Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
In-situ-Detection of Potassium Atoms in High-Temperature CoalCombustion Systems using Near-Infrared-Diode Lasers
E. Schlosser, T. Fernholz, H. Teichert, V. Ebert
Institute of Physical Chemistry, University of Heidelberg,
Im Neuenheimer Feld 253, 69120 Heidelberg, Germany
Tel:.+49-6221-54-5004
Fax: +49-6221-54-5050,
E-mail: [email protected]
Abstract: Direct tunable diode laser absorption spectroscopy at 769.9 nm and 767.5 nm was used to
measure potassium atom concentrations in situ in the high temperature (up to 1650 K) flue gas of two
different pulverized coal dust combustion systems (atmospheric or pressurized (12 bar)). Two laser types
(Fabry-Pérot (FP) and vertical-cavity surface-emitting lasers (VCSEL)) were used for the spectrometer
and characterized with respect to the magnitude and linearity of their static and dynamic wavelength
tuning properties. The wide continuous current-induced tuning range of the VCSEL of 20 cm-1 (compared
to 1 cm-1 for the FP) make this laser ideal for species monitoring in high pressure processes. Two
VCSELs were time-multiplexed to realize the simultaneous detection of the potassium D1 and D2 lines.
Several oxygen absorption lines in the A-band, which are in close spectral vicinity of the potassium lines
were detected simultaneously, showing the possibility of multi-species detection with one laser.
Using the FP-DL for the atmospheric process and the VCSEL for the high pressure process, the pressuredependent coefficients for spectral broadening as well as a shift of the potassium line in the flue gas were
determined to be (0.18 ± 0.01) cm-1/atm and (-0.060 ± 0.003) cm-1/atm (at 1540 K and 11.2 bar). The total
width and shift of the D1 line (11.2 bar/1540 K) were 60 GHz and –20 GHz respectively.
The potassium atom concentration was determined continuously for several days in both plants under
various operation conditions. Typical concentrations in the atmospheric plant - referenced to standard
conditions throughout the entire paper - were around 2 µg m-3 at STP with a range of 50 ng m-3 at STP to
30 µg m-3 at STP. Averaging 100 scans for each concentration value, we achieved a time resolution of
1.7 s and a detection limit of 10 ng m-3, which corresponds to a fractional absorption in the 10-3 to 10-4
range. A strong anti-correlation with the oxygen concentration could be verified. At the 12 bar plant the
concentration was again typically around 2 µg m-3 but potassium levels up to 60 µg m-3 were observed.
Here, a strong dependence of the K-signal on the type of fuel could be verified.
Keywords:
in situ detection, potassium, alkali corrosion, direct tunable diode laser absorption spectroscopy,
VCSEL, FP-DL, flue gas, pulverized coal combustion
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
1/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
1 Introduction
Coal will continue to be an important primary energy source worldwide. In order to minimize CO2-emission the
efficiency of the energy conversion needs to be increased. Modern combined-cycle power plants yield
efficiencies of up to 58 % [1] by a combined use of traditional steam turbines and additional flue gas turbines
that work at temperatures of up to 1650 K. Due to the stringent requirements on the purity of the flue gas this
technique is currently used only in combination with natural gas combustion. However, there is a strong interest
to transfer this technique to pressurized coal combustion systems. But in order to achieve long service lifetimes
of the expensive flue gas turbines it is necessary to efficiently remove fly ash and soot particles generated by the
combustion of the pulverized coal. Alkali metal (sodium and potassium) compounds that evaporate due to the
high temperatures are of concern because they promote severe corrosion of the turbine blades even in minute
concentrations above 24 ppb (weight) [2]. Gas turbine manufacturers set the limit for the alkali vapor
concentration in flue gas to 6.5 ppb (weight) [3]. To meet the flue gas quality requirements, a completely new
technology of ultra high efficiency flue gas cleaning systems is being developed, which is supposed to work
under high temperature (>1500 K) and high pressure conditions (>10 bar), i.e. without any major temperature
reduction or pressure loss in the flue gas. One important task in building coal-fired combined-cycle power plants
is to find an industrial sensor to monitor the effectiveness of the hot gas filter. A fast detection system would
allow the flue gas turbine to be protected, e.g. by bypassing the flue gas in case of a filter malfunction or by
active control of the combustion process.
Presently off-line, wet chemical, time-integrated analysis of the residues collected via extractive sampling
techniques using Atomic Absorption Spectroscopy is the only commercial technique for the detection of alkalimetals. Faster techniques to detect alkali on-line are being explored, e.g. Excimer Laser Induced Fragmentation
Fluorescence (ELIF), Surface Ionisation (SI), and Plasma Excited Alkali Resonance Line Spectroscopy
(PEARLS). ELIF detects gaseous alkali species, while SI and PEARLS additionally detect alkali on aerosol
particles [4]. But SI and PEARLS are to a certain extent extractive methods because they require the transfer of a
hot gas sample into the detector, which makes them susceptible to the typical problems of extractive techniques,
like adsorption or chemical changes in the sample during the extraction process. ELIF on the other hand is purely
optical and measures in situ in the flue gas duct [5]. But it relies on a relatively bulky and costly excimer laser.
Therefore there is a need for an in situ alkali sensor, which is fast, robust, compact, and cost-efficient. These
properties are offered by NIR-diode-laser based in situ species monitors, which have been demonstrated for
many combustion species [6]. However, due to the lack of suitable diode lasers, no such sensor seems feasible
for alkali compounds. Instead we propose to detect atomic potassium (K) which is generated through the thermal
dissociation of the alkali. Diode lasers have been used to detect atomic potassium in commercial laboratory
atomic absorption spectrometers with higher sensitivity than with hollow cathode lamps [7]. However, no in situ
detection in industrial combustion processes has been reported. Since a fast increase in alkali concentration will
be associated with a sharp increase in K atom density, this signal may serve as a fast indicator for a filter
malfunction.
We developed an absorption spectrometer for fast in situ monitoring of K atoms which offers high specificity,
sensitivity and temporal resolution. It is based on inexpensive diode lasers and accesses the potassium D lines.
