PHOTONIC SENSORS / Vol. 6, No.

2, 2016: 127‒131

High Sensitivity Gas Sensor Based on IR Spectroscopy

Technology and Application
Hengyi LI
Department of Electronic Information Engineering, Jincheng College of Sichuan University, 611731, Chengdu, China
*
Corresponding author: Hengyi LI E-mail: [email protected]

Abstract: Due to extremely effective advantages of the quantum cascade laser spectroscopy and
technology for trace gas detection, this paper presents spectroscopy scanning, the characteristics of
temperature tuning, system resolution, sensitivity, and system stability with the application of the
presented gas sensor. Experimental results showed that the sensor resolution was ≤0.01cm–1
(equivalent to 0.06 nm), and the sensor sensitivity was at the level of 194 ppb with the application of
H2CO measurement.
Keywords: Quantum cascade laser; IR spectroscopy; high sensitivity; formaldehyde

Citation: Hengyi LI, “High Sensitivity Gas Sensor Based on IR Spectroscopy Technology and Application,” Photonic Sensors, DOI:
2016, 6(2): 127–131.

1. Introduction technology and instrumental integration system.

This paper aims at the issues existing in the gas
Infrared laser absorption spectroscopy is an
detection technology and instrument system research,
extremely effective tool for detecting trace gases.
and the key technologies in gas spectroscopy
The demonstrated sensitivity of this technique is at
detection based on the QCL were carried out.
the level of parts per billion (ppb). Presently, the
usefulness of the laser spectroscopy approach is
2. Spectrum detection and gas sensor
performance
limited by the availability of convenient tunable
sources in the spectroscopically important As to the experimental setup, the laser light was
“fingerprint” region from 3 μm to 20 μm [1]. Recent collimated by an off-axis parabolic gold mirror and
measurements with QC-DFB (distributed feedback then passed through a 20 m multi-pass white cell
quantum cascade laser) lasers [2, 3] have filled with the sample gas for detection. The
demonstrated the usefulness of these devices for radiation passed through a germanium Fabry-Pérot
highly selective real-time trace gas concentration etalon with a free spectral range of 500 MHz. The
measurements based on the absorption spectroscopy longitudinal modes supported by an etalon with the
with the sensitivity of several ppb [4‒11]. length d and refractive index n were separated by the
Due to the characteristics of high output power, free spectral range (FSR), νFSR=c/2nd. A typical
narrow line-width, room temperature operation, and signal was detected when a 200-ns-long laser pulse
so on, quantum cascade lasers (QCLs) have become at a wavelength around 8 μm passed through an
the research focus in the fields of trace gas detection empty cell, which was used to find the nonlinear

Received: 22 September 2015 / Revised: 13 November 2015

© The Author(s) 2015. This article is published with open access at Springerlink.com
DOI: 10.1007/s13320-015-0290-8
Article type: Regular
128 Photonic Sensors

relation between the time and frequency axes. “ ” in Fig. 2, with linear fit (represented by “ ”)
of the experimental data, and the calculated
2.1 Spectrum scanning Range
temperature tuning coefficient was ‒0.083cm–1/K.
The spectroscopy scanning range was
determined both by pulse current tuning and 1253.5 Wavenumber tunable value
Linear fit of wavenumber tunable value

Wavenumber tunable value (cm−1)

temperature tuning [11]. In order to express the 1253.0

spectrum scanning characteristics, 121 mTorr H2CO 1252.5

with the purity of 99.8% was introduced into the gas 1252.0
cell to obtain the absorption spectrum at different 1251.5 y = a + b*x
Equation

temperatures. Keeping the H2CO pressure and Weight

Residual sum
No Weighting
1.99038E-4
1251.0 of Squares

operation temperature unchanged, a series of Adj. R-square 0.99997

Value Standard Error

1250.5 Intercept 1252.10192 0.00208

scanning spectra were collected when tuning the Wavenumber

tunable value Slope -0.08309 1.77745E-4

1250.0
operation temperature from ‒15 ℃ to 20 ℃ with an −20 −15 −10 −5 0 5 10 15 20 25
Temperature (Celsius degree)
increment of 5 ℃. The scanning spectrum is shown
in Fig. 1, from which we can see that the scanning Fig. 2 Relationship between the QCL central wavenumber
and temperature.
range will move to the longer wavelength with an
increase in operation temperature. 2.3 Sensor spectrum resolution
Spectral resolution refers to the detection system
which can distinguish the minimum wavelength
difference, which is based on the Rayleigh criterion,
taking the absorption spectral full width at half
maximum (FWHM) as the spectral resolution of the
detection system. Figure 3 shows the Fourier
transform spectrometer (FTIR) spectrum collected
during laser operating temperature changing from
‒15 ℃ to +20 ℃, in which “ ” indicates the
formaldehyde absorption spectrum taken from this
experimental system, and “ ” indicates the
spectral measurement range of 1250 cm–1 –
1255 cm–1.
1.4 Spectrum from FTIR
Spectrum from this work 0.10
Fig. 1 Absorption line of H2CO at the sweeping temperature 1.2
A
from ‒15℃ to 20℃ . 1.0 E
D 0.05
2.2 Temperature tuning coefficient 0.8
Absorbance

