This document summarizes research using terahertz time-domain spectroscopy and tomography to sense and image within sooty flames. A bespoke high-pressure burner was designed to produce methane diffusion flames at different air co-flow rates, allowing line-of-sight THz measurements through the flames. THz time-domain spectroscopy measurements were taken of flames at various air co-flow rates from 3-12 l/min. The time delay values extracted from the THz waveforms revealed that optical density along the sensing path initially increases linearly with co-flow, then levels off after 12 l/min, indicating changes in radial temperature distribution. Varying the air co-flow provides a way to study flame properties with potential for

1. T
Sensing in Sooting Flames: THz Time-Domain
Spectroscopy and Tomography
H. G. Darabkhani, M. Banuelos-Saucedo, J. Young, M. Stringer, P. Wright, Q. Wang,
Y. Zhang, R. E. Miles and K. B. Ozanyan, Senior Member, IEEE
Abstract— We present line-of-sight measurements with
broadband THz radiation through a methane diffusion flame,
with the motivation to develop a modality for sensing and
imaging within sooty flames.
We present the design details of a bespoke high-pressure
burner suitable for THz line-of-sight measurements, as well as
hard-field tomography imaging from 4 projections. Further we
introduce the setup for THz time-domain spectroscopy (TDS)
and show the results of TDS measurements of a methane-air
diffusion flame at different values of peripheral air co-flow at 3,
5, 10 and 12 l/min. The time-domain waveforms of the THz
electric field transmitted through the flame are processed to
extract the values of the time delay for each case. The dependence
of these values on the air co-flow reveals that after an initial
linear increase, the optical density along the sensing path levels
out after 12 l/min. This is interpreted in terms of the radial
temperature distribution of the gas flow. The effect of these
parameters on the correct measurement setup for flame sensing
and imaging at various geometries is briefly discussed.
I. INTRODUCTION
he motivation for this work is to develop methodology for
performing spectroscopy and imaging in the presence of
sooting. An example of such an environment is a high-
pressure flame, typical of airplane turbofan engines. We have
already demonstrated the feasibility of temperature sensing
inside such engines [1], as well as temperature and
concentration imaging by tomography from molecular
absorption of water vapor [2], therefore there is a lot to be
gained by applying tomography to such subjects. The work we
Manuscript received August 10, 2012. This work was supported by the UK
Engineering and Physical Sciences Research Council (EPSRC). Infrastructure
support from the Photon Science institute is gratefully acknowledged.
M. B. Saucedo, J. Young, P. Wright and K. B. Ozanyan are with the
School of Electrical and Electronic Engineering and the Photon Science
Institute, The University of Manchester, M13 9PL, UK. (email:
K.Ozanyan@Manchester.ac.UK,)
H. G. Darabkhani and Y. Zhang were with the School of Mechanical,
Aerospace and Civil Engineering, The University of Manchester, M13 9PL,
UK. Y. Zhang is currently with the Mechanical Engineering Department, The
University of Sheffield, UK( e-mail: YZ100@ Sheffield.ac.UK).M. Stringer
and R. E. Miles were with School of Electronic and Electrical Engineering,
The University of Leeds, UK.
report here paves the way to hard-field tomography at high
pressures, needed for engine design optimization.
We have recently demonstrated that soot particles are
transparent to THz radiation [3]. The necessary requirements
for performing hard-field tomographic imaging with THz
beams have also been set out [4] and hard-field THz
tomography imaging has been demonstrated on styrofoam
phantoms with refractive index similar to that in diffusion
flames. Therefore, the utilization of THz radiation is a
promising approach for addressing the challenges of sensing
and imaging in highly sooting environments.
In this work, we present line-of-sight (LOS) measurements
with broadband THz radiation through a diffusion flame, as a
sensing modality for imaging through sooty flames. This is
justified, since the path-integrals required for the Radon
transform in hard-field tomography involve a large number of
such LOS measurements – in principle they can be taken with
a single fixed THz-source/subject/THz-detector measurement
train and varying the sensing geometry by rotating and
translating the subject.
