Professional Documents
Culture Documents
com
Proceedings
of the
Combustion
Institute
Proceedings of the Combustion Institute 33 (2011) 3225–3232
www.elsevier.com/locate/proci
Abstract
Pure rotational CARS thermometry, complemented by UV emission measurements and ICCD imaging,
is used to study kinetics of low temperature plasma assisted fuel oxidation and ignition in a repetitive nano-
second pulse discharge in hydrogen-air mixtures, with number of pulses in a 40 kHz burst varying from a
few to a few hundred. Time-resolved OH emission, coupled with gated ICCD images of the plasma and the
flame, demonstrate that volumetric ignition of H2–air mixtures occurs in a spatially uniform plasma. The
results are shown to agree well with predictions of a new hydrogen-air plasma chemistry model, which
incorporates non-equilibrium plasma processes, H2–air chemistry, non-empirical scaling of nanosecond
pulse energy coupled to the plasma, and quasi-one-dimensional conduction heat transfer. In particular,
the results demonstrate that the heating rate in low temperature hydrogen-air plasmas is much faster than
in air plasmas, primarily due to energy release from exothermic reactions of fuel with O and H atoms gen-
erated in the plasma. Kinetic sensitivity analysis is used to identify dominant plasma and chemical pro-
cesses of hydrogen oxidation, demonstrating that additional heat release in these reactions is a key
factor in ignition kinetics. Kinetic modeling calculations demonstrate that removal of the radical genera-
tion processes by the nanosecond pulsed plasma from the model completely blocks subsequent exothermic
chemical reactions, thus making ignition impossible.
Ó 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
1540-7489/$ - see front matter Ó 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.proci.2010.05.100
3226 Y. Zuzeek et al. / Proceedings of the Combustion Institute 33 (2011) 3225–3232
compared to other types of electric discharges, 40 kHz repetition rate. For all measurements pre-
due to the very low duty cycle operation [2,3]. sented in this paper, the pulser was operated in a
Several groups have recently reported the use repetitive burst mode, generating sequences of up
of nanosecond pulse discharges for fundamental to 1000 pulses at pulse repetition rate of 40 kHz
PAC studies [4]. Bozhenkov et al. [5] have shown and burst repetition rate of 10 Hz. This rate
more than one order of magnitude ignition delay matches the repetition rate of the CARS system,
time reduction in H2/Ar/air mixtures preheated and ensures that each gas sample in the flowing
in a shock tube. Other recent examples include discharge cell experiences only a single burst. To
large increases in the critical co-flow velocity [6] produce breakdown in the discharge on the first
and NO production rate [7] in methane–air PAC pulse, the test cell was irradiated by a deuterium
flames, and fuel lean flame stability limit extension UV lamp, providing pre-ionization. The equiva-
[8] in premixed propane–air flames, to / = 0.3. lence ratio in air–fuel mixtures was varied from
Uddi et al. [9] have used two photon absorption / = 0.05 to 1.0.
laser induced fluorescence (TALIF) to study low The pure rotational CARS apparatus, pat-
temperature generation and loss kinetics of atomic terned after that of Alden et al. [15], has been
oxygen in air, methane–air, and ethylene–air mix- described in detail previously and will be only
tures at 60 Torr. The results were found to agree summarized here. A broad-band Ti:sapphire laser
well with predictions of a hydrocarbon–air plasma [16] creates two, orthogonally polarized, 15 ns,
chemistry model, supplemented by GRI Mech 3.0 30 mJ/pulse, CARS “pump” beams at 780
[10] and the hydrocarbon oxidation mechanism nm, which are overlapped in a focused planar
of Wang et al. [11]. Finally, Zuzeek et al. [12] have phase matching geometry with a 532 nm “probe”
recently presented time-resolved Coherent Anti- beam, using a 400 mm focal distance lens. Pure
Stokes Raman Spectroscopy (CARS) measure- rotational CARS is generated with vertical polar-
ments of heat release in air and ethylene–air mix- ization, which allows for discrimination from
tures. Results were found to agree well with stray horizontally polarized probe light using a
predictions of the plasma chemistry kinetic model polarizer. As detailed in [12] this results in
of Uddi et al. [9], modified to incorporate a new detected CARS resonant and non-resonant inten-
model for energy coupling in repetitively pulsed sities which are a factor of 9/16 and 1/9, respec-
nanosecond plasmas [13], as well as quasi-one- tively, of the values which would be obtained if
dimensional heat transfer. all polarizations were aligned parallel [17]. The
This paper presents new experimental CARS CARS beam is spectrally resolved (resolution
thermometry results for heat release in nanosec- 1–2 Å) with a ½ m spectrometer (1800 lines/
ond repetitively pulsed H2–air discharges [14]. mm grating), and a gated ICCD camera. Typi-
Sensitivity analysis predicts that the initial heating cally, for burst durations of 10 ms or less, the
rate, at temperatures of approximately 300 K, is CARS signal was averaged up to 600 bursts
controlled by the low temperature processes (60 s). For longer bursts, five such spectra were
O + HO2 ? OH + O2 and OH + H2 ? H2O + added together.