0.6 C
0.00
When the QCL operation temperature changed, 0.4 B
the central emission wavelength would move, so it is 0.2
−0.05
necessary to test the temperature tuning coefficient 0.0

for the detection system. Formaldehyde was used as −0.2

−0.10
the detection gas to get the correlation of the central 1250.5 1251.5 1252.5 1253.5 1254.5
Wavenumber (cm−1)
emission wavelength with a function of tuning Fig. 3 Comparison of the absorption spectroscopy done in
temperature, which is represented by scatter dots this work and FTIR.
Hengyi LI: High Sensitivity Gas Sensor Based on IR Spectroscopy Technology and Application 129

10
The calculation of formaldehyde absorption lines
A, B, C, D, and E at 20 ℃ showed that the largest 9 Output power

width was 0.01 cm–1, which was exactly in the most 8 Linear fit

Output power (mW)

intensive areas, in which it would be more 7

reasonable to calculate the resolution. Other results 6 Equation y = a + b*x

are shown in Table 1. So the resolution of this

Weight No weighting
5 Residual sum 0.01063
of squares
spectroscopic detection system was ≤ 0.01 cm–1 4 Adj. R-square 0
Value Standard error
(equivalent to 0.06 nm), which could also be seen 3 Book3_B
Intercept
slope
6.97847
0
0.01882
--

from the comparison between the two results. It is 2

1 2 3 4 5 6
obvious that the QCL detection system has a Time (min)
significantly higher resolution than the FTIR (a)
spectrometer.
0.32
Table 1 Linewidth of experimental formaldehyde spectrum. Absorbance

0.30 Linear fit

Line A B C D E
Absorbance 0.28
FWHM (cm–1) 0.0100 0.0098 0.0097 0.0098 0.0096
0.26 Equation y = a + b*x

2.4 Gas sensor stability Weight

Residual
No weighting
1.39699E
0.24 0
Adj. R-s
The system stability renders the basic response Value Standard error
0.22 Book18_ Intercept 0.2746 6.82396E-4
throughout the detection time, mainly measured B slope 0 --

through the zero drift and interval drift. The zero 0.20
1 2 3 4 5 6
drift refers to the absence of the measured gas and Time (min)
the output changes throughout the detection time of (b)

the system. The interval drift refers to the changes in Fig. 4 System stability measured by (a) zero drift: output
power of QCL varying with time and (b) interval drift:
the continuous output in the case of measured gas absorbance varying with time.
presence over the measuring time. In the absence of
any absorbed gas existing, the zero drift is expressed 2.5 Sensor sensitivity
by the output light intensity versus time shown in Selecting the appropriate line and fitting
Fig. 4(a). In the case of H2CO gas absorption function is very important for the calculation of
as presence, the interval drift is analyzed detection limit. Usually, we choose the Voigt
from the relationship of the specific absorption function for nonlinear fitting and also the integration
peak intensity varying with time, as shown in of spectral lines in the region, and then we obtain the
Fig. 4(b). integrated absorption intensity area for calculating
It can be seen from the linear fit results that the detection sensitivity. The measured gas pressure is
quantum cascade laser output light intensity of the fixed, by varying the buffer gas pressure, so we can
standard deviation is 0.0188, the calculated “zero get different concentrations of formaldehyde.
drift” relative standard deviation is 0.26%, and the As shown in Fig. 5(a), the absorption spectrum
stability of the standard deviation of gas absorption of 60 mTorr H2CO in 200 Torr N2 was measured and
peak is 0.0007. Thus the calculated “interval drift” then fitted with the nonlinear Voigt function and
expressed by relative standard deviation is 0.24%. non-absorption part fitted with linear function. In
Both results illustrate a good stability of the QCL Fig. 5, “ ” indicates original absorption lines of
detection system. H2CO, “ ” indicates Voigt function fitting, and
130 Photonic Sensors

“ ” is a linear fit to the noise signal. As seen from temperature tuning coefficient was ‒0.083cm–1/K,
Fig. 5, the Voigt function fits well with the the absorbance reproducibility relative standard
experimental results. Figure 5(b) shows fitting deviation was less than 0.0051%, the system
residuals with the deviation in the range of 0.4%. resolution was less than 0.01 cm–1, and the system
After data processing and analysis, the calculated detects sensitivity reached 194 ppb, which was
minimum detectable concentration is 187 ppb. In this comparatively high sensitivity for detection trace
system, the signal to noise ratio (SNR) is calculated gas or toxic gases.
to be 1549, and the corresponding concentration is Acknowledgment
300 ppm, then the detection sensitivity is 194 ppb
when SNR is 1. The authors wish to thank the financial support
from the National Natural Science Foundation of
60 mTorr H2CO in 200 Torr N2 China (61505020) and the Fundamental Research
0.3 Voigt fit of absorbance
Linear fit of absorbance Funds for the Central Universities
Value Standard error
Absorbance y0 -0.00156 5.71109E-4 (ZYGX2013J007).
xc 1253.1360 1111.06902
0.2
Absorbance