In addition to demonstrating measurements of single path-
integrals, we apply the physics of the THz emission from
biased-gap antennas to obtain more accurately and precisely
the THz pulse delay through the flame: by integrating the THz
electrical field waveforms recorded as a time-domain
correlation. This allows us to collect, at different air co-flows,
evidence on the dynamics of behavior of the hot-gas core of
the flame and the enveloping cooler-gas jacket. In addition to
its phenomenological value, this is essential for the definition
of well-behaved path-integrals for use in tomography
reconstruction of flames. The reported results imply that
imaging in pulse delay contrast is possible in the presence of
soot particles. This, together with previously published
tomography work, also indicates that soot alone will not be a
hindrance to achieving imaging contrasts of water vapor
concentration, temperature and pressure.
II. BACKGROUND
The paper reports on the investigation into heavily sooted
diffusion flames using THz spectroscopy at different co-flow
rates. A change in the combustion flow field, by varying fuel
978-1-4577-1767-3/12/$26.00 ©2012 IEEE

2. or air flow rates, results in the change of different aspects of
the flame properties, i.e. flame geometry, combustion stability,
soot emission and temperature field. Although much work has
been reported in the literature relating to the combustion of
fuel jets in still air or in parallel co-flowing streams [5,6], the
effect of air co-flow on the laminar non-lifted diffusion flame
properties is left almost unattended to. Darabkhani and Zhang
[5] reported that the flame dynamics and combustion
characteristics of a co-flow diffusion flame are strongly
affected by the co-flow air velocity. It was observed that when
the co-flow velocity reaches a certain value, the buoyancy-
driven flame oscillations may become completely suppressed.
That work demonstrated that the co-flow of air is able to push
the location of the instability initiation point beyond the
visible flame to create a very steady laminar flow region in the
reaction zone.
The resolution and sensitivity of THz imaging and
spectroscopy are still rather limited compared to existing
shorter wavelength optical and infrared diagnostics. At the
same time, the alternative diagnostic techniques such as laser
induced fluorescence (LIF), polarization spectroscopy and
Coherent anti-Stokes Raman Spectroscopy (CARS) [7] are
difficult, and sometimes near to impossible, to apply in
moderately to heavily sooted combustion environments, owing
to strong absorption, spectral interference from particulate
scattering, and fluorescence from large molecules. Therefore,
the case for applying THz-TDS of combustion species under
such conditions remains strong. As shown below, the major
combustion product with which the THz photons interact is
water vapor. To our knowledge, the observations in this work
are the first in literature on the effect of air co-flow velocities
on the transmission of THz radiation (0.2 – 2.5 THz) through
the flame zone.
III. EXPERIMENTAL SETUP
The context of the sensing task requires a dedicated burner,
capable of high pressures where soot often becomes a
problem, as well as allowing providing he views necessary for
tomographic imaging. The air co-flow burner used in this
study was designed to produce a classic Burke–Schumann [8]
laminar diffusion flame. This burner was specifically designed
and fabricated for the THz-TDS experiments and THz
tomography experiments in flames. The burner was capable of
pressures up to 30 bar and has eight 45 mm diameter
windows, as shown schematically in Figure 1, to enable THz
tomography based on 4 simultaneous views with no moving
parts. The window material is 20 mm thick high-resistivity
float-zone silicon (HRFZ-Si), which allows access for
wavelengths up to 1000 µm, and is commonly used for THz
measurements and spectroscopy [4]. The nitrogen port shown
in Figure 1 is used for pressurizing the chamber. It was
completely shut during the atmospheric pressure experiments.
The high pressure capability of the burner was not used to
obtain the results in this paper as the reported work is at
atmospheric pressure. Since the measurements are based on
simple LOS transmission, only two opposite windows were
used.
