Frequency Comb Spectroscopy
of CO2, CH4, H2O, and Isotopes Over a 2km Outdoor Path:
Concentration Retrievals Using Different Absorption Models
Greg Rieker1,2, Fabrizio Giorgetta1, Ian Coddington1, William Swann1, Laura
Sinclair1, Chris Cromer1, Esther Baumann1, Alex Zolot1, Nathan Newbury1
Jonathan Kofler3, Gabrielle Petron3, Colm Sweeney3, Pieter Tans3
1National
Institute of Standards and Technology, Boulder, CO
2University of Colorado-Boulder
3National Oceanic and Atmospheric Administration, Boulder, CO
Aim of This Work
Demonstrate dual frequency comb technique for high resolution
absorption spectroscopy in a well-characterized outdoor environment
 270 cm-1 bandwidth with 0.0033 cm-1 point spacing (410-8 cm-1 res.)
 Fit with several absorption models
Wavelength (nm)
1650
1645
1640
1635
Absorbance
0.2
Partial CH4 Tetradecad
(+ H2O, CO2)
1615
1610
1605
1600
3001300001 band of CO2
(+ hotbands, H2O, HDO, 13CO2)
0.1
0.0
6060
6080
6100
6200
Wavenumbers (cm-1)
6220
6240
2
Demonstration: 2km Laser Path Above NIST
Calibrated pressure, wind direction, & velocity sensors at each end
Cavity ringdown instrument calibrated to World Monitoring
Organization (WMO) scale on 30m tower near center of path
Temperature extracted from the CO2 fit
3
Overview of Talk
Frequency Combs
+
Dual-comb Detection
Measurements of
Outdoor Absorption
Spectra
Comparison of
Retrievals with
Calibrated Sensor
Mole Fraction (ppm)
CO2
CH4
H2O
DCS
404.7 ± 0.8
Tower Sensor
397.6 ± 0.06
Difference
7.1 1.78%
1.878 ± 0.009 1.874 ± 0.002 0.004 0.20%
3223 ± 22
3168 ± 95
55 1.74%
Dual comb
spectrometer
4
Laser
Frequency
Comb
Intensity
Frequency Combs: Why are they special?
frequency
comb
“teeth”
nn-1 nn nn+1
broad
spectrum
coherent
& bright
calibrated
frequency scale
laser
rf synthesizer
“Revolutionized Optical Frequency Metrology”
5
Spectroscopy with a Comb Source
gas
Detector
absorption
profile
Detector
???
Intensity
Frequency
Comb
frequency
100 MHz
Grating Spectrometer
FTIR Spectrometer
VIPA Spectrometer
6
Spectroscopy with a Comb Source
gas
Detector
absorption
profile
Detector
???
Intensity
Frequency
Comb
frequency
100 MHz
Grating Spectrometer
FTIR Spectrometer
VIPA Spectrometer
None can resolve 100-MHz comb teeth
“Smear out comb”  Might as well use a light bulb!
7
Dual-Comb Spectroscopy
Comb Tooth by Comb Tooth Readout
gas
absorption
profile
Intensity
Frequency
Comb 1
Detector
Single
Photodiode
frequency
Frequency
Frequency
Comb
2 2
Comb
Local oscillator
Keilmann, et al, Opt. Lett. 29, 1542 (2004)
Schiller, Opt. Lett. 27, 766 (2002)
Schliesser, Opt. Exp, 13, 9029 (2005)
Coddington, PRL , 100, 013902 (2008)
Giaccari, Opt. Express, 16, 4347 (2008)
Bernhardt, Nat. Photon., 4, 55-57 (2009)
Coddington, Opt. Lett, 35, 1395 (2010)
Coddington, PRA, 043817 (2010)
8
Deschenes, Opt. Express, 23358 (2010)
Heterodyne Detection
Detector
Laser 1
Laser 2
Detector signal:
(Frequency Domain)
Optical frequency (THz)
Df
Magnitude
Laser light:
Laser 2 Laser 1
Df
0
RF frequency (MHz)
frep/2
9
Coherent Dual-Comb Detection
Laser light:
frep - frep=Df
Frequency
Comb 2
frep
LO Source
Comb 1 &2
tooth spacing
differs by Df
Detector
Frequency
Comb 1
frep
Optical frequency (THz)
Df
Magnitude
Detector signal:
(Frequency Domain)
0
RF frequency (MHz)
frep/2
10
Coherent Dual-Comb Spectroscopy
Laser light:
frep - frep=Dfrep
Detector
absorption
profile
Frequency
Comb 2
frep
LO Source
Comb 1 &2
tooth spacing
differs by Dfrep
gas
Frequency
Comb 1
frep
Optical frequency (THz)
Dfrep
Magnitude
Detector signal:
(Frequency Domain)
0