The specifications of the device and first measurements of the absolute potassium density in two coal-fired
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
2/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
combustion systems, one at atmospheric pressure and one at up to 16 bar, are presented. The sensor is based on a
new diode laser type, a vertical–cavity surface-emitting laser (VCSEL), which offers the ultra wide tuning
properties that are needed to cover the strongly pressure broadened potassium line. The tuning properties of such
a laser have been determined. In addition, we report the first simultaneous in situ detection of both potassium
D-lines and of potassium and oxygen. Process parameters influencing the release of the alkali metal atoms (the
oxygen-concentration and the temperature of the flue gas) are discussed.
2 Experimental
2.1 Direct Absorption Measurement Principle
The measurement principle of absorption spectroscopy is based upon detection of the wavelength-specific
intensity loss of monochromatic light propagating through the measurement volume containing a number density
N of absorbers. The measured laser intensity I(λ) after passage through an absorbing medium of the thickness z
is described by Lambert-Beer’s law:
(1)
I (λ , T ) = I 0 ( λ ) ⋅ e
−σ ( λ ,T ) ⋅ N ⋅ z
where I0(λ) is the initial intensity and σ(λ,Τ) is the wavelength and temperature dependent absorption cross
section. Spectrally integrating σ over all wavelengths yields the integrated line strength S(T). Determination of
S(T) instead of the peak absorption coefficient by a spectrally resolved measurement has the advantage of being
insensitive to line broadening (e.g. the line width changes mainly with pressure but also with composition and
temperature).
Continuously tunable diode lasers are versatile light sources to realize this spectral integration. By varying the
laser temperature, the emission wavelength can be coarsely tuned to the position of the absorption line. In
addition, the laser can be finely tuned across the absorption line by a fast (kHz) periodic modulation of the
injection current as shown in Figure 1. At the same time, this current modulation leads to an often undesired
intensity modulation I0(λ) of the laser light.
One powerful application of tunable diode laser absorption spectroscopy (TDLAS) is to determine absolute
species concentrations in situ, that is, directly in the probe volume, completely avoiding any gas sampling, which
is a common source of error. This technique was used to demonstrate the simultaneous multi-species detection
(CH4, CO2, H2O, O2) and optical temperature measurement by two-line absorption techniques in power plants
and waste incinerators [6]. It is also possible to detect potassium (K) atoms via strong electronic dipole
transitions: the well known D1 (4s2S1/2 → 4p2P1/2 ) and D2- line (4s2S1/2 → 4p2P3/2) at 769.90 nm and 766.49 nm.
A common problem for in situ absorption measurements are the numerous, strong and uncontrollable
disturbances of the optical signal, which are often 100 to 1000 times larger than the molecular absorption signal.
They are caused by a quickly fluctuating optical obscuration (Tr(t)) of the measurement path and by emission of
broadband background radiation (E(t)). Taking these disturbances into account, equation 1 can be rewritten:
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
3/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
(2)
I (λ , T ) = I 0 (λ ) ⋅ e
− S (T )⋅ g ( λ − λ0 )⋅ N ⋅ z
⋅ Tr (t ) + E (t )
∞
, with
∫ g (λ )d λ = 1
−∞
with g(λ) being the normed form function of an absorption line, centered at wavelength λ0. The time-dependent
transmission Tr(t) of the measurement path is mainly generated by scattering or absorption by soot or dust. In
addition, refractive index gradients, mechanical vibrations, and alignment shifts reduce the amount of laser light
that reaches the detector, while E(t) caused by gray body radiation of particles and the furnace walls can increase
the detector signal. Artificial or daylight, as well as the dark current of the photodiode add to E(t), too. All these
effects must be removed before applying Beer’s law.
Various measures have to be taken to minimize the effect of the disturbances. Rigid mounts for lasers and
detectors are needed, which are preferably connected directly to the walls of the process vessel in order to ensure
a permanent alignment despite frequent variations of the vessel temperature. Background radiation can be
minimized by the use of narrow-band optical filters. The ability to separate the various disturbances and to
determine their magnitude is most important in order to correct absorption signals and in order to make the
absolute level of the concentration signal virtually insensitive to the degree of optical obscuration (Tr(t)) or
emission of background radiation(E(t)).
Figure 1: Measurement principle of direct tunable diode laser spectroscopy. Disturbances (Tr(t), E(t)) have to be
compensated before applying Beer’s Law. Tr(t) indicates the unspecific loss of the initial laser light (I0) due to scattering
and refraction, E(t) is the level of background light and gray body radiation reaching the detector. The integrated specific
absorption is proportional to the number density of absorbers and to the intensity of the transmitted light.
If Tr(t) can be assumed to be constant during the scan, it is possible to correct for the transmission variations
during the acquisition of an absorption profile by dividing the raw scan by the baseline signal. This assumption
can be assured by means of the high-speed (kHz) wavelength modulation capabilities of the diode laser, which
allow a modulation frequency faster than typical transmission fluctuations. In addition, the amplitude modulation
can be used to analyze and to correct for transmission changes as well as to remove the remainders of the
emission which are not removed completely by the narrow-band filters.
2.2 Diode Laser Characterization
Two types of diode lasers were used for the detection of potassium via the D1/D2 absorption lines: A FabryPérot-type AlGaAs-laser (FP-DL) and a vertical-cavity surface-emitting laser (VCSEL, Centre Suisse
d’Électronique et de Microtechnique (CSEM)). FP-DLs are conventional edge emitting diode lasers. The FP-DL
used by us (Mitsubishi-ML4405) was initially developed for the first laser pointers. Because of the old
manufacturing process wavelengths up to 770 nm can be selected, even though the target wavelength was
750 nm. FP-DL have the serious disadvantage of discontinuous, staircase-like wavelength tuning (Figure 2).
These so-called “mode hops” make it necessary to characterize and to select the diode laser carefully with
respect to a coincidence of a high quality, single-mode emission behavior and coverage of the desired absorption
line. Static wavelength tuning properties were measured using a wavemeter (Burleigh WA-1000). Figure 2
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
4/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
shows two regimes of the wavelength tuning (time resolution 1 s) of a FP-DL tuned by a linear current ramp.