Gaseous methane (CH4) fuel was supplied from a
compressed gas cylinder regulated by a needle valve and
measured by a calibrated mass flow meter with 1% full scale
accuracy.
Co-flow air from a compressed dry air cylinder was
supplied into the burner and diffused using a layer of glass
beads, after which a honeycomb structure with 1.5 mm
diameter holes was used to straighten the flow. The air co-
flow was controlled by a needle valve to access a range of
mass flow rates from 1 to 20 l/min through a coaxial air exit
nozzle with a shroud diameter of 37.8 mm. The flame on a
fuel nozzle with an exit diameter of 4.6 mm was stationary
and stable under all co-flow conditions. Over the range of
examined flow conditions, the RMS flame tip flicker was less
Pressure Gauges
Filter
Cooling Block
Back-pressure Valve
View Port
Flow Metres
Optical Accesses
Figure 1. Schematic of the high-pressure burner facility. In
this experiment the burner was run at atmospheric pressure
and only one opposite pair of windows were used for line-of-
sight transmission measurements.
than 1% of the flame height
THz-TDS measurements on the methane diffusion flames
were taken at different air flow rates (5, 7.5, 10, 12.5 and 15
l/min) under a constant fuel co-flow rate of 200 ml/min. To
distinguish the effects of combustion (combustion products,
temperature gradients, etc.) from possible changes in optical
density and lensing effects in the cold methane/air co-flow
itself, two reference groups of data were collected from the
flows without a flame (henceforth referred to as cold flows):
a) Air flow only, with flow rates of 5, 7.5, 10, 12.5, 15 l/min
through the air jacket nozzle in the burner; b) Co-flow of air at

3. 5, 10, 15 l/min and methane at 200 ml/min through their Griot) with a 500 nm step accuracy. Each THz field scan has a
Gas
l/min
Rate
m3
/s
Flow
Rate
mg/s
Velocit
m/s
y
Fr No.
Methane 0.2 3.33E-06 2.27 0.230 61.2 1.249
Air 5 8.33E-05 100 0.077 171.27 0.019
Air 7.5 1.25E-04 150 0.115 256.91 0.043
Air 10 1.67E-04 200 0.153 342.55 0.076
Air 12.5 2.08E-04 250 0.192 428.18 0.119
Air 15 2.50E-04 300 0.230 513.82 0.171
respective nozzles in the burner, without ignition.
Table 1 shows the physical parameters of the fuel and air
streams for the reported experiments, including the reference
cases of cold co-flows. The mean fuel jet exit velocity is
determined by the flow rate and the nozzle cross-section 0.23
m/s with Reynolds number (Re) of 61. The air exit velocities
are in the range from 0.077m/s to 0.23 m/s with Re from 171
to 514. This substantiates the assumption that all flows were in
a laminar mode during all sets of experiments. The maximum
fuel jet Froude number (Fr), as a measure to compare inertia
and gravitational forces in the flow stream, is calculated to be
1.249 and the maximum Fr of the co-flow air is 0.171 at 15
l/min of air flow rate. Both Re and the Fr of the air streams
have been calculated based on the hydraulic diameter of the
annulus air port (DH= Do-Di=31.5 mm).
TABLE I
FUEL AND AIR PARAMETERS IN CO-FLOW AIR EXPERIMENTS
range of approximately 67 ps, sampled at 2048 points,
corresponding to a 10 μm step. To improve the signal-to-
noise, the antenna bias is modulated at 10 KHz and the
balanced detector output is processed by a lock-in amplifier
(EG&G 7265). The measurement system is integrated in a
LabView environment. The transmitted THz spectra are
obtained from the electrical field time domain waveform by
applying a Fast Fourier Transform (FFT) with appropriate
zero padding. This results in a useful bandwidth of 0.2-2.5
THz with a signal-to-noise ratio of 30 dB.
