Vol. 126 (2014) ACTA PHYSICA POLONICA A No.

Passive Open-Path FTIR Measurements

and Spectral Interpretations for in situ Gas Monitoring
and Process Diagnostics
*
S. Ci¦szczyk
Institute of Electronics and Information Technology, Lublin University of Technology,
Nadbystrzycka 38A, 20-618 Lublin, Poland
(Received May 7, 2014; in nal form July 2, 2014)
This article aims to present the possibilities for the passive spectral measurements and qualitative interpre-
tations of inside-the-object gases. The basic features and applications of open-path Fourier transform infrared
spectroscopy for measuring atmospheric and process exhaust gases are discussed. Methods for modelling and in-
terpreting the spectra used for in situ measurements are presented. As an example, a spectrum simulation was
investigated using CO2 , a gas present in substantial quantities in various exhaust gases. The spectrum varied sig-
nicantly with temperature allowing its use for diagnostics purposes. Thus, examples of measurements made using
passive transmission with a heated object as a radiation source are presented. A baseline correction procedure was
necessary to obtain a transmission (or absorbance) spectrum. The gas content and temperature were determined
based on these measurements.

DOI: 10.12693/APhysPolA.126.673

PACS: 07.07.Df, 07.57.Ty, 33.20.Ea, 07.05.Tp

1. Introduction the ow. Changes in the rotational line intensity can

mirror the state of ow [10]. Due to its high absorption,
Optical methods are widely applied for gas measure- the CO2 spectral band between 2200 and 2400 cm−1 is
ments in various industrial elds and environmental mon- used to estimate the temperature. The temperature pro-
itoring [1]. With the aid of special equipment, one can le of the spectrometer line of sight is measured using a
process diagnostic information at a reaction site [24]. spectral remote-sensing method [11, 12]. However, this
Under harsh industrial environments, spectral measure- method can introduce a strong absorption from atmo-
ments can provide valuable information. However, de- spheric CO2 causing signicant errors. Thus, the spectral
spite their vast potential, in situ spectroscopic measure- band near 2270 cm−1 , in which this eect is negligible,
ments are not commonly used largely due to the dicul- is applied [13]. The measurement of CO2 using a spec-
ties in obtaining information from the measured spectra. trally narrow quantum cascade detector with a Fourier
As with many indirect measurements a mathematical re- transform spectrometer is also proposed [14].
lationship between the desired process parameters and Fourier transform infrared (FTIR) spectroscopy is
the directly measured values must be found [5, 6]. widely used to measure various gases, often in the anal-
During pyrolysis and combustion, the gas content and ysis of gases derived from the thermal degradation of
temperature must be monitored. The predominant gases biomass [15, 16]. This technique is also used to measure
in these processes are carbon dioxide and water vapour. the scattering by particles, such as water vapour [17].
Gases such as CO, CH4 , NOx and others may also be Light extinction of particles can signicantly impact the
present in lower concentrations. Carbon dioxide is the absorption measurements of the medium [18, 19]. The
main greenhouse gas that causes climate change. Car- primary advantage to this method is its ability for si-
bon dioxide is often measured using a non-dispersive in- multaneous qualitative and quantitative analyses of mul-
frared (NDIR) technique, which can be applied to mon- tiple components. Open-path FTIR (OP-FTIR) spec-
itor stack emissions [7] or the air quality inside build- troscopy measures the transmission path in the atmo-
ings. One problem with such measurements is the de- sphere. The integrated concentration along the length
pendence of the sensor parameters on the temperature; of the measured path is then quantied. The spectro-
however, attempts have been made to alleviate this is- meter is characteristically independent from the radia-
sue [8]. New NDIR sensor systems can monitor CO2 con- tion source. There is no sampling system separate from
tent over a broad range from the ppm scale to 20 vol.% the test object and no need to store, dry or cool the sam-
for a 32 mm optical path length [9]. Spectral measure- ples. The active version utilises an additional radiation
ments of carbon dioxide can also be used to characterise source.
In the passive version, the radiation derives from nat-
ural sources, such as the sun or hot gases. However,
* e-mail: s.cieszczyk@pollub.pl this case requires the use of an emission or absorption
spectrum. If the studied gas layer is located between