A 0.01052 1.86554
wG
wL
0.0124
0.07577
0
1.71059E-4
Open Access This article is distributed under the terms
Value Standard of the Creative Commons Attribution 4.0 International
error
0.1
Absorbance Intercept -4.13479 5.50551
License (http:// creativecommons.org/licenses/by/4.0/),
µ=0.0853
Slope 0
E-4 --
E-5 which permits unrestricted use, distribution, and
SD=5.506 E-5 reproduction in any medium, provided you give
0.0 appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons license, and
1252.9 1253.0 1253.1 1253.2 1253.3 1253.4 1253.5 1253.6 indicate if changes were made.
Wavenumber (cm−1)
(a)
References
0.010
Residual of absorbance
[1] J. Faist, F. Capasso, D. L. Sivco, C. Sirtori, A. L.
0.008
Hutchinson, and A. Y. Cho, “Quantum cascade laser,”
0.006
Science, 1994, 264(158): 553‒556.
Residual of absorbance

0.004 [2] L. Wang and T. R. Sharples, “Intrapulse quantum

0.002 cascade laser spectroscopy: pressure induced line
0.000 broadening and shifting in the ν6 band of
−0.002 formaldehyde,” Applied Physics B, 2012, 108(2):
−0.004 427‒435.
−0.006
[3] T. Steck, N. Glatthor, T. von Clarmann, H. Fischer, J.
M. Flaud, and B. Funke, “Retrieval of global upper
−0.008
tropospheric and stratospheric formaldehyde (H2CO)
−0.010
1253.35 1253.40 1253.45 1253.50 1253.55 1253.60 distributions from high-resolution MIPAS-envisat
Wavenumber (cm−1) spectra,” Atmospheric Chemistry & Physics, 2008,
(b) 8(3): 463‒470.
Fig. 5 Formaldehyde spectrum and the residual signal: (a)
[4] X. Chen, L. Cheng, D. Guo, Y. Kostov, and F. S.
Voigt fit of 60 mTorr formaldehyde in 200 Torr N2 and (b) Choa, “Quantum cascade laser based standoff
standard error less than 0.4% between the experimental results photoacoustic chemical detection,” Optics Express,
and Voigt function fit. 2011, 9(21): 20251‒20257.
[5] L. Wang and S. Thomas-Reoben, “Monitoring
3. Conclusions hydrogen sulfide using a quantum cascade laser based
In this work, the gas sensor performance based trace gas sensing system,” Chinese Physics Letters,
2011, 28(6): 1183‒1187.
on the IR spectroscopy technology was carried out. [6] A. Wisthaler, E. C. Apel, J. Bossmeyer, A. Hansel, W.
The spectrum range was 1249 cm–1 – 1257 cm–1, the Junkermann, and R. Koppmann, “Technical note:
Hengyi LI: High Sensitivity Gas Sensor Based on IR Spectroscopy Technology and Application 131

intercomparison of formaldehyde measurements at the [9] C. L. Heald, A. H. Goldstein, J. D. Allan, and A. C.

atmosphere simulation chamber SAPHIR,” Aiken, “Total observed organic carbon (TOOC) in the
Atmospheric Chemistry & Physics Discussions, 2008, atmosphere: a synthesis of North American
7(6): 15619‒15650. observations,” Atmospheric Chemistry & Physics,
[7] K. Chance and J. Orphal, “Revised ultraviolet 2008, 8(7): 2007‒2025.
absorption cross sections of H2CO for the HITRAN [10] L. Wang, “Formaldehyde and methane spectroscopy
database,” Journal of Quantitative Spectroscopy & measurements based on mid-IR quantum cascade
Radiative Transfer, 2011, 112(9): 1509‒1510. laser system,” Journal of Infrared and Millimeter
[8] I. R. Burling, R. J. Yolelson, S. K. Akagi, S. P. Waves, 2014, 33(6): 591‒597.
Urbanski, C. E. Wold, D. W. T. Griffith, et al., [11] G. Duxbury, N. Langford, M. T. McCulloch, and W.
“Airborne and ground-based measurements of the Stephen, “Quantum cascade semiconductor infrared
trace gases and particles emitted by prescribed fires in and far-infrared lasers: from trace gas sensing to
the United States,” Atmospheric Chemistry & Physics, non-linear optics,” Chemical Society Reviews, 2005,
2011, 11(23): 12197‒12216. 34(11): 921‒934.