Volume Flow Mass
Re No.
We use a standard set-up for THz-TDS [4], shown
schematically in Figure 2. Ultra-short pulses of 80 fs duration,
spectrally centered at 800 nm, are delivered by a Coherent
Mira 900D Ti:Sapphire laser, pumped by a Verdi-V18
Nd:YVO4 laser. The total average power of 300 mW is split
90/10 between the pump and probe beams respectively. The pump
beam is focused on a biased photoconductive SI GaAs antenna
[9] to generate vertically polarized THz pulses. The optical
path of the THz beam is shown in Figure 2 as a thicker green
streak between “Photoconductive THz emitter” and “ZnTe
crystal”. The THz radiation is collected and collimated by a
gold-coated parabolic mirror into a 10 mm diameter beam,
which is then transmitted through the co-flow burner. The
transmitted THz beam is focused by a second gold-coated
parabolic mirror onto a 1 mm thick (110)-oriented ZnTe
crystal, where it is combined with the collinearly propagating
horizontally polarized and delayed probe beam, entering
through a hole in the second parabolic mirror. The THz-
induced changes in the polarization of the probe beam are
analyzed by a quarter-wave plate followed by a Wollaston
prism and recorded with a balanced detector (Nirvana 2017) as
a function of delay time, yielding the electrical field time
variation of the transmitted THz pulse. The time delay is
controlled by passing the probe beam through a hollow retro-
reflector mounted on a mechanical translation stage (Melles
Figure 2. Experimental setup for THz-TDS. Ultra-short
pulses are split in a pump beam to excite the THz emitter and
a probe beam with controlled delay, to allow time correlation
between the IR (probe) and transmitted THz pulses.
To reduce the effects of ambient water in the beam path,
two purging boxes were installed in the path of the THz beam,
as shown in Figure 2, just before and after the silicon
windows, and purged by a flow of dry air. Reference spectra,
with no fuel supply to the chamber, were taken for the 5 dry
air flows specified in Table 1.
IV. RESULTS AND DISCUSSION
The time domain signals recorded for methane diffusion
flames at different air-flow rates (see Table 1) are shown in
Figure 3 over a time window of 3 ps, shifted with 5.5 ps from
the origin. The no-flame reference measurements, in the
presence of an air-flow (5, 7.5, 10, 12.5 and 15 l/min) only and
an air-fuel cold co-flow (methane at 0.2 and air at 5, 10, 15
l/min), exhibited a negligible spread in the time at which the
signal crosses the zero and therefore are not shown.

4. 5, 10, 15 l/min and methane at 200 ml/min through their Griot) with a 500 nm step accuracy. Each THz field scan has a
Signal(V)
7.0E-02
5.0E-02
3.0E-02
1.0E-02
-1.0E-02
-3.0E-02
-5.0E-02
Ref
Meth200-air5 l/min M
eth200-air7.5 l/min Me
th200-air10 l/min Me
th200-air12.5 l/min Me
th200-air15 l/min
5.5 6.0 6.5 7.0 7.5 8.0 8.5
Time (ps)
effect of increasing the air flow rate, observed in Figure 4, is
unlikely to be caused by the methane-air co-flow per se. The
most likely cause is the radial gradient in the refractive index
due to the thermal gradients across the flame. Indeed, at higher
co-flows the diffusion rate increases, resulting in a shorter
optical path through hot gas; this causes variation in the
optical path length through hot and cold gas, and consequently
in the distortion of the beam as it traverses the flame. Thus the
decrease in the swing amplitude in Figure 3 can be ascribed to
the increased THz beam steer or divergence as it passes
through the flame at higher flow rates. Such beam “walk”
and/or changes in the filling of the detector aperture results in
a drop in transmission. Substantiation of this speculation can
be sought by acquiring additional evidence on the change in
the optical path through cold and hot gas, e.g. through
Figure 3. THz time domain signals of reference case and
methane diffusion flames at different air flow rates (5, 7.5, 10,
12.5 and 15 l/min
The waveforms in Figure 3 exhibit a clear trend towards
reduction of the swing amplitude between the negative and
positive peak, simultaneous with a time shift towards longer
delays, as the air co-flow is increased. For a more precise plot
of the estimated time delay, we applied additional processing
on the THz field waveforms, which was introduced and
explained in detail in earlier work [4]: the values for the pulse
delay for each value of the air flow were calculated using
parabolic interpolation on the time-integrated waveforms in
Figure 3. The result is plotted in Figure 4, showing a linear
relationship between the time delay of the THz pulse and the
co-flow values. After an initial linear increase, the optical
Figure 4. THz pulse delay as a function of the air flow.