(673)
674 S. Ci¦szczyk

the layers or an element of higher temperature, then an lar method for analysing open-path spectra is the par-
absorption spectrum is obtained. If the measured layer tial least-squares (PLS) method [31]. Signicant non-
has a temperature higher than the ambient temperature, linearity and thus Beer's law derogating can occur if the
then an emission spectrum will be observed. The pas- absorbance the spectral lines exceeds 0.5 [32]. This situ-
sive method is less sensitive than the active method, par- ation limits the possibility of using linear CLS and PLS
ticularly when the temperature in the gas layer of in- methods or forces the use of weaker spectral lines. For
terest is near the ambient temperature. The advantage quantitative analysis of the spectra, a neural network can
to these approaches, in comparison with active open- be used as a non-linear method [33].
path absorption measurements, especially under indus-
trial conditions, is that they do not require adjustments
to the source or spectrometer. Exemplar measurements
for assessing re gases are obtained at 180 ◦C [20]. This
method also works well for analysing the thermal decom-
position of sewage sludge [21]. An interesting practical
application of the passive technique is the quantication
of exhaust gases from aircraft engines [22]. OP-FTIR
is also frequently used to measure the gaseous emissions
from the combustion of biomass [23].
In industry, emission spectroscopy is typically used to
measure the radiation spectra of ames. This technique
is clearly most commonly associated with temperature
measurements. However, attempts have been made to
determine the contents of both soot and a homogeneous
layer containing water vapour and CO2 gases [24]. This Fig. 1. CO2 transmittance at 300 K (innite resolu-
method is attractive primarily because it is capable of tion).

remote sensing; for example, it can be used in environ-

ments such as volcanoes where other methods become
dicult [25].
The semi-active method introduces a radiation source
whose size corresponds to that of the spectrometer eld of
view [26]. This method does not require accurate align-
ment between the source and the spectrometer. The ra-
diation source has a lower temperature than that in the
classic active system; however, its temperature exceeds
that of the passive system.

2. Modelling of gas spectra

Calibration involves building a model between the

spectrum and the content of the specic components.
Classical calibration requires the measurement of high- Fig. 2. CO2 transmittance changes due to the concen-
quality calibration samples with an appropriate concen- tration path length at 300 K (1 cm
−1
resolution).
tration range adapted to the specic application. Special
equipment, such as a ventilation test chamber, may be In the open-path measurements, however, the domi-
used for this purpose [27]. During the calibration process nant approach uses a forward model, by which the spec-
one must maintain a constant temperature, pressure and trum can be determined based on the temperature, pres-
pressure uniformity along the gas cell length. sure, and gas content. To determine content, one must
The classical least-squares (CLS) technique can be solve the inverse problem, which typically uses a non-
used to precisely measure atmospheric gases. Spectral linear least-squares method. Because a mathematical so-
ranges are selected such that the absorbance does not lution to the inverse of the forward model is impossible,
exceed 0,4. For CO, this range is 20202200 cm−1 ; an iterative process is applied. The synthetic data from
for 12 CO2 , it is 20202306 cm−1 ; and for 13 CO2 , it is direct modelling are compared with the measured data
20202290 cm−1 [28]. To continuously monitor atmo- (observations) until achieving an assumed t. Similar
spheric gases the range of 21502320 cm−1 for CO2 , CO mathematical approaches for inverse problems are used
and N2 O retrieval is used [29]. The CLS method can in various scientic elds. The solution method is of-
be applied only over small ranges of the concentration ten independent of the physical problem under consider-
path length product because the concentration can be ation. Many inversion methods are available for which
underestimated at higher values [30]. Another popu- comparisons can be found in the literature [3436]. A
 Passive Open-Path FTIR Measurements and Spectral Interpretations. . . 675

