross-section for CH2O was derived from calcula- ment was used to quantify the crosstalk from DME on O2 and ac-
tions of static polarizabilities by Bacskay et al. [33] in the molecule counted for it in the DME/air measurements. The mole fraction of
fixed coordinate system. These were converted to the polarizability DME was specified by 1  Xair, where Xair was determined from
tensor invariants by the measured O2 concentration adding the corresponding parts
of N2, CO2, Ar and H2O in air. By this procedure mole fractions
1 of DME, O2, N2, CO2, Ar, and H2O in the jet were known and sub-
a 2 ¼ ðaxx þ ayy þ azz Þ2 ðA:8Þ sequently used to compute the Rayleigh cross-section of DME
9
from the corresponding Rayleigh measurement to match the tem-
and perature reading of the thermocouple. The corresponding Eqs.
(A.11)–(A.16) used for these calculations are given below. This
1n 2
o procedure was repeated for various DME/air ratios on different
Da2 ¼ ðaxx  a yy Þ þ ðayy  azz Þ2 þ ðazz  a xx Þ
2
: ðA:9Þ
2 days, comparing shot-averaged and single-shot evaluation. Finally,
the resulting value for the Rayleigh cross-section of DME relative
These were inserted into to N2 was 8.9 ± 0.35. Within the error margins this value coincides
with values listed in the rightmost column of Table A.5. Uncer-
ð@ rRay =@ XÞi a 2 þ ð7=45ÞDa2i tainties of this method are dominated by the remaining uncer-
¼ 2i ðA:10Þ
ð@ r Ray =@ XÞ N2 a
 N þ ð7=45ÞDa2N
2 2
tainty in the O2 concentration. This uncertainty was below 0.5%
difference in successive measurements of O2 in air. Assuming
to obtain the differential Rayleigh cross-section for CH2O relative to ±0.5% in the O2 concentration yields an uncertainty in the relative
nitrogen. To minimize systematic deviations due to use of polariz- DME cross-section of ±0.35. This is in agreement with deviations
abilities of different excitation wavelengths as well static N2-values between different measurements.
for a
 and Da were used from Pecul and Rizzo [44]. With this proce- reff ¼ TIRay =cRay ðA:11Þ
dure one obtains a relative CH2O Rayleigh cross-section of 1.99 reff ¼ X air rair þ ð1  X air ÞrDME ðA:12Þ
listed in the forth column of Table A.5. The depolarization ratio 1
was calculated by Eq. (A.4). Note that in these calculations a static X air ¼ XO ðA:13Þ
0:20914 2
polarizability was used because of lack of values at 532 nm. How- 3 1 T 101:325kPa
ever, the wavelength dependence is rather small. Comparing for X O2 ¼ C O2  22:413996  10 m3 mol
273:15K p
example polarizabilities of N2 at 315 nm versus the static value ðA:14Þ
the deviation is about 4% Pecul and Rizzo [44]. The impact is even 
1 cRam;O2 DME
smaller when cross-sections relative to N2 are considered, as the C O2 ¼ I Ram;O2  ðIRam;DME
cRam;O2 c
mean polarizability a  rises with increasing excitation frequencies  Ram;DME 
npix; DME npix;O2
similarly for the other molecules. The change in the anisotropy  IRam;bgr  IRam; bgr ðA:15Þ
npix;bgr npix;bgr
Da in this context can be completely neglected.
TIRay  cRay X air rair
Based on Eq. (A.10) differential Rayleigh cross-sections were ) rDME ¼ ðA:16Þ
cRay ð1  X air Þ
calculated additionally for CO2, O2, CO, H2, C2H4, C2H6, and C2H2
using literature values from Bridge and Buckingham [40] and
Bogaard et al. [42] for an excitation wavelength of 632 nm. The
reff Effective Rayleigh cross-section in
anisotropy Da of C2H4 is not listed by Bridge and Buckingham
the probe volume
and was calculated using Eq. (A.4). Resulting values are listed in
T Temperature in probe volume
the forth column of Table A.5. A comparison to the values based
determined by thermocouple
on Eq. (A.6) and 532 nm in the fifth column shows a close agree-
ment with typical deviations in the order of 1%. (continued on next page)
2558 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