Continuous linear tuning is achieved at low modulation amplitudes (t=100 s-1000 s, ∆I=4 mA). At higher
modulation depths (t>1100 s, ∆I=16 mA) mode hops occur, so that the linear tuning sections are interrupted by
tuning gaps of about 0.26 nm. It has to be noted that the occurrence of a mode hop can depend on the tuning
direction. In our case there are two mode hops for the up-scan and only one large mode hop for the down-scan,
skipping one resonator mode. The inset shows the wavelength versus the drive current for the same data. Such a
”mode map” is used to check whether a diode laser can be used for species detection within a certain spectral
interval.
Figure 2: Quasi-static current-induced wavelength tuning of a FP-DL (ML4405) showing tuning discontinuities or “mode
hops”. The inset shows the same wavelength data versus the injection current.
The FP-DL used in our experiments to detect the D2 line at 766.49 nm shows a threshold of 43.5 mA and a
amplitude modulation coefficient of (0.26 ± 0.01) mW/mA both measured at 312.5 K. The static current and
temperature-tuning coefficients measured in a mode hop-free region are (–0.24 ± 0.01) cm-1/mA (at 296 K) and
(–1.01 ± 0.01) cm-1/K. A mode hop of this laser is typically 5 cm-1 wide. By changing the laser temperature from
280 K to 305 K, the wavelength is varied by 108 cm-1 (at 50.4 mA), however less than a quarter of this spectral
interval can be accessed because of mode hops. The static, large-scale temperature-tuning coefficient, including
all mode hops, is (–3.37 ± 0.05) cm-1/K. The typical continuous tuning range is ~ 1.9 cm-1. Limited by the
maximum allowed current modulation and the reduced dynamic tuning, the maximum continuous tuning range
at 1 kHz modulation is only 1 cm-1. Even though FP-DLs are relatively inexpensive and can successfully be used
for TDLAS, the need to find a suitable laser is time-consuming because many diodes have mode hops at the
desired wavelength. Out of a batch of 60 FP-DLs (ML4405) 28 lasers emitted between 764 nm and 770 nm at
room temperature. Only 12 lasers emitted light near 769.9 nm and 15 near 766.5 nm choosing different operating
temperatures and currents. Just three (5%) of them (1 for the D1 and 2 for the D2 line of potassium) had a
suitable mode map that allowed a mode hop-free, single-mode scan of the potassium absorption line profile.
In principle there is the possibility to use other laser structures with better tuning properties. For example an
external resonator (XC-DL) can be used to force the diode to emit at the target wavelength, however its tuning
speed is greatly decreased due to the necessary mechanical movement of the feedback mirror [8]. DFB-DL
(distributed feedback DL) are frequently used for the detection of oxygen at 760 nm [5, 9, 10, 11, 12] but they
are presently not available for wavelengths around 770 nm.
However a more recently developed laser type, the VCSEL, is an excellent alternative even though the output
power is low (100 µW-1 mW) compared to FP-DLs (3 mW-20 mW). VCSELs have been used recently for the
sensitive detection of oxygen at atmospheric pressures exploiting their small specific amplitude modulation [13].
They also offer an extremely wide current-induced, single-mode tuning range which has been exploited under
laboratory conditions for the detection of oxygen at pressures of up to 10.9 bar [12].
VCSELs have a very thin gain region and a very short resonator length. They are pumped along the resonator
axis by an annular ring contact. The ring contact, with an inner diameter between 2 µm and 20 µm forms the
emitting aperture. This leads to a circular, relatively weakly diverging (~12°) beam which is free of astigmatism.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
5/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
While the beam profile of the FP-DL using an AR-coated, molded aspheric lens (NA=0.55, f=4.51mm) is
elliptical (2 × 6 mm²), it is circular and 3 mm in diameter for the VCSEL.
The emitted laser light of our VCSELs is polarized with a power-dependent polarization contrast (power (p-pol)
divided by power (s-pol)) of up to 120 [13], which is important for minimizing etalon noise in the signal. The
polarization dependent emission spectrum was investigated with a calcite Glan-Thomson prism and a wavemeter
that was modified in order to obtain the emission spectrum [13]. A side mode could be detected at higher output
power, which was located 0.23 cm-1 from the main mode and which had a polarization orientation perpendicular
to the main mode. This property can be used to suppress the side modes using polarization filters and extend the
usable single mode tuning range.
The VCSEL used for the detection of the D2 line has a very low threshold current (5.6 mA, 298 K) compared to
the FP-DLs and a single mode output power of 262 µW (7.0 mA, 298 K). The amplitude modulation coefficient
is highly non-linear with a value of ~ (0.20 ± 0.01) mW/mA for low currents. It is zero at about 8mA and
becomes negative above 9 mA. The static current and temperature-tuning coefficients are (6.2 ± 0.3) cm-1/mA (at
298 K) and (0.96 ± 0.05) cm-1/K (at 7mA), respectively. By changing the current between 5.5 mA and 9.5 mA
the wavelength can be tuned by 26.4 cm-1. The same laser can be tuned by a change in temperature from 280 K
to 305 K by 24 cm-1.
For the line fitting process the deviation from linear tuning needs to be small or it has to be compensated by
calibrating the tuning rate. Therefore the dynamic tuning rate and its deviation from linearity was characterized
with emphasis on the specific setting used for our in situ measurements. This was done using a 10 cm air-spaced
etalon with a free spectral range of 0.05cm-1.
Figure 3: Dynamic wavelength tuning of a VCSEL measured at 1.3 kHz modulation frequency using a 10 cm air spaced
etalon. The insets show the etalon structure for the same time interval at the beginning and at the end of the sweep.
Figure 4: Deviation from linear tuning of a 770 nm-VCSEL at 1.3 kHz. Only 14 cm-1 of the tuning range were evaluated,
since the laser is tuned from below threshold at the beginning of the graph.