H, where HO2 is formed by three body recombi- Inference of rotational temperature and O2
nation of O and H2. At intermediate tempera- mole fraction is performed using the Sandia
tures, 500–600 K, OH formation from chain National Laboratories CARS code [18]. Prior to
branching processes increases, with rapid concur- fitting, experimental spectra are corrected to
rent increase in heat release due to plasma chem- account for the finite spectral width (400 cm1
ical reactions, leading to rapid temperature rise – FWHM) of the broad-band Ti:sapphire pump
and, in some cases, volumetric ignition. beams, as described in [12]. CARS spectra are
modeled as pure air since the H2 (S(0)) transition
falls outside the spectral region of nitrogen and
2. Experimental oxygen. Uncertainty in rotational–translational
temperatures, as discussed in Section 4, is gener-
Experiments are conducted in a 220 mm ally found to be in the range 20–200 K.
long 22 mm width 10 mm height quartz In order to detect ignition and gauge discharge
channel/plasma flow reactor [9,12], at 40 Torr uniformity, time-resolved emission measurements
pressure and flow velocity 0.8 m/s, controlled and plasma imaging have been performed. Emis-
by mass flow controllers. Rectangular copper sion measurements detect time-resolved OH emis-
plate electrodes, 14 mm wide by 65 mm long and sion with a photomultiplier tube and bandpass
rounded at the corners to reduce the electric field filter centered at 310 ± 2 nm, captured with a dig-
nonuniformity, are flush mounted to the top and ital oscilloscope. Time response of emission mea-
bottom of the 1.75 mm thick quartz channel, surements (10 ls) was controlled using a 50 kX
which acts as a dielectric barrier. A Chemical termination resistor. Broad-band visualization
Physics Technologies power supply produces images are obtained with a gated ICCD camera
20 kV peak voltage, 25 ns duration pulses at and UV lens.
Y. Zuzeek et al. / Proceedings of the Combustion Institute 33 (2011) 3225–3232 3227
Fig. 2. ICCD images of the plasma and the flame in H2–air excited by a m = 40 kHz discharge at / = 1 and P = 40 Torr.
Emission spans entire discharge volume.
and do not, therefore, illustrate discharge devel- The bottom two images in Fig. 2 (labeled
opment during a single burst. “Flame”) were taken with an 18 ls gate and timed
The top four images in Fig. 2 were obtained to collect only emission that occurs after the dis-
with a 2 ls intensifier gate, which was timed to charge emission has decayed (3 ls delay after the
include the 25 ns discharge event. While some discharge pulse). The observed uniform emission
filamentary structure is observed for the first (principally OH) is further evidence that volumet-
pulse, all images for pulse numbers exceeding ric ignition has occurred in this case after 10 ms.
2–4 are observed to be uniform, including Figure 3 is similar to Fig. 2, except that the images
images obtained for pulse numbers up to 1000 shown were obtained at 100–3000 ls (with a
(not shown in Fig. 2). This result is in contrast 100 ls gate) after the final pulse in a 15 ms
to our previous nanosecond pulse plasma images (600 pulse) burst for u = 1 at 70 Torr. Uniform
obtained in ethylene–air mixtures, in which large OH emission, again spanning the complete dis-
scale filaments were detected after 50–100 charge volume, is clearly evident, indicating con-
pulses, depending upon equivalence ratio [12,24]. tinuous, combustion-induced, emission for a
While some evidence of filamentation was period of several milliseconds after application
observed after 100–200 pulses at pressures of of the final discharge pulse.