IRay Integrated Rayleigh scattering sured and literature data may result from impurities of the gases
signal as remarked in Bogaard et al. [42].
cRay Rayleigh temperature calibration
constant, determined in pure air Appendix B. Raman scattering from DME and intermediate
Xi Mole fraction of species i in probe species (CH4, CH2O, OH, C2H2, C2H4, C2H6, CH3)
volume
rair Relative Rayleigh cross-section of
air (0.96925) B.1. Raman spectra
rDME Relative Rayleigh cross-section of
DME This section provides the spectroscopic details on Raman scat-
0.20914 Mole fraction of O2 in air tering from DME and important intermediate species. This informa-
containing 0.0015 water tion is essential for all conclusions derived in Section 3 regarding
C O2 Concentration of O2 in the data processing of the present Raman/Rayleigh measurement
103 m3 mol (mol/l) in DME/air flames. In the following treatment selected hydrocarbon
p Pressure in laboratory, measured Raman bands and crosstalks to other Raman channels are exam-
with manometer ined. For this purpose Raman scattering from DME, CH4, C2H4, or
cRam;O2 Raman calibration constant for C2H6 diluted by 91% N2 were measured in electrically heated jets.
O2, determined in pure air The temperature was varied in steps of 100 K from 295 to 820 K
cRam;O2 DME and measured by Rayleigh-scattering. A thin-film polarizer was
Ratio of Raman calibration
cRam; DME
constants for crosstalk of DME on placed in front of the Raman spectrometer, such that only the polar-
O2 ðcRam ; O2 DMEÞ and DME ized or depolarized part was detected, respectively. Figure B.24
(cRam,DME). Determined in DME/N2 shows two relevant spectral ranges at 295 K (top) and 820 K (bot-
(9% DME) mixture. tom), respectively. The spectra comprise Raman shifts from 800
IRam;O2 Signal (counts  (averaged dark to 1800 cm1 and 2650 to 3300 cm1 containing the most intense
image)) on Raman O2 channel Raman bands. Relative signal ratios between the molecules scatter
IRam,DME Signal on Raman DME channel within ±10%. These uncertainties were caused inter alia by flow
IRam,bgr Signal on Raman background controllers, remaining uncertainties in crosstalks, temperature
channel measurements via Rayleigh scattering, and laser shot energy cor-
npix,i Number of pixels of Raman rection. Signals at the low wavenumber end were attenuated by
channel i decreasing transmission/detection efficiency by roughly 15%. From
22.143996  103m3mol1 Molar volume of ideal gas Fig. B.24 it is obvious that DME and the intermediate species C2H4,
(273.15 K, 101.325 kPa) C2H6 cause significant crosstalks at lower Raman shifts. The C2 fluo-
rescence/broadband interference channel around 760 to 1090 cm1
is affected particularly by DME and C2H6. The CO2 and O2 channels
are heavily influenced by C2H4, DME, and C2H6 but much less by
The relative Rayleigh cross-sections of CH4, C2H4, C2H6, or C3H8
CH4. For this reason CH4/air flames are less sensitive to this specific
were determined by diluting these gases with 91 ± 1.5% N2. The
crosstalk. Another crosstalk on N2 was observed in DME/air mix-
mixture composition was determined by relying on the flow con-
tures, causing a systematic increase of the N2/O2 ratio by approxi-
troller settings. Thus, differences to literature values particularly
mately 5% (for 28.1% DME in air, see N2 channel in Fig. 16).
for C2H4 and C2H6 may be caused by flow controller uncertainties.
Due to the limited spectral resolution of the transmission spec-
trometer and the complexity of the DME molecule, exact assign-
A.6. Other possible sources of experimental errors ments of the DME Raman bands evident from Fig. B.24 are
difficult. However, four separated bands I–IV are observed here
The large solid angle of the first collection lense has a different and are briefly discussed relying on bands assignments by Allan
impact on the angular dependence of the Rayleigh signal for spe- et al. [47], Blom et al. [48], and Hameka [49]:
cies with different depolarization ratios. Therefore, it affects even
values normalized to nitrogen. The effect was computed using I. 2750–3100 cm1: This is the CH-stretching region that is
Eqs. (53)–(55) from Miles et al. [37] and appropriate integration mainly constituted of five strongly polarized Raman peaks
over ±15° detection angle corresponding to the used collection at 2823 cm1, 2872 cm1, 2926 cm1, 2963 cm1 and
lense. It is found +0.1% for the values of oxygen and carbon dioxide 2999 cm1 as reported by Blom et al. [48]. Here, two Raman
and much smaller for all other molecules. Accordingly, this effect is peaks are most notably because of their apparent change in
negligible compared to other experimental uncertainties. A slightly relative signal intensities at higher temperature. First, the
bigger impact (<0.5%) can be due to variations in the index of dominating peak in cold gas here observed at 2820 cm1,
refraction in the beam path of the laser and the scattered light which probably results from Fermi resonance of symmetric
which cause slight differences in the optical imaging for different out-of-plane (C-O-C plane) CH2 stretch and in-plane CH
gases measured. In the same order all experimentally derived val- stretch vibrations as stated by Allan et al. [47]. Second, the
ues were particularly affected by the background treatment of the peak at 2926 cm1 which is very likely assigned to the
Rayleigh image. Here background contributions to the Rayleigh asymmetric CH2 out-of-plane stretch mode (Allan et al.).
signal were estimated from pixel rows above and below the line This peak is notable because it is increasing with tempera-
Rayleigh image: representative background intensities were calcu- ture, even exceeding the peak at 2820 cm1 for tempera-
lated by averaging few rows that are clearly separated from the tures above 1300 K. This observation is not compulsory an
Rayleigh image. Pixel-wise interpolation in vertical pixel direction attribute of DME alone, as species such as CH4 and C2H6
provided then an estimation of the background underlying the are formed at higher temperatures. The superposition of
Rayleigh image and was subtracted as the first step in data post- CH4 and C2H6 Raman spectra can be estimated from
processing. Another source for minor deviations between mea- chemical kinetic calculations evaluating gas compositions
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2559