The laser (D1, 770 nm) was modulated at 1.3 kHz from below threshold. Figure 3 shows the etalon structure. Its
contrast is intensified with increasing the laser power. The inset shows the etalon structure at the beginning and
at the end of the ramp for equal time intervals. The laser clearly tunes faster at the end of the sweep. The
following procedure was used to extract the time dependent tuning behavior: The detector signal is normalized
by a second order polynomial which was fitted to the raw signal. The zero crossings of the normalized signal
with negative slope were then detected and indexed by a LABVIEW program. The dynamic tuning curve of the
laser can be drawn (Figure 4) by plotting the time at each zero crossing versus the fringe index number
multiplied by the free spectral range of the etalon (0.05 cm-1). Finally, a first and a second order polynomial was
fitted to the tuning curve. The difference between fit and experiment is shown in the lower part of Figure 4. A
slightly non-linear tuning can be seen, which is well described by the second order fit. The maximum deviation
from linearity was ± 0.3 cm-1 for a 14 cm-1 section of the total tuning range. The dynamic current tuning
coefficient of the 770 nm-VCSEL (Figure 3, Figure 4) is rising for one scan from 5.0 cm-1/mA to 6.8 cm-1/mA
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
6/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
within the 1.3 kHz current ramp for a 2.9 mA peak-to-peak saw-tooth shaped current modulation. The 767 nmVCSEL shows for the same settings a dynamic current tuning coefficient from 5.4 cm-1/mA to 5.7 cm-1/mA. The
static tuning coefficient for the latter is 6.2 cm-1/mA in comparison.
This data will be used to correct the influence of non-linear tuning on the absolute absorber concentration or to
determine the maximum error which would be caused by ignoring the non-linearity.
3 In situ Measurements of Potassium
3.1 Setup
In situ measurements of potassium were performed at two types of pulverized-coal-fired (PCF) plants: one
operating at atmospheric pressure with a total power of 250 kW and one operating at 10 bar to 16 bar with a total
power of 1 MW. The flue gas channel was optically accessible downstream from the hot gas filter via two
pressurized, air-purged windows. The measurement location was chosen upstream from
the first heat
exchangers in order to ensure a high temperature and a sufficient level of thermally dissociated potassium. Laser,
optics, and detector were attached to two small steel platforms that were directly connected to the walls of the
flue gas channel. The alignment was occasionally corrected to compensate for distortions of the channel walls.
Figure 5: Left: Schematics of the 250 kW-atmospheric PCF combustor with flue gas filter showing the sampling location of
the in situ laser spectrometer for the detection of atomic potassium. Inset: Cross-section of the flue gas duct and location of
the laser measurement path.
The absorption path across the hot flue gas (300 mm at the atmospheric and 140 mm at the high pressure plant,
respectively) was short compared to the total absorption path (970 mm / 1760mm) within the high pressure
access flanges through the insulation of the vessel. By varying the purging air supply and by substituting the
purging air by nitrogen, it was confirmed that the absorption path length is well defined and remains unchanged
by gas turbulence.
The frequency and amplitude behavior of the background emission and the transmission of the in situ path was
analyzed and used to adapt the data processing scheme. The amount of background light reaching the detector
was in the order of 3 µW at 1 bar, while the variable thermal emission was in most cases below 225 µW, which
is 16 % of the average FP-DL output power (1.45 mW). The frequency spectrum of emission and transmission
(measured over 0.2 µs to 2 s time intervals) showed only minor disturbances above 1 kHz. At 12 bar the thermal
emission reaching the detector was usually 13% of the average laser power (~130 µW), but was occasionally 5.4
times larger than the laser (Figure 8).
The combined laser/peltier-element driver (Profile, LDC 1000) provided a low-noise (<0.4 µA rms., 10 Hz –
6 MHz) current source for the laser and the laser temperature control loop (∆T=10 mK). The laser current was
modulated via a function generator (SRS DS345) using a symmetric triangular modulation. A repetition rate of
1.302 kHz was chosen to tune the laser across the absorption line. To allow the quasi simultaneous detection of
two independent absorption features, e.g. both potassium D-lines or potassium and oxygen, a time-multiplexed
setup with two lasers was realized. Here both laser beams were superimposed and directed onto one single
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
7/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
detector. The first laser was linearly tuned across the absorption line within 383µs and then its injection current
was kept at a constant level just below the threshold while the sequence was repeated for the other laser. This
saw-tooth-shaped modulation signal was generated by two programmable function generators which were
synchronized and triggered by a pulse generator. The two modulation signals were interleaved and phase shifted
in a way that a complete saw-tooth signal is observed at the detector. The repetition frequency of a complete
scan with both lasers was 1.302 kHz. The laser beams were directed through the probe volume onto a silicon
photodiode using a focusing mirror. A low-pressure reference cell with pure potassium vapor was either placed
within the probe beam or 5% of the probe beam were split off and the reference signal was detected by a separate
detector. The photocurrents were preamplified (Femto DLPCA 100) and digitized at 1.0 Msample/s by an eightchannel, simultaneously sampling 12-bit A/D converter on a PCI-board (T112-8, Imtec). Usually 100
consecutive scans were averaged for further noise reduction. A background polynomial of second order and a
Lorentzian line shape function was fitted to the scans. Based on this fit extensive corrections where made to
remove disturbances by transmission and emission changes. Finally the area underneath the absorption line was
determined and used to calculate the absolute concentration values, assuming the ideal gas law. Further
parameters of this calculation were the operating parameters of the power plant (pressure and temperature), the
tuning properties of the laser and the K line strength, which was calculated using CRC [14] data.
3.2 Results and Discussion
3.2.1 250 kW Atmospheric Pulverized-Coal-Fired (PCF) Power Plant
The in situ spectrometer was used under industrial conditions to detect K atoms in the flue gas channel of an
atmospheric 250 kW-coal combustor (Figure 5). A FP-DL offered a sufficiently wide continuous tuning range
(1 cm-1) to measure the pressure broadened potassium line as shown in Figure 6. Wider tuning would have been
beneficial, but was limited by mode hops. The D1-K absorption line at 1 atm and 1170 K showed a full width at
half maximum (FWHM) of (5.2 ± 0.5) GHz (0.17 cm-1). The reference absorption line from the low pressure Kvapor cell had a width of (0.90 ± 0.05) GHz. This value includes the hyperfine structure splitting of the D1 line.