70 Torr or higher, these structures gradually dif-
fuse, resulting in uniform plasmas after 300 4.2. Pure rotational CARS
pulses (7.5 ms), i.e. before ignition occurred (see
Fig. 1) [24,25]. High thermal diffusivity of hydro- CARS is a well known diagnostic technique for
gen, as well as its relatively high ionization poten- high spatial and temporal resolution measurement
tial (15.6 eV), compared to ethylene (10.5 eV), is of temperature and major species distributions in
the most likely the explanation of the observed flames and plasmas, and is ideally suited as a
improvement in discharge stability. method to provide more quantitative data which
Fig. 3. ICCD images in / = 1 H2–air at 70 Torr, obtained at differing times after final pulse of 15 ms 40 kHz burst.
Emission spans entire discharge volume.
Y. Zuzeek et al. / Proceedings of the Combustion Institute 33 (2011) 3225–3232 3229
can be benchmarked against plasma chemical oxi- Fig. 4 also shows that at burst durations in the
dation kinetic modeling codes. Figure 4 shows a approximate range 15–20 ms (600–800 pulses), the
comparison of experimental CARS temperatures experimental CARS temperatures (as well as the
obtained in this work as a function of burst dura- modeling predictions) for both u = 0.5 and 1 exhi-
tion at 40 kHz discharge repetition rate in air and bit a maximum, and then decay. Furthermore, as
in H2–air at / = 0.05, 0.5, and 1.0, along with pre- shown in Fig. 5, the O2 mole fractions for the stoi-
dictions from the plasma kinetic modeling code. chiometric mixture, obtained from the Sandia
In all cases, P = 40 Torr and the initial tempera- CARS code (which floats this parameter) decrease
ture is T = 300 K. Note that the raw CARS spec- very rapidly in the region of 15–20 ms duration
tra are similar in intensity and precision to those bursts. Figure 5 also shows that the O2 mole frac-
of our previous work in ethylene–air mixtures tion decay agrees well with the kinetic model pre-
[12] and, therefore, will not be reproduced here. dictions, which also predicts large concurrent
Note also that the Sandia CARS code does not decrease in H2 mole fraction. The observed tem-
provide any estimate of uncertainty in the fitting perature maximum, coupled with the observed
parameters. To provide an estimate of experimen- rapid loss in O2, is a very strong indication that
tal precision, ten spectra were obtained and fit for ignition, accompanied by almost complete fuel
each measurement condition presented in Fig. 4. oxidation, has occurred in these mixtures, a result
The statistical uncertainty (95% confidence inter- which is consistent with the emission data and
val) in the mean, defined as 2.23 (“t” value) times images shown in Figs. 1–3.
the experimental standard deviation, divided by While the experimental temperature rise in the
the square root of 10, is shown for representative region of 15–20 ms is more gradual then that pre-
data in Fig. 4. dicted, Figs. 4 and 5 show good overall agreement
Focusing first on the experimental data in between the model and the experiments. It is
Fig. 4, it can be seen that the initial heating rate, believed, although not definitely determined, that
for burst durations less than 3–4 ms (120–160 the burst to burst reproducibility of the ignition
pulses), is approximately equal for each of the event is not highly reproducible in time, leading
mixtures to within the experimental uncertainty. to averaging of the temperature profile in the
For burst durations exceeding approximately 5– vicinity of the maximum. Nonetheless the overall
7 ms (200–280 pulses) clear trends become dis- quantitative agreement demonstrates that the
cernable. In particular, the rate of heating for mix- plasma kinetic model adequately describes the
tures at u = 0.5 and 1.0 can be clearly seen to energy balance in the repetitively pulsed nanosec-
exceed that for pure air and the u = 0.05 mixture. ond discharge, including the energy coupling to
A somewhat closer examination of the data in the plasma, energy release during exothermic
Fig. 4 shows that accelerated heating for the plasma chemical reactions, and energy loss by
u = 0.5 and 1.0 mixtures begins to occur at tem- heat transfer to the test section walls. It is stressed
peratures of approximately 500–600 K. This will that this level of agreement is attained without
be discussed in more detail in Section 5. recourse to any adjustable parameters in the
model.