200 600
DME 295 K 295 K
CH4 500

Raman signal (arb.u.)

Raman signal (arb.u.)

150 C2H4
C2H6 400

100 O2
300
CO2
200
50
100

0 0
F560 CO2 O2 in future extended Hydrocarbons

800 1000 1200 1400 1600 1800 2700 2800 2900 3000 3100 3200 3300
Raman Shift (cm-1) Raman Shift (cm-1)
80
DME 820 K 160 820 K
CH4
Raman signal (arb.u.)

Raman signal (arb.u.)

60 C2H4 apparent shift due
C2H6 to optical bowing 120
40 bowing center
80 on strip 37

20 40 max. apparent
shift on strip 1

0 0
F560 CO2 O2 in future extended Hydrocarbons

800 1000 1200 1400 1600 1800 2700 2800 2900 3000 3100 3200 3300
Raman Shift (cm-1) Raman Shift (cm-1)

Fig. B.24. Polarized Raman spectra from DME, CH4, C2H4, and C2H6, CO2, O2 at 295 K (top) and 820 K (bottom). Binning regions for LIF interference (F560), CO2, O2, and for the
hydrocarbon channel are marked by dash-dotted lines. The hydrocarbon channel already used in previous CH4/air flame measurements spans from 2800 to 3270 cm1. In
future studies, however, the low-wavenumber edge of the hydrocarbon channel will be shifted to 2740 cm1 to minimize bowing effects and is already marked in the figure.
Thus, in addition to CH4 and C2H4, Raman bands of DME and C2H6 are contained nearly completely in the hydrocarbon channel. The effect of optical bowing is illustrated by an
apparent shift (25 cm1) of the spectra (bottom, dotted lines) with respect to the binning regions. On bottom right just shown for DME for the sake of clarity.