The expected Doppler broadening FWHM is 0.773 GHz (300 K). Additionally, the atmospheric D1 line in the
flue gas showed a red shift of the line center of (–1.2 ± 0.1) GHz (0.04 cm-1).
Figure 6: D1-K-absorption line measured in situ in the flue gas at 1 atm and 1170 K, and in a low pressure reference cell at
room temperature using a FP-DL. Mode hops limit the tuning range to about 1 cm-1.
K number densities found under typical combustion conditions were 8·109 atoms cm-3 corresponding to 2 µg m-3
(number densities -in units of atoms cm-3- being stated under high temperature process conditions while
concentrations -in units of g m-3- are always referenced to standard temperature and pressure conditions, STP).
We achieved a detection limit of about 4·107 atoms/cm3 (or 10 ng m-3, resp. 7.5 ppt (weight)). The temporal
evolution of the K concentration over a time period of 8 hours is depicted in Figure 7 in combination with the
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
8/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
oxygen concentration in the flue gas which was measured by an extractive, paramagnetic sensor. A, B and C
correspond to time periods where the combustor was co-fired with oil. The amount of oil which was used with
respect to period A was twofold during B and fourfold during C. A close anti-correlation between the oxygen
and the K-signal is observed. During severely O2-depleted and thus chemically reducing situations, we found
that the K-concentration can increase within very short time by a factor of more than 100 and even lead to
complete absorption of the laser light. This K-O2-correlation was expected since potassium is readily oxidized,
and the equilibrium concentration of atomic potassium is reduced by the presence of O2. The oxygen dependence
could be corrected if the equilibrium constant and its temperature behavior were known and a precise oxygen
signal with the same temporal response as the laser signal would be available. Such a simultaneous measurement
of species concentration and temperature has been demonstrated for other species within a gas-fired power plant
using TDLAS [6]. A setup for the simultaneous detection of oxygen and potassium is currently under
investigation. First results are presented later in this paper.
The K signal is expected to show further correlations with other combustion parameters. It is expected that the K
signal increases significantly: a) with rising flue gas temperatures due to stronger dissociation, b) with higher
alkali content of the coal, and c) with smaller concentrations of other possible reaction partners in the flue gas
(H2O, CO2, SOx).
Figure 7: In situ detection of the concentration of atomic potassium (squares, top) in the flue gas duct of a 250 kW-coalcombustor by a laser in situ spectrometer. O2-concentration (line, below) as determined by an extractive, paramagnetic O2sensor. e1: Combustion stopped and restarted, e2: shortly no averages, e3: burner stopped.
3.2.2 1000 kW Pressurized Pulverized-Coal-Fired Power Plant
We investigated the possibility of an in situ detection of atomic potassium in a high pressure PCF-plant at
pressures above 10 bar. Above atmospheric pressure, FP-DLs or DFBs do not provide sufficient wavelength
tuning depth and are best replaced by VCSELs. The spectrometer access for probing the flue gas of the 1000 kW
coal combustor was located downstream from the liquid ash separator. The absorption path length was 140 mm.
Both potassium absorption lines, D1 and D2, were probed quasi simultaneously by time multiplexing two
VCSEL lasers as shown in Figure 8. The 767 nm-VCSEL (right half of saw-tooth profile) had a strong nonlinear
amplitude modulation (L-I-curve) which distorted the background severely. During normal operation the
absorption was weak and the background radiation caused by gray body radiation was much smaller than the
laser light intensity (lower graph). When the fuel supply was changed, the oxygen concentration was depleted
and high concentrations of potassium atoms were generated. This led to nearly 100% absorption of the laser light
at the line center. At the same time, the background emission caused by glowing soot particles is significantly
increased (twice as large as the laser intensity, upper graph). Since the drive current dropped for each laser and
each scan below the threshold, non-laser light can easily be subtracted before the final evaluation.
Figure 8: Quasi-simultaneous time-multiplexed in situ detection of both potassium D-lines with two independent VCSELs
(D1 line left trace, D2 right trace and twice as strong). The lower graph depicts typical absorption depth under standard
conditions. The upper graph shows a high K-concentration event with significant background radiation.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
9/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
The D2 line is twice as strong as the D1 line. However, oxygen absorption lines due to the A-Band
(b1Σ+g←X3Σ-g, P21Q20, P21P21) are located in the immediate spectral vicinity of the D2 line center. These
additional lines have to be included in the fitting procedure for mid-range and low K concentrations. The
fractional absorption of these O2-lines is in the order of 10-3. O2 is probed partially outside the pressurized vessel
(ambient air), in the low-temperature pressurized air in the access pipes, and in the hot flue gas. The observed
profile is therefore the result of two different line shapes (1 bar, 12 bar) and two different line center positions.
This complicated situation drastically increases the number of fitting parameters needed for the O2-lines, so that
the sensitivity and accuracy will be reduced. The oxygen lines (P31P31, P33Q32, and P33P33) which are visible
in the profile of the D1 line are shown in Figure 9. They are substantially weaker and much better separated from
the K line, so that they can be handled much easier. For mid-range concentrations they can even be ignored (in
Figure 9 they show up in the difference).
Because of the wide spectral scan the relative position of the line within the scan was very stable and hardly
needed any correction. The minimum detectable fractional absorption was of the order of 10-4 for the O2
absorption lines and 10-3 for heavily broadened K- lines.
Figure 9: Potassium D1-absorption lines measured in situ in a pressurized PCF-plant at 11.2 bar and 1540 K and in a room
temperature reference cell. Three rotational resolved O2-absorption lines (P31P31, P33Q32, and P33P33) are visible in the
profile.