forward model should be veried for known measurement

conditions. Open-path FTIR spectrometers are primar-
ily used to measure atmospheric gases. Spectra obtained
at temperatures that exceed those encountered under at-
mospheric conditions are used to verify the spectra calcu-
lated via simulation [38]. Experimental results are rarely
presented in the literature. To determine the absorp-
tion coecients of gases at high temperatures, a special
cuvette is required. Accurate CO2 measurements at tem-
peratures from 300 to 1550 K conrm the validity of the
spectral model created using the line-by-line method and
narrow-band databases, especially up to 600 K [39]. Sim-
ilar high-resolution measurements have been made for in-
dustrial applications at temperatures up to 1773 K [40].
Fig. 3. Inuence of temperature on transmittance for
−1
The spectral measurements at elevated temperatures also
10000 ppm m CO2 (5 cm resolution).
demonstrate good accuracy (up to 3% dierence in the
transmissivity) for dierent spectral models with CO2
and H2 O used either in the spectroscopy or combustion
processes for radiative transfer modelling [4143]. The
temperature can be determined from the absorption spec-
trum using the ratio of two absorption lines provided that
optically thin conditions are utilised [44]. For example,
the 20202100 cm−1 spectral range is used to determine
of CO and CO2 [45]. Simulations of issues at hand can be
performed to correctly analyse the measurement data [46]
and the impact of various factors such as the tempera-
ture on the retrieval process [47]. For this purpose, the
HITRAN database is utilised [48]. Figures 14 provide
the CO2 transmissivity spectra for atmospheric concen-
trations and temperatures of exhaust gases created by
dierent processes. The gas spectrum exhibits a strong
Fig. 4. Inuence of temperature on transmittance of
−1
characteristic absorption band at 23002400 cm−1 . In
100000 ppm m CO2 (5 cm resolution).
this range, the transmittance is zero for concentrations
above 1000 ppm m. The 22002300 cm−1 range is sen-
sitive to temperature changes depending on the concen-
direct model must take into account all gas and instru- tration. As depicted in Fig. 5 temperature inuence on
mental parameters because it can introduce errors, which the CO spectrum is also signicant.
can be signicant if the instrument is misaligned [37].
Hence, the most important factor for interpreting the 3. Measurements and their analyses
measurements is the creation of an appropriate direct
model and an iterative procedure. The accuracy of the The main drawback to passive FTIR is the need for
baseline correction. Algorithms can automatically per-
form this procedure; however, most require operator in-
tervention. Manual correction to on-line measurements
is dicult or even impossible. To perform an unattended
operation, an automatic correction algorithm such as a
wavelet transform is necessary [49]. The most common
method utilises a polynomial held by the points of the
spectrum without rotational line absorption [50], which
implies the need for measurements with relatively high
resolutions of at least 1 cm−1 . At lower resolutions, the
individual spectral lines may overlap making it dicult
to nd a point on the spectra with no absorption. Vis-
ibility of the individual rotational lines must be visible
for compiling the background shape. Another method
for determining the background involves the removal of
all components from the spectra via iteration [51].
Fig. 5. Inuence of temperature on transmittance of Measurements were made using a Bruker OPAG 33
−1
7000 ppm m CO (1 cm resolution). spectrometer with a resolution of 1 cm−1 . The measure-
676 S. Ci¦szczyk

ment site was used to study the processes of biomass

gasication and combustion. All measurements were ob-
tained using the passive technique. The gas absorption
features are shown in the spectrum because they were
located between the hot regions of the structure and the
spectrometer (Fig. 6). A typical spectrum is presented
in Fig. 7. To retrieve the gas content using such mea-
surements, the baseline spectra must rst be estimated.
A typical spectrum obtained after such an operation is
presented in Fig. 8. A comparison between the atmo-
spheric CO2 and the gases from biomass gasication in
an 8 m path length is provided in Fig. 9. As indicated,
the absorption of CO2 in the 23002400 −1 spectral range
is characteristically high. Signicant CO absorption and
several lines representing water vapour are also visible. Fig. 8. Measured spectra after baseline correction.