Number of pulses
T, K 0 200 400 600 800 1000 1200 Mole fractions
1200 P=40 torr P=40 torr, φ =1
φ=1
0.2
600
0.1
300
0.0
0 5 10 15 20 25 30
Time, msec 0 5 10 15 20 25 30
Time, msec
Fig. 4. Comparison of CARS temperatures in a repet-
itively pulsed nanosecond discharge in air and in Fig. 5. Experimental and calculated O2 mole fractions at
hydrogen-air at P = 40 Torr with kinetic model predic- the conditions of Fig. 4. Also shown are calculated H2
tions. m = 40 kHz, burst repetition rate 10 Hz. mole fractions for the same conditions.
3230 Y. Zuzeek et al. / Proceedings of the Combustion Institute 33 (2011) 3225–3232
5. Kinetic sensitivity analysis for total thermalized energy from a single dis-
charge pulse as a function of initial temperature.
To provide insight into the chemical processes At 300 K the additional thermalized energy
which are primarily responsible for the observed (beyond the 0.28 meV/molecule coupled directly
low temperature oxidation, heat release, and igni- from the discharge) is minimal, approximately
tion in hydrogen-air plasmas, a sensitivity analysis 10%. However, as the temperature is increased,
has been performed, the principal results of which the additional thermal energy due to net exother-
will be summarized below. A more complete anal- mic plasma chemical reaction increases rapidly,
ysis can be found in [24,25]. reaching a factor of approximately three greater
At the lowest temperatures relevant to this than the pulse input energy at 600 K. It is this
work, 300–600 K, a detailed analysis is given rapid increase in plasma chemical oxidation which
in [25]. Summarizing it is found that at 300 K causes the observed rapid temperature rise of the
plasma chemical oxidation proceeds, principally, u = 0.5 and 1.0 mixtures, relative to u = 0.05
by the following mechanism. First, O and H are and pure air, observed in Fig. 4 after 5 ms
formed during, and after, the discharge pulse (200 pulses), and leads, ultimately, to the observed
(on a time scale of 100–1000 ns) primarily via ignition.
electron impact and dissociating collisions with Figures 7 and 8 summarize the most significant
electron metastable states of N2. Subsequent fuel sensitivity analysis results for burst mode kinetics
oxidation during the post-discharge stage (on a and oxidation. Figure 7 plots time-resolved tem-
time scale from 1 ls to 1 ms) occurs primarily peratures predicted by the model at u = 1.0,
via a three process sequence, P = 40 Torr using both the full H2–air chemistry
H þ O2 þ M ! HO2 þ M ð1Þ model, and a reduced kinetic mechanism consist-
ing of reactions (1)–(8) plus the additional
O þ HO2 ! OH þ O2 ð2Þ process,
OH þ H2 ! H þ H2 O ð3Þ OH þ HO2 ! H2 O þ O2 ð9Þ
In other words, at this low temperature, OH for- The full and reduced kinetic mechanisms result
mation via the well known “high temperature” in very similar predictions, both of which agree
chain processes, well with the experimental data. Figure 8 shows
Oð3 PÞ þ H2 ! H þ OH ð4Þ predictions of the full kinetic model for the tem-
poral evolution of key species for the same condi-
H þ O2 ! O þ OH ð5Þ
tions as Fig. 7. It can be seen that H, O, and OH
is extremely slow, while processes (2) and (3) are mole fractions are predicted to peak at 18 ms,
relatively fast, with activation barriers of consistent with the observed maximum in the
14.35 kJ/mol (1660 K) and essentially zero, experimental CARS temperatures observed in
respectively [14]. Fig. 4. In addition, while not plotted here, it is
As the initial temperature is increased, to 500– shown in [24] that the reduced kinetic model
600 K, the rates of processes (4) and (5), as well as
of the following three processes,
Energy, meV/molecule
H þ HO2 ! H2 O þ O ð6Þ 1.5
H þ HO2 ! OH þ OH ð7Þ H2-air, P=40 torr
900
6. Summary and conclusions