The D1 line showed at 11.2 bar and 1540 K a pressure broadening of (60 ± 4) GHz (FWHM) and a red shift of
the line center of (–20 ± 1) GHz. The pressure-induced broadening coefficient is (0.18 ± 0.01) cm-1/atm and the
line shift coefficient is (-0.060 ± 0.003) cm-1/atm in flue gas at 1540 K. The wavelength tuning range of the
VCSEL shown in this graph was 13 cm-1. The air broadening of the O2(P33Q32) line at 1540 K and 11.2 bar was
calculated using the HITRAN database to be 0.287 cm-1 (FWHM). In comparison, the broadening of the K(D1)
in flue gas was measured to be 2.00 cm-1 (FWHM) under those conditions. Potassium therefore shows
approximately a seven times larger pressure broadening than O2. This may be a consequence of the higher
reactivity of the potassium atoms.
The flue gases were probed continuously for 64 hours. Taking 100-sweep-averages a time resolution of 1.7 s was
achieved. Figure 10 shows the potassium concentration measured during the start-up procedure of the plant in
combination with the coal feed and the flue gas temperature. The power plant was heated up with oil as fuel that
naturally has a low alkali content. Then the operation was switched to pulverized coal as fuel as soon as the
combustion was stable and the temperature was sufficiently high. During the oil-fired start-up procedure of the
plant, the concentration was only 20 ng m-³. Under coal-fired conditions, the K-concentration rose from less
than 0.1 µg m-³ to up to 1.5 µg m-³. This was expected since the alkali was set free from the coal during
combustion. The potassium concentrations typically observed were around 1 µg m-³. The noise is reduced
significantly at 12:00 h by improving the alignment of the laser.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
10/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
Figure 10: In situ K-concentration measured in the flue gas duct downstream the filter of a high pressure PCF-plant.
Temperature and coal feed are shown in order to mark the changes during the transition between oil- and coal-fired
combustion.
The small K-concentration peaks visible in Figure 10 were caused by variations in the flue gas temperature and
in the O2-concentration. A strong K-concentration peak is generated during the fuel change from oil to coal.
Such peaks can be generated for short periods of less than 10 s by temporally insufficient air supply, which leads
to reducing conditions of the flue gas. The potassium concentration momentarily reaches 60 µg m-³ and then falls
back to normal. The maximum K-concentration during standard operation conditions was usually below 2 µgm-³.
4 Conclusion
Direct tunable diode laser absorption spectroscopy is a versatile tool for fast, non intrusive in situ gas analysis,
especially for reactive species like potassium.
We employed FP-DL and, for the first time, VCSELs under industrial conditions to measure absolute
concentrations of potassium atoms in situ in the flue gas of two coal combustion systems. Very wide wavelength
tuning was needed because of the severe broadening of the potassium line in the flue gas. Due to their limited
tuning range and complicated step-like tuning behavior, FP-DLs were only suitable for the atmospheric
combustion process. These lasers had to be selected from a batch of more than 60 lasers. The yield for a laser
suitable for the D1 or D2 line was only 5%. For the high-pressure process with up to 16 bar fast current tuning
with continuous tuning ranges of 10 to 20 cm-1 was needed to recover the K line, which is only possible with
VCSELs. For both laser types we investigated magnitude and linearity of the static temperature and current
tuning and of the dynamic wavelength tuning at tuning frequencies in the kHz range. The VCSEL proved to be
superior to the FP-DL especially in current tuning width (up to 26cm-1) and in linearity (± 0.3 cm-1 for a 14 cm-1
sweep at 1.3 kHz).
Both potassium lines, at 769.9 nm (D1) and 767.5 nm (D2), and several rotationally resolved oxygen lines of the
A-band in the spectral vicinity were investigated in situ at operating pressures of 1 bar and up to 16 bar. Despite
the higher absorption coefficient of the D2 line, the D1 has to be preferred because of less interference by
oxygen lines. The pressure-induced broadening and shift of the K line was determined in the flue gas and
resulted at 1540 K and 11.2 bar in broadening coefficients of (0.18 ± 0.01) cm-1/atm and line shifts of
(-0.060 ± 0.003) cm-1/atm. The full-width of the D1 line at 11.2 bar and 1540 K was 60 GHz and the shift was
-20 GHz. As far as we know, no experimental or theoretical data for the line width or the shift of potassium D1
or D2 lines in flue gas are available in literature. Experimental data for broadening and shift by N2 (and noble
gases) at 410K are: Half-width at half maximum divided by density (γ/n) is (1.30±0.02)·10-20 cm-1/cm-3 and the
shift by density (β/n) is (-9.71±0.21)·10-21 cm-1/cm-3 for the D1-line[13]. We found for the broadening at 1170 K
and 1 bar γ/n=(1.36±0.14)·10-20 cm-1/cm-3 and at 1540 K and 11.2 bar γ/n=(1.89±0.10)·10-20 cm-1/cm-3. We found
for
the
shift
at
1170 K
11.2 bar β/n=(-1.26±0.10)·10
-20
and
-1
1 bar
β/n=(-6.38±0.64)·10-21 cm-1/cm-3
and
at
1540 K
and
-3
cm /cm .
The potassium atom concentration was determined continuously for several days (up to 64 h) under industrial
conditions in an atmospheric PCF-plant and a 12 bar PCF-plant under various operation conditions. Typical
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
11/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
concentrations in the atmospheric plant were around 2 µg m-3, with a range of 50 ng m-3 to 30 µg m-3. A strong
anti-correlation with the oxygen concentration could be verified. At the 12 bar-plant the concentration was again
typically around 2 µg m-3, but up to 60 µg m-3 were observed. Here, a strong dependence of the K-signal on the
type of fuel could be verified. Averaging 100 scans for each concentration value, we achieved a time resolution
of 1.7 s and a detection limit of 10 ng m-3, which corresponds to a fractional absorption in the 10-3 to 10-4 range.
In conclusion, we could show that potassium can be detected in situ with high sensitivity. Because of the high
reactivity of potassium atoms, it would be very difficult to provide similar properties with extractive techniques.
Therefore the spectrometer should be suitable for the surveillance of filter malfunctions. Further measurements
to verify this function are underway in cooperation with our industrial partners.