Fig. 6. Setup involving an FTIR spectrometer and

biomass gasication of ue gases.

Fig. 9. Comparison of the measured spectra with cold

atmospheric CO2 and warm, high-path length CO2 and
CO concentrations.

due to interference from the CO. The inversion procedure

yielded a temperature of 360 K and a CO2 concentration
of 21500 ppm m.
Figure 12 provides a comparison of the measured and
calculated spectra at 1 cm−1 resolution (without smooth-
ing). The large values in the dierence (residual) spec-

Fig. 7. Measured single-beam spectra for CO2 and

CO.

Figure 10 provides the spectrum of CO2 measured at

a 1 cm−1 resolution and smoothed by moving the aver-
age lter. Because the measured concentration and path
length for atmospheric CO2 is low, absorption occurs in
the range of 23202370 cm−1 . Figure 10 clearly demon-
strates the alignment between the measured spectrum
and that derived from the direct model. This gure also
indicates the correct baseline estimation obtained using a
polynomial. The estimated value of the CO2 concentra-
tion is 950 ppm m (for 300 K). After smoothing the rota-
tional line is no longer visible, but ltering aids in increas-
ing the quality of the match between the measured and Fig. 10. Comparison between the measured and calcu-
calculated spectra for low signal-to-noise passive spectra. lated spectra from the forward model after an iterative
The large mist between 2200 and 2250 cm−1 in Fig. 11 is tting for atmospheric CO2 .
 Passive Open-Path FTIR Measurements and Spectral Interpretations. . . 677

Fig. 11. Comparison between the measured and calcu- Fig. 13. Comparis on between the measured and cal-
lated spectra from the forward model after an iterative culated rotational lines spectra from the forward model
tting for warm CO2 derived from biomass gasication. after an iterative tting for warm CO derived from
biomass gasication.

correction for background spectra can improve the out-

come of the algorithm for determining gas content. The
correctness of the baseline may be evaluated using the
dierence between the measured and calculated spectra,
which is particularly obvious in spectra containing evenly
spaced rotational lines such as that of CO. The residual
spectrum has high-frequency components, which after in-
tegration within the bands approach or reach zero. These
high-frequency components in the dierence spectra can
be interpreted as errors that arise from the simplicity of
the ILS utilised. Therefore, a correct determination of
the baseline requires low values of low-frequency compo-
Fig. 12. Comparison of the measured and calculated
nents. The simulations, measurements and their com-
spectra using the forward model after an iterative tting parison demonstrate the possibility of determining the
for warm CO derived from biomass gasication. content and temperature of gases via passive FTIR ab-
sorption with heated construction elements serving as a
radiation source.
trum are related to the water vapour absorption lines.
Figure 12 clearly shows four such lines. Using detailed
References
analysis for each CO rotational line (Fig. 13), a good t
can be obtained between the measurements and calcula-
tions, although the direct model only takes into account [1] J. Hodgkinson, R.P. Tatam, Meas. Sci. Technol. 24,
012004 (2013).
the shape of the ILS associated with the window applied
to the interferogram. No ILS correction resulted from the [2] J. Bak, S. Clausen, Meas. Sci. Technol. 13, 150
(2002).
eld of view of the spectrometer. The example utilised
an estimated path length of 7100 ppm m at 360 K. [3] S. Rego-Barcena, R. Saari, R. Mani, S. El-Batroukh,
M.J. Thomson, Meas. Sci. Technol. 18, 3479 (2007).