at intermediate temperatures around 1300 K and corre- and is one quarter relative to CH4. The transferability from this nar-
sponding pure gas spectra. This is supported experimentally row peak value to the relative intensity across the entire hydrocar-
by comparing to Raman spectra from laminar jet flames. bon channel is somehow speculative and causes uncertainties.
However, the methyl radical occurs only at low concentrations
(Fig. 3), and its contribution to the hydrocarbon channel at inter-
II. 1300–1600 cm1: This range is due to CH3 deformation
mediate temperatures is rather small. At high temperatures, how-
modes. The observed DME-peak at 1453 cm1 is formed by
ever, its concentration relative to other hydrocarbons is high but its
three overlapping CH3 deformation modes, two symmetric
absolute concentration level is very low.
and one asymmetric at 1442 cm1, 1453 cm1, and 1462.5
cm1, respectively.
III. 1050–1300 cm1: CH3 rocking mode rCH3(a2) with a peak at
B.2. Relative Raman signal intensities
1131 cm1 and mCOC + rCH3(b1) at 1099 cm1, which are
both comparably weak.
Following these general discussions of species-specific Raman
IV. 800–1050 cm1: COC-stretching region and a peak at
spectra, this section provides relative Raman signal intensities that
930 cm1.
were obtained from integration of the different detection channels
Another very weak and polarized Raman scattering feature is at one center strip. In addition to the species DME, CH4, C2H4, and
observed around 2550 cm1 (not shown) that is not mentioned C2H6 measured in this study, contributions by other species such as
in the literature cited above. However, a very similar but narrower CH2O, C2H2, OH, and indirect contributions from Ar, O, and H are
feature is observed for CH4, too, which is assigned to a sublevel of discussed. Tables B.6, B.7, B.8, B.9 summarize integrated Raman
the pentad polyad (Boudon et al. [50]). signal intensities IRam,i/IRam,DME of species i relative to DME. Infor-
The other species CH4, C2H4, and C2H6 contribute differently to mation on species not measured in this work is addressed below.
the regions denoted by I to IV. The m3 vibration at 994.6 cm1 and CH2O: Information on the Raman scattering from formaldehyde
m11 at 1468.1 cm1 of C2H6 coincide with the regions IV, III, and II is available for a temperature of 423 K measured by Wiegeler and
from DME, respectively. The first very strong m3 peak of C2H4 at Bleckmann [54] while Bruna et al. [55] report on ab intio calcula-
1342.4 cm1 from symmetrical CH2 deformation/bending vibration tions. All six normal mode vibrations are Raman active, but only
contributes to regions III and II. The m2 peak at 1623.3 cm1 from three of them exhibit significant intensities. The strongest one orig-
C@C vibration (see [51]) causes severe crosstalk on O2, especially inates from the symmetric C–H stretch m1 at 2782.2 cm1 and con-
at higher temperatures and small O2-concentrations which is tributes partly to the blue end to the hydrocarbon channel but only
extremely difficult to compensate. The crosstalk from m2 from at the outside strips (approximately strip 1–5), due to the optical
CH4 [52] on O2 around 1530 cm1 is quite small compared to the bowing effect. This peak were searched in from shot-averaged
other hydrocarbons. The m1 band from CH3 has a Raman shift of and spectrally resolved data recorded in laminar jet flames of
3002 cm1. The 2m2 band at 1284 cm1 scatters on the CO2 channel, various fuels (CH4, DME, C2H4, C2H6, C3H8). Only in case of C2H4
and m1 + 2m2 is observed at 4286 cm1 [53], where the channel to with the most separated Raman band located around 3020 cm1,
monitor the background starts. The relative intensity of the a small peak is observable near 2780 cm1. This might be explained
CH3-m1 band has been measured by CARS by Hädrich et al. [36] by the fact that CH2O is mainly occurring at temperatures below
2560 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

Table B.6 Table B.8

Relative integrated Raman signals IRam,i/IRam,DME for the hydrocarbon channel. Relative integrated Raman signals IRam,i/IRam,DME for crosstalk on the CO2 channel.