Most interesting is our ability to extend the spectrometer to the detection of multiple species (6). Especially for a
(indirect) measurement of alkali we would need to measure in situ potassium, the temperature and the major
reaction partners (e.g. O2, H2O, CO2) of potassium. First measurements to determine two species with close lying
absorption lines quasi-simultaneously have been demonstrated in this paper for the case of potassium and
oxygen by time-multiplexing two VCSELs. Other species, which are spectrally better separated can be “added”
to the spectrometer with a dichroic beamsplitter-setup.
5 Literature
[1]
B. Hillebrand, RWI-Papiere 47: 1-47 (1997)
[2]
L. Scandrett, and R. Clift, J. Inst. Energy 57: 391-397 (1984)
[3]
E. Schlosser, T. Fernholz, H. Pitz, W. Christmann, and V. Ebert, Proceedings of the 10th International IUPAC
Conference on High Temperature Materials Chemistry, 10.-14. April 2000, Forschungszentrum Jülich, Germany,
Schriften des Forschungszentrums Jülich, Reihe Energietechnik /Energy Technology, Vol.15, Part I, 27-30 (2000)
[4]
V. Häyrinen, R. Hernberg, R.Oikari, U.A. Gottwald, P.B. Monkhouse, K.O. Davifson, B.Lönn, K. Engvall, J.B.C.
Petterson, P.Lehtonen, and R. Kuivalainen, 6th International Conference on Circulating Fluidized Beds, Würzburg, 873,
(1999)
[5]
U. Gottwald, and P. Monkhouse, Appl. Phys. B 69, 151-154, (1999)
[6]
V. Ebert, T. Fernholz, C. Giesemann, H. Pitz, H. Teichert, J. Wolfrum, and H. Jaritz, Proc. Comb. Inst. 28, 423-430,
(2000); and
V. Ebert, J. Fitzer, I. Gerstenberg, K.-U.Pleban, H. Pitz, J. Wolfrum, M. Jochem, and J. Martin, Proc. Comb. Inst. 27,
1301-1308, (1998)
[7]
K. Niemax, H. Groll, and C. Schnürer-Patschan, Spectrochimica Acta Rev. 15(5): 349-377 (1993)
[8]
D. Wandt, M. Laschek, K. Przyklenk, A. Tünnermann, and H. Welling, Opt. Commun. 130: 81-84 (1996)
[9]
N.A. Morris, J.C. Connolly, R.U. Martinelli, J.H. Abels, and A.L. Cook, IEEE Phot. Tech. Lett 7(5): 455-457 ( 1995)
[10] C. Corsi, M. Gabrysch, and M. Inguscio, Opt. Comm. 128: 35-40 (1996)
[11] V. Weldon, J. O'Gorman, J. J. Perez-Camacho, and J. Hegarty, Electron. Lett. 32: 219-221 (1996).
[12] J. Wang, S.T. Sanders, J.B. Jeffries, and R.K. Hanson, Appl. Phys. B 72: 865-872 (2001)
[13] P. Vogel, and V. Ebert, Appl. Phys. B 72: 127-135 (2001)
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
12/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
[14] D.R. Lide (Editor), „Handbook of chemistry and Physics“, CRC BOCA RATON , 75rd Ed. (1995)
[15] N. Lwin, and D.G. McCartan, J.Phys.B 11(22): 3841-3847 (1995)
Acknowledgement:
The valuable support of Prof. Dr. J. Wolfrum, University of Heidelberg is acknowledged. The authors would
also like to thank Mr. Bartmann (E.ON Engineering), Dr. Christmann (E.ON Engineering) and Dr. Förster (Saar
Energie) for their assistance during the measurements. The studies of E.S. were kindly sponsored by the State of
Baden Württemberg (Landesgraduiertenförderungsgesetz). V.E. would like to acknowledge the Heidelberger
Akademie der Wissenschaften for their financial support.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
13/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
Figures
Figure 1: Measurement principle of direct tunable diode laser spectroscopy. Disturbances (Tr(t), E(t)) have to be
compensated before applying Beer’s Law. Tr(t) indicates the unspecific loss of the initial laser light (I0) due to
scattering and refraction, E(t) is the level of background light and gray body radiation reaching the detector. The
integrated specific absorption is proportional to the number density of absorbers and to the intensity of the
transmitted light.
Figure 2: Quasi-static current-induced wavelength tuning of a FP-DL (ML4405) showing tuning discontinuities
or “mode hops”. The inset shows the same wavelength data versus the injection current.
Figure 3: Dynamic wavelength tuning of a VCSEL measured at 1.3 kHz modulation frequency using a 10 cm air
spaced etalon. The inset shows the etalon structures for the same time interval at the beginning and at the end of
the sweep.
Figure 4: Deviation from linear tuning of a 770 nm-VCSEL at 1.3 kHz. Only 14 cm-1 of the tuning range were
evaluated, since the laser is tuned from below threshold at the beginning of the graph
Figure 5: Left: Schematics of the 250 kW-atmospheric PCF combustor with flue gas filter showing the sampling
location of the in situ laser spectrometer for the detection of atomic potassium. Inset: Cross-section of the flue
gas duct and location of the laser measurement path.
Figure 6: D1-K-absorption line measured in situ in the flue gas at 1 atm and 1170 K, and in a low pressure
reference cell at room temperature using a FP-DL. Mode hops limit the tuning range to about 1 cm-1.
Figure 7: In situ detection of the concentration of atomic potassium (squares, top) in the flue gas duct of a
250 kW-coal-combustor by a laser in situ spectrometer. O2-concentration (line, below) as determined by an
extractive, paramagnetic O2-sensor. e1: Combustion stopped and restarted, e2: shortly no averages, e3: burner
stopped.
Figure 8: Quasi-simultaneous time-multiplexed in situ detection of both potassium D-lines with two independent
VCSELs (D1 line left trace, D2 right trace and twice as strong). The lower graph depicts typical absorption depth
under standard conditions. The upper graph shows a high K-concentration event with significant background
radiation.