The delay values are calculated from the maxima of the
integrated THz waveforms shown in Figure 3. The exact
values are shown as labels next to the experimental points.
density along the sensing path levels out above 12 l/min.
The power spectra (not shown) of the THz pulses
transmitted through the methane flame at different air flow
rates, normalized to their recorded reference spectra without a
flame (cold flow), confirm that the main interaction of the
incident THz radiation with the flame is due to water vapor, as
expected from the polar character of the H2O molecule.
schlieren imaging.
The observations in this work indicate substantial challenges
for hard-field THz tomography in flames: beam steering can
induce cross-talk in detector arrays, potentially compromising
the precision of the Radon transform. However, we have
recently demonstrated THz tomography in time-of-flight
contrast [4], which overcomes problems caused by the
“softening” of the probe field. Therefore, it is conceivable
that such tomography can be implemented with measurements
similar to those reported above.
REFERENCES
[1] Wood, M.P. and Ozanyan, K.B., Temperature mapping from molecular
absorption tomography, Proc. IEEE SENSORS 2011, pp. 865-869, 2011.
[2] York T., McCann H. and Ozanyan K.B., Agile Sensing Systems for
Tomography, IEEE Sensors J .2011, vol.11 pp. 2507
[3] Bassi, J., Naftaly, M., Miles, B., and Zhang, Y., The investigation of
sooty flames using terahertz waves. J. Flow. Meas. Instrum.,.vol. 16
pp. 341-345 (2005).
[4] Ozanyan, K.B.; Wright, P.; Stringer, M.R.; Miles, R.E. Hard-Field THz
Tomography, , IEEE Sensors J. 2011, vol.11, no.10, pp. 2507-2512,
2011.
[5] Gohari Darabkhani, H. and Zhang, Y. Suppression Dynamics of a
Laminar Oscillating Diffusion Flame with Co-flow Air. in the
Proceedings of the World Congress on Engineering (WCE) 2010, The
International Conference of Mechanical Engineering (ICME), 30 June -
2 July, 2010, Imperial College, London, UK,.
[6] Chen, Y.-C. and Bilger, R. W., Stabilization mechanisms of lifted
laminar flames in axisymmetric jet flows. Combust. Flame, 2000. 123(1-
2), p. 23-45.
[7] Tobai, J. and Dreier, T., Measurement of relaxation times of NH in
atmospheric pressure flames using picosecond pump-probe degenerate
four-wave mixing. Journal of Molecular Structure, 1999. 480-481, p.
307-310.
[8] Burke, S. P. and Schumann, T. E. W., Diffusion flames. Proc. Combust.
Inst., 1948. 1(2), p. 2-11. (2011).Tomography Shu, Z., Aggarwal, S. K.,
Katta, V. R., and Puri, I. K., Flame-vortex dynamics in an inverse
partially premixed combustor: The Froude number effects. Combust.
Flame, 1997. 111(4), p. 276-286.
[9] Nahata A., Weling A. S., and Heinz T. F., A wideband coherent
terahertz spectroscopy system using optical rectification and electro-
optic sampling. Applied Physics Letters, , 1996. 69(16), p. 2321-2323.