[4] S. Rego-Barcena, R. Mani, R. Saari, M. J. Thomson,

4. Conclusions Metall. Mater. Trans. B 40, 158 (2009).

[5] J. Mroczka, D. Szczuczy«ski, Metrol. Meas. Syst. 16,

The temperature of gases measured in industrial pro- 333 (2009).
cesses using in situ methods can vary to a greater extent [6] D. Szczuczy«ski, J. Mroczka, Opt. Appl. 39, 521
than those measured using conventional atmospheric OP- (2009).
FTIR. Therefore, three spectral windows wider than that [7] Y.W.Sun, C. Liu, K.L.Chan, P.H. Xie, W.Q. Liu,
found for atmospheric gas retrieval are needed. If, how- Y. Zeng, S.M. Wang, S.H. Huang, J. Chen,
ever, a wider spectral window is utilised to analyse pas- Y.P. Wang, F.Q. Si, Atm. Measur. Techn. 6, 1993
sive gas absorption spectra, a proper background spectra (2013).
must be selected. The use of conventional approxima- [8] J. Park, H. Cho, S. Yi, Procedia Eng. 5, 303 (2010).
tion via the polynomial method does not always produce [9] J. Hodgkinson, R. Smith, W.O. Ho, J.R. Sael,
the desired results. An iterative procedure that includes R.P. Tatam, Sens. Act. B 186, 580 (2013).
678 S. Ci¦szczyk

[10] W. Wong, Z. Yang, A. Muriel, ISRN Spectr. 2013, [32] L. Shao, P.R. Griths, P.M. Chu, T.W. Vetter, Appl.
681654 (2013). Spectrosc. 60, 254 (2006).

[11] H. K. Kim, T. H. Song, J. Quant. Spectr. Ra- [33] S. Ci¦szczyk, Metrol. Meas. Syst. 2, 287 (2013).
diat. Transf. 93, 369 (2005). [34] J. Mroczka, D. Szczuczy«ski, Appl. Opt. 49, 4591
[12] T. Song, Heat Transf. Eng. 29, 417 (2008). (2010).

[13] H. Grosch, A. Fateev, K.L. Nielsen, S. Clausen, [35] J. Mroczka, D. Szczuczy«ski, Appl. Opt. 51, 1715
J. Quant. Spectr. Radiat. Transf. 130, 392 (2013). (2012).

[14] D. Hofstetter, J. Di Francesco, L. Hvozdara, [36] J. Mroczka, D. Szczuczy«ski, J. Quant. Spectr. Ra-
H. P. Herzig, M. Beck, Appl. Phys. B 103, 967 diat. Transf. 129, 48 (2013).
(2011). [37] M.D. Coleman, T. Gardiner, Vibrat. Spectrosc. 51,
[15] A.J. de Castro, A.M. Lerma, F. Lopez, M. Gui- 177 (2009).
jarro, C. Diez, C. Hernando, J. Madrigal, Infrared [38] T. Fleckl, H. Jager, I. Obernberger, J. Phys. D:
Phys. Technol. 51, 21 (2007). Appl. Phys. 35, 3138 (2002).
[16] A.V. Sepman, L.P.H. Goey, Biomass Bioen. 35, 2903 [39] M.F. Modest, S.P. Bharadwaj, J. Quant. Spectr.
(2011). Radiat. Transf. 73, 329 (2002).
[17] R.M. Hashmonay, M.G. Yost, Env. Sci. Techn. 33, [40] V. Evseev, A. Fateev, S. Clausen, J. Quant. Spectr.
1141 (1999). Radiat. Transf. 113, 222 (2012).
[18] M. Czerwi«ski, J. Mroczka, T. Girasole, G. Gouesbet, [41] V. Becher, S. Clausen, A. Fateev, H. Splietho, Int.
G. Grehan, Appl. Opt. 40, 1514 (2001). J. Greenhouse Gas Contr. 5S, S76 (2011).
[19] M. Czerwi«ski, J. Mroczka, T. Girasole, G. Gouesbet, [42] V. Becher, S. Clausen, A. Fateev, H. Splietho, Int.
G. Grehan Appl. Opt. 40, 1525 (2001). J. Greenhouse Gas Contr. 5S, S66 (2011).
[20] A. Stec, P. Fardell, P. Blomqvist, L. Bustamante- [43] V. Becher, A. Goanta, H. Splietho, Int. J. Green-
Valencia, L. Saragoza, E. Guillaume, Fire Safety J. house Gas Contr. 11, 34 (2012).
46, 225 (2011).
[44] G.P. Jellison, D.P. Miller, Opt. Eng. 45, 016201
[21] J.H. Ferrase, S. Chavez, P. Arlabosse, N. Dupuy, (2006).
Thermochim. Acta 404, 97 (2003).
[45] M.J. Wooster, P.H. Freeborn, S. Archibald, C. Op-
[22] K. Schafer, C. Jahn, P. Sturm, B. Lechner, M. Bacher, penheimer, G. J. Roberts, T.E.L. Smith, N. Goven-
Atm. Env. 37, 5261 (2003). der, M. Burton, I. Palumbo, Atm. Chem. Phys. 11,
[23] I.R. Burling, R.J. Yokelson, D.W.T. Grith, 11591 (2011).
T.J. Johnson, P. Veres, J.M. Roberts, C. Warneke, [46] J.Mroczka, Measurement 46, 2896 (2013).
S.P. Urbanski, J. Reardon, D.R. Weise, W.M. Hao,
J. Gouw, Atm. Chem. Phys. 10, 11115 (2010).
[47] S. Ci¦szczyk, Metrol. Meas. Syst. 3, 513 (2013).