Figure 9: Potassium D1-absorption lines measured in situ in a pressurized PCF-plant at 11.2 bar and 1540 K and
in a room temperature reference cell. Three rotational resolved O2-absorption lines (P31P31, P33Q32, and
P33P33) are visible in the profile.
Figure 10: In situ K-concentration measured in the flue gas duct downstream the filter of a high pressure PCFplant. Temperature and coal feed are shown in order to mark the changes during the transition between oil- and
coal-fired combustion.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
14/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
Initial Laser Intensity
I0(t,λ)
Detector Signal (V)
Tr(t0)
0
e-αmol (λ)
Raw Signal
Tr(t1)
E(t) Offset
time
λ decreasing
λ increasing
Figure 11: Measurement principle of direct tunable diode laser absorption spectroscopy. Disturbances (Tr(t),
E(t)) have to be compensated before applying Beer’s Law. Tr(t) indicates the unspecific loss of the initial laser
light (I0) due to scattering and refraction, E(t) is the level of background light and gray body radiation reaching
the detector. The integrated specific absorption is proportional to the number density of absorbers and to the
intensity of the transmitted light.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
15/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
768.4
768.4
768.2
λ (nm)
768.2
λ (nm)
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
Mode Map
mode hops
768.0
768.0
767.8
Current (mA)
767.8
767.6
46
48
50
54
52
56
767.6
0
200
400
600
800
Time (s)
1000
1200
1400
1600
Figure 12: Quasi-static current-induced wavelength tuning of a FP-DL (ML4405) showing tuning discontinuities
or “mode hops”. The inset shows the same wavelength data versus the injection current.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
16/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
0
-2
Time(ms)
1.100 1.102 1.104 1.106 1.108
-3.0
-3 -0.5
-4
-5
-3.5
-1.0
1.350 1.352 1.354 1.356 1.358
1.05
1.10
1.15
Detector Signal (V)
Detector Signal (V)
-1
1.20
1.25
Time (ms)
1.30
1.35
Figure 13: Dynamic wavelength tuning of a VCSEL measured at 1.3 kHz modulation frequency using a 10 cm
air spaced etalon. The inset shows the etalon structures for the same time interval at the beginning and at the end
of the sweep.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
17/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
Figure 14: Deviation from linear tuning of a 770 nm-VCSEL at 1.3 kHz . Only 14 cm-1 of the tuning range were
evaluated, since the laser is tuned from below threshold at the beginning of the graph.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
18/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
Pulverized
Coal
Detector
Si-PD
Sending
Unit
NIR-DL
flue
gas
Combustion
Chamber
300mm
970mm
TDLAS
Filter
Slag
Flue Gas
Temperature
Measurement
Extractive Alkali
Measurement
Bypass
Figure 15: Left: Schematics of the 250 kW-atmospheric PCF combustor with flue gas filter showing the sampling location
of the in situ laser spectrometer for the detection of atomic potassium. Inset: Cross-section of the flue gas duct and location of
the laser measurement path.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
19/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
-1
Transmission (%)
12989.2
100.0
12989.0
Wavenumber (cm )
12988.8
12988.6
12988.4
12988.2
Mode Hops
99.0
FWHM = 0.90 GHz
FWHM = 5.2 GHz
98.0
Background
Scan (Reference)
Scan (in situ)
Fit (in situ)
97.0
96.0
400
300
Line Shift
200
100
Samples
0
Figure 16: D1-K-absorption line measured in situ in the flue gas at 1 atm and 1170 K, and in a low pressure
reference cell at room temperature using a FP-DL. Mode hops limit the tuning range to about 1 cm-1.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
20/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
3.0
e1
e2
e3
O2 (vol%)
15
K (µg/m³)
2.0
10
1.0
5
A
B
0.0
02:10
03:10
C
5l/h oil
2.5l/h oil
04:10
05:10
10l/h oil
06:10
07:10
08:10
09:10
0
10:10
Figure 17: In situ detection of the concentration of atomic potassium (squares, top) in the flue gas duct of a
250 kW-coal-combustor by a laser in situ spectrometer. O2-concentration (line, below) as determined by an
extractive, paramagnetic O2-sensor. e1: Combustion stopped and restarted, e2: shortly no averages, e3: burner
stopped.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
21/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
3
Photo Detector Signal (V)
D1
2
D2
540% Emission, 100% Absorption
13% Emission, 3.7% Absorption
1
0
-400
0
samples (pixel)
400
Figure 18: Quasi-simultaneous time-multiplexed in situ detection of both potassium D-lines with two
independent VCSELs (D1 line left trace, D2 right trace and twice as strong). The lower graph depicts typical
absorption depth under standard conditions. The upper graph shows a high K-concentration event with
significant background radiation.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
22/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
99.5
99.0
98.5
12993
Transmission (%)
100.0
wavelength (cm-1)
12990
12987
12984
12981
O2(P33Q32)
O2(P33P33)
O2(P31P31)
FWHM
Scan (in situ)
Fit (in situ)
Scan (Reference)
Difference (abs.%)
98.0
97.5
0.2
0.1
0.0
line
shift
300
200
100
Figure 19: Potassium D1-absorption lines measured in situ in a pressurized PCF-plant at 11.2 bar and 1540 K
and in a room temperature reference cell. Three rotational resolved O2-absorption lines (P31P31, P33Q32, and
P33P33) are visible in the profile.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
23/24
30.04.02
Presented at 3rd International Conference on Tunable Diode Laser Spectroscopy,
0.5
0.0
00:00
Temperature (K)
T
K
C
06:00
12:00
Time (hh:mm)
1600
1550
1500
Coal Feed (kg/h)
1.0
Potassium Concentration
(µg/m³)
July 8-12, 2001 Zermatt, Switzerland, accepted for publication in Spectrochimica Acta 2002,(in print)
18:00
1450
100
50
0
24:00
1400
1350
Figure 20: K-concentration measured in the flue gas duct downstream the filter of a high pressure PCF-plant.
Temperature and coal feed are shown in order to mark the changes during the transition between oil- and coalfired combustion.
2001-07-Spectrochimica Acta-Zermatt_TXT+PCS final revised acc proofs.doc
24/24
30.04.02