Molecule Relative integrated Raman signalsfor 2798–3263 cm1 Molecule Relative integrated Raman signals for 1127–1481 cm1
295 K 820 K 295 K 820 K
DME 1 1 DME 1 1
CH4 0.64 0.64 CH4 0.064 0.09
C2H4 0.55 0.57 C2H4 2.83 3.28
C2H6 1 1 C2H6 0.51 0.54
CH3 0.16a CH3 0.02a
C2H2 0 a
CH2O 0 From Hädrich et al. [36], m1, 450 K.

a
From Hädrich et al. [36], m1, 450 K.

Table B.9
Relative integrated Raman signals IRam,i/IRam,DME for crosstalk on the O2 channel.
Table B.7
Relative integrated Raman signals IRam,i/IRam,DME for crosstalk on the F560 channel. Molecule Relative integrated Raman signals for 1484 to 1657 cm1

Molecule Relative integrated Raman signals for 761–1093 cm1 295 K 820 K

295 K 820 K DME 1 1

CH4 0.53 0.27
DME 1 1
C2H4 4.41 3.21
CH4 0 0.003
C2H6 1.19 0.85
C2H4 0.203 0.91
CH3 0.07a
C2H6 1.43 1.37
CH3 0 a
From Hädrich et al. [36], m1, 450 K.

100
DME+4HCs ← H2O crosstalk

2.2
F560 ← H2 crosstalk

2 80
1.8
60
1.6
1.4 40
1.2 Strip 1
Strip 37 20
1 Strip 60
0.8 0
N2, CO ← DME+4HCs crosstalk

2.2
1.4
2
N2 ← CO crosstalk

1.2 1.8

1.6
1
1.4
0.8
1.2
0.6 1
500 1000 1500 2000 2500 500 1000 1500 2000 2500
T (K) T (K)

Fig. C.25. Crosstalk curves supplementing information provided in Section 4.1.

1800 K and the spectral region below 2800 cm1 is dominated by sidering the intensity ratio of acetylene [54,56], its spectral
scattering from other intermediate hydrocarbons. The peaks from broadening at flame temperature [57], maximum concentrations
the C–O stretch m2 at 1745.1 cm1 (between O2 and CO) and CH2 deduced from laminar flame calculations, and experimental spec-
stretch m3 at 1500.2 cm1 (on O2) are ten times weaker than the tra, it was concluded that no signal will be detected on any of
feature at 2780 cm1 and were neglected here. the adjacent channels.
C2H2: Acetylene is considered separately here because it does Ar, O, H, OH: The atomic species Ar, O, and H are not Raman
not contribute to any Raman channels in the present experiments. active but contribute to Rayleigh scattering. Assuming room tem-
Information on fundamental vibration modes and corresponding perature, N2, and O2 calibration on exact concentrations in air,
Raman shifts for acetylene were taken from Herzberg [51]. The neglecting Argon would result in a 0.5% too high effective Rayleigh
C„C vibration Raman shift of 1973.8 cm1 is positioned between cross-section and temperature in the unburnt DME/air mixture of
the channels of O2 (O2 channel ends between 1635 and T2. This deviation decreases to about 0.1% at flame temperature.
1657 cm1) and CO (CO channel starts between 1990 and O and H atoms occur only in a narrow temperature band around
2015 cm1). The Raman band of the C–H vibration is about nine 2100 K and their negligence would evoke an error at these high
times weaker and has a Raman shift of 3373.7 cm1, which is temperatures of about +0.8%. However, the major part of this inac-
located close to the H2O channel starting at 3409–3430 cm1. Con- curacy will decrease to 0–0.3% when concurrently assuming parts
 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562 2561

0.1 Appendix D. Sensitivity of change in the Raman crosstalks

w/o correction CO2 DME + 4HCs and O2 DME + 4HCs exemplified on the L2
0.08
configuration