[48] L.S. Rothman, I.E. Gordon, Y. Babikov, A. Barbe,

[24] J. Lim, Y. Sivathanu, J. Ji, J. Gore, Combust. Flame
D. Chris Benner, P.F. Bernath, M. Birk, L. Bizzocchi,
137, 222 (2004).
V. Boudon, L.R. Brown, A. Campargue, K. Chance,
[25] C. Oppenheimer, P. Bani, J.A. Calkins, M. R. Burton, E.A. Cohen, L.H. Coudert, V.M. Devid, B.J. Drouin,
G. M. Sawyer, Appl. Phys. B 85, 453 (2006). A. Faytl, J.-M. Flaud, R.R. Gamachem, J.J. Harrison,
[26] T.J. Johnson, B.A. Roberts, J.F. Kelly, Appl. Opt. J.-M. Hartmann, C. Hill, J.T. Hodges, D. Jacque-
43(3), 638 (2004). mart, A. Jolly, J. Lamouroux, R.J. Le Roy, G. Li,

[27] J. Raisanen, R. Niemela, Ann. Occup. Hyg. 46, 501 D.A. Long, O.M. Lyulin, C.J. Mackie, S.T. Massies,

(2002). S. Mikhailenko, H.S.P. Müllert, O.V. Naumenko,

A.V. Nikitin, J. Orphal, V. Perevalov, A. Perrin,
[28] M.B. Esler, D.W.T. Grith, S.R. Wilson, L.P. Steele,
E.R. Polovtseva, C. Richard, J. Quant. Spectr. Ra-
Analyt. Chem. 72, 206 (2000).
diat. Transf. 130, 4 (2013).
[29] S. Hammer, D.W.T. Grith, G. Konrad, S. Vardag,
[49] L. Shao, P.R. Griths, Environm. Sci. Technol. 41,
C. Caldow, I. Levin, Atm. Measur. Techn. 6, 1153
7054 (2007).
(2013).
[50] L. Zhi-Ming, L. Wen-Qing, G. Ming-Guang. T. Jing-
[30] T.E.L. Smith, M.J. Wooster, M. Tattaris,
Jing, Z. Tian-Shu, X. Liang, W. Xiu-Li, Chin. Phys.
D.W.T. Grith, Atm. Measur. Techn. 4, 97
B 17, 4184 (2008).
(2011).
[51] D. Hong, S. Cho, Appl. Spectrosc. 57, 299 (2003).
[31] D.L. Bjorneberg, A.B. Leytem, D.T. Westermann,
P.R. Griths, L. Shao, M.J. Pollard, Trans. ASABE
52, 1749 (2009).