Exemplarily, the data evaluation of the L2 configuration was

0.06 made with and without the additional background correction
2
XCO

implied in the CO2 DME + 4HCs crosstalk (compare Fig. 12). Fig-
0.04 ure D.26 shows the impact on the CO2 mole fractions and z⁄-pro-
files in mixture fraction space. A value of up to 0.012 smaller
mole fractions is obtained using the background correction. Inde-
0.02
pendent of this correction all other scalars are very close to the
calculation. Therefore, it is assumed that this obvious mismatch
0 in the CO2-profile must be caused by an uncorrected interference
0.1 signal on CO2. The impact on any other of the profiles shown in
Fig. D.26 is negligible. This is exemplified on the most sensitive
0.05 z⁄-profile, where still no significant difference is observable. The
impact of the additional correction on O2 DME + 4HCs (compare
0
Fig. 12) was much smaller and is therefore not shown explicitly.
z* = F*H - F*C

-0.05
References
-0.1
[1] TNF, International Workshop on Measurement and Computation of Turbulent
-0.15 Nonpremixed Flames (TNF), <http://www.ca.sandia.gov/TNF>.
[2] D.G. Pfuderer, A.A. Neuber, G. Früchtel, E.P. Hassel, J. Janicka, Combust. Flame
106 (3) (1996) 301–317.
-0.2 [3] M.M. Tacke, S. Linow, S. Geiss, E.P. Hassel, J. Janicka, Proc. Combust. Inst. 27
(1998) 1139–1148.
-0.25 [4] C. Schneider, A. Dreizler, J. Janicka, E.P. Hassel, Combust. Flame 135 (1–2)
0 0.2 0.4 0.6 0.8 1 (2003) 185–190.
F* [5] R.S. Barlow, J.H. Frank, Proc. Combust. Inst. 27 (1998) 1087–1095.
[6] A.N. Karpetis, R.S. Barlow, Proc. Combust. Inst. 29 (2002) 1929–1936.
Fig. D.26. Sensitivity of the additional correction on CO2 DME + 4HCs crosstalk, [7] R.S. Barlow, A.N. Karpetis, Flow Turbul. Combust. 72 (2004) 427–448.
which was applied on the crosstalk derived from experimental spectra and species [8] R.S. Barlow, J.H. Frank, A.N. Karpetis, J.-Y. Chen, Combust. Flame 143 (4) (2005)
distributions from laminar flame calculations following Eq. (4). Errorbars denote 433–449.
the measurement precision as one standard deviation from 100 single laser shots. [9] R.S. Barlow, H.C. Ozarovsky, A.N. Karpetis, R.P. Lindstedt, Combust. Flame 156
(11) (2009) 2117–2128.
[10] S.H. Stårner, R.W. Bilger, R.W. Dibble, R.S. Barlow, Symp. Int. Combust. Proc. 23
(1991) 645–651.
of the hydroxyl radical being detected on the Raman channel for [11] A.R. Masri, R.W. Dibble, R.S. Barlow, Combust. Flame 89 (2) (1992) 167–185.
water or when even neglecting the OH contribution completely [12] C. Dreyer, T. Parker, M.A. Linne, Appl. Phys. B: Lasers Opt. 79 (2004) 121–130.
[13] W. Meier, O. Keck, Meas. Sci. Technol. 13 (2002) 741–749.
as in Sutton et al. [31]. In fact, the main Raman band from OH at [14] F. Rabenstein, A. Leipertz, Appl. Opt. 37 (21) (1998) 4937–4943.
3568.4 cm1 coincides with the Raman channel for water. Assum- [15] J. Egermann, T. Seeger, A. Leipertz, Appl. Opt. 43 (29) (2004) 5564–5574.
ing IRam; OH ¼ 0:5IRam;H2 O as described by Linow et al. [58], consider- [16] P.A. Nooren, M. Versluis, T.H. van der Meer, R.S. Barlow, J.H. Frank, Appl. Phys.
B: Lasers Opt. 71 (2000) 95–111.
ing common OH/H2O ratios around 5–6% at locations of the highest
[17] R.W. Dibble, A.R. Masri, R.W. Bilger, Combust. Flame 67 (3) (1987) 189–206.
OH concentrations would lead to a maximum overestimation of [18] A. Brockhinke, A.T. Hartlieb, K. Kohse-Höinghaus, D.R. Crosley, Appl. Phys. B:
X H2 O of 2–3%. Generally, this Raman crosstalk on H2O was partly Lasers Opt. 67 (1998) 659–665.
compensated due to a H2O calibration in flat flames with similar [19] R.S. Barlow, A.N. Karpetis, J.H. Frank, J.-Y. Chen, Combust. Flame 127 (3) (2001)
2102–2118.
equilibrium OH concentrations, and was not treated particularly [20] P.C. Miles, Appl. Opt. 38 (9) (1999) 1714–1732.
here, too. [21] F. Fuest, R.S. Barlow, D. Geyer, F. Seffrin, A. Dreizler, Proc. Combust. Inst. 33 (1)
(2011) 815–822.
[22] A.N. Karpetis, R.S. Barlow, Proc. Combust. Inst. 30 (2005) 665–672.
[23] R.S. Barlow, G.H. Wang, P. Anselmo-Filho, M.S. Sweeney, S. Hochgreb, Proc.
Appendix C. Supplementary crosstalk curves Combust. Inst. 32 (1) (2009) 945–953.
[24] C.-P. Chou, J.-Y. Chen, C.G. Yam, K.D. Marx, Combust. Flame 114 (3–4) (1998)
420–435.
Additional crosstalk curves are presented completing the infor- [25] J.A. Miller, R.J. Kee, M.D. Smooke, J.F. Grcar, The computation of the structure
mation provided in Section 4.1. The crosstalks F560 H2 and and extinction limit of a methane–air stagnation point diffusion flame, Spring
Meeting of the Western State Section of the Combustion Institute WSS/CI84-
N2 CO were calculated following the procedure outlined by 10, 1984.
Fuest et al. [21]. The crosstalk curve for DME + 4HCs H2O was [26] K. Seshadri, F.A. Williams, Int. J. Heat Mass Transfer 21 (1978) 251–253.
transferred from previous CH4/air flame measurements. So far, no [27] A.E. Lutz, R.J. Kee, J.F. Grcar, F.M. Rupley, OPPDIF: a Fortran program for
computing opposed-flow diffusion flames, Tech. Rep., Sandia National
measurements without N2 were taken. Hence, the calibration of Laboratories Report SAND96-8243, 1997.
the crosstalk from DME + 4HCs on N2 relies on a procedure that [28] Z. Zhao, M. Chaos, A. Kazakov, F.L. Dryer, Int. J. Chem. Kinet. 40 (2008) 1–18.
follows two steps: at first crosstalk upon the O2 channel was esti- [29] E.W. Kaiser, T.J. Wallington, M.D. Hurley, J. Platz, J. Curran, W.J. Pitz, C.K.
Westbrook, J. Phys. Chem. A 104 (35) (2000) 8194–8206.
mated from DME/N2 flows, then DME/air flows were investigated [30] R.J. Kee, J.F. Grcar, M.D. Smooke, J.A. Miller, A Fortran program for modeling
using the crosstalk-corrected O2 mole fraction to estimate DME steady laminar one-dimensional premixed flames, Tech. Rep., Sandia National
contributions to the N2 channel. Because the relative O2/N2 con- Laboratories Report SAND85-8240, 1992.
[31] G. Sutton, A. Levick, G. Edwards, D. Greenhalgh, Combust. Flame 147 (1–2)
centrations in air are known, excessive signal on the N2 channel
(2006) 39–48.
was attributed to DME-contributions. A more straight forward [32] W.C. Gardiner Jr., Y. Hidaka, T. Tanzawa, Combust. Flame 40 (1981) 213–219.
strategy would be to measure DME/He jets and will be done in [33] G.B. Bacskay, S. Saebø, P.R. Taylor, Chem. Phys. 90 (3–4) (1984) 215–224.
future experiments. Note that the crosstalk contribution [34] R.W. Bilger, S.H. Stsrner, R.J. Kee, Combust. Flame 80 (2) (1990) 135–149.
[35] A.E. Lutz, R.J. Kee, J.A. Miller, SENKIN: A Fortran program for predicting
N2 DME + 4HCs is assumed to be independent on temperature homogeneous gas phase chemical kinetics with sensitivity analysis, Sandia
because of lack of more detailed information (see Fig. C.25). National Laboratories Tech. Rep. SAND87-8248, 1988.
2562 F. Fuest et al. / Combustion and Flame 159 (2012) 2533–2562

[36] S. Hädrich, S. Hefter, B. Pfelzer, T. Doerk, P. Jauernik, J. Uhlenbusch, Chem. Phys. [49] H.F. Hameka, J. Mol. Struct.: THEOCHEM 226 (3–4) (1991) 241–249.
Lett. 256 (1–2) (1996) 83–86. [50] V. Boudon, M. Rey, M. Lodte, J. Quant. Spectrosc. Radiat. Transfer 98 (3) (2006)
[37] R.B. Miles, W.R. Lempert, J.N. Forkey, Meas. Sci. Technol. 12 (2001) R33–R51. 394–404.
[38] D.A. Long, The Raman Effect: A Unified Treatment of the Theory of Raman [51] G. Herzberg, Molecular Spectra and Molecular Structure, Infrared and Raman
Scattering by Molecules, Wiley, 2002. Spectra of Polyatomic Molecules, vol. II, Krieger Publishing Company, Malabar,
[39] C.D. Carter, Laser-based Rayleigh and Mie Scattering Methods, Handbook of Florida, 1991.
Fluid Dynamics and Fluid Machinery, Wiley, New York, 1996. [52] E. Jourdanneau, F. Chaussard, R. Saint-Loup, T. Gabard, H. Berger, J. Mol.
[40] N.J. Bridge, A.D. Buckingham, Proc. R. Soc. Lond. A 295 (1966) 334–349. Spectrosc. 233 (2) (2005) 219–230.
[41] R.L. Rowell, G.M. Aval, J.J. Barrett, J. Chem. Phys. 54 (5) (1971) 1960–1964. [53] P.B. Kelly, S.G. Westre, Chem. Phys. Lett. 151 (3) (1988) 253–257.
[42] M.P. Bogaard, A.D. Buckingham, R.K. Pierens, A.H. White, J. Chem. Soc., Faraday [54] W. Wiegeler, P. Bleckmann, J. Mol. Struct. 66 (1980) 273–280.
Trans. 1 74 (1978) 3008–3015. [55] P.J. Bruna, M.R.J. Hachey, F. Grein, J. Mol. Struct.: THEOCHEM 400 (1997) 177–
[43] W.F. Murphy, J. Chem. Phys. 67 (12) (1977) 5877–5882. 221.
[44] M. Pecul, A. Rizzo, J. Chem. Phys. 116 (4) (2002) 1259–1268. [56] D.A. Stephenson, J. Quant. Spectrosc. Radiat. Transfer 14 (12) (1974) 1291–
[45] M. Sneep, W. Ubachs, J. Quant. Spectrosc. Radiat. Transfer 92 (3) (2005) 293– 1301.
310. [57] A.V. Mokhov, S. Gersen, H.B. Levinsky, Chem. Phys. Lett. 403 (4–6) (2005) 233–
[46] Y.P. Raizer, Gas Discharge Physics, Springer, Berlin, 1997. 237.
[47] A. Allan, D.C. McKean, J.P. Perchard, M.L. Josien, Spectrochim. Acta, Part A 27 [58] S. Linow, A. Dreizler, J. Janicka, E.P. Hassel, Meas. Sci. Technol. 13 (2002) 1952–
(8) (1971) 1409–1437. 1961.
[48] C.E. Blom, C. Altona, A. Oskam, Mol. Phys. 34 (2) (1977) 557–571.