RF frequency (MHz)
frep/2
11
Coherent Dual-Comb Spectroscopy
Comb 1 &2
tooth spacing
differs by Dfrep
frep - frep=Dfrep
Detector
absorption
profile
Frequency
Comb 2
frep
LO Source
Laser light:
gas
Frequency
Comb 1
frep
Optical frequency (THz)
Transmitted
Intensity:
Detector signal:
(Frequency Domain)
Magnitude
Dfrep
0
Optical
frequency
(THz)
RF frequency
(MHz)
frep/2
12
Example “Raw” Dual-Comb Data over 2km Path
Wavelength (nm)
1660
Transmitted Intensity
1.2
1640
1620
CH4 absorption features
(+ weak H2O and CO2)
1.0
1600
CO2 absorption features
(+ weak H2O, HDO, and hotband)
0.8
0.6
0.4
0.2
0.0
180
182
184
186
188THz
Frequency (THz)
Broadband intensity variation due to combination of
– Spectral variation of comb output intensity
– Spectral variation of transmission optics
– Selective optical filtering
13
Example “Raw” Dual-Comb Data over 2km Path
Wavelength (nm)
1660
1640
1620
1600
Transmitted Intensity
1.0
CH4 absorption features
(+ H2O and CO2)
CO2 absorption features
(+ H2O, HDO, and hotband)
0.0
6000
6050
6100
6150
6200
6250
Wavenumber (cm-1)
0.0033 cm-1 point spacing, each point representing absorption on two
Hz-width comb teeth <0.0016 cm-1 apart
 Absorption model simulations include the tooth separation, however
 Negligible instrument effects on these spectra
14
Example “Raw” Dual-Comb Data over 2km Path
Wavelength (nm)
1660
1640
1620
1600
1.0
Transmitted Intensity
CH4 absorption features
(+ H2O and CO2)
CO2 absorption features
(+ H2O, HDO, and hotband)
0.0
6000
6050
6100
6150
6200
6250
Wavenumber (cm-1)
170 minute average under windy, well-mixed conditions:
– Less than 0.3% variation in temperature, pressure & concentration
15
Absorption Model Fitting
Wavelength (nm)
1650
1645
1640
1635
Absorbance
0.2
Partial CH4 Tetradecad
(+ H2O, CO2)
1615
1610
1605
1600
3001300001 band of CO2
(+ hotbands, H2O, HDO, 13CO2)
0.1
0.0
6060
6080
6100
6200
6220
6240
Wavenumbers (cm-1)
Fitting Procedure:
 Create absorption models @ multiple temperatures + measured pressure
 Fit to measured data with 1 free parameter per species (representing mole
fraction) + temperature + polynomial baseline (to remove background)
16
Fit with Hitran 2008 + Voigt
Absorbance noise typically < 5×10-4 for 170 minute average
 System not yet optimized
 Laser far below eye-safe limit
 Order of magnitude improvement will be achieved with latest systems
17
Fit with Hitran 2008 + Voigt
18
Fits with Other Models
Hitran 2012:
 CO2 - Similar residuals to Hitran 2008
 CH4 – Larger residuals compared with Hitran 2008
19
Fits with Other Models
Line-by-line Voigt:
 Allowed collisional width, area, and line center to float
20
Fits with Other Models
Line-mixing & speed dependent Voigt*:
 Only for CO2
 Residual changes significantly
* Thompson, et al. JQSRT 113, 2265–2276 (2012), Devi, et al. J Mol Spectrosc 245, 52–80 (2007), Predoi-Cross, et al. Can. J. Phys. 87, 517–535 (2009)
21
Concentration Retrievals
CO2
Mole Fraction Retrieval (ppm)
Hitran 08 Hitran 12 LM / SD
Toth
408.7
407.7
404.7
406.2
Systematic Unc. (ppm)
excluding spectral model
0.8
0.21%
CH4
1.878
1.985
--
--
0.009
0.45%
H2O
HDO
13
CO2
3223
1.13
3217
0.97
---
---
22
0.13
0.73%
11%
4.5
4.4
--
4.2
1.7
37%
Variation between models exceeds our model-independent systematic
uncertainty
22
Concentration Retrievals
CO2
Mole Fraction Retrieval (ppm)
Hitran 08 Hitran 12 LM / SD
Toth
408.7
407.7
404.7
406.2
Systematic Unc. (ppm)
excluding spectral model
0.8
0.21%
CH4
1.878
1.985
--
--
0.009
0.45%
H2O
HDO
13
CO2
3223
1.13
3217
0.97
---
---
22
0.13
0.73%
11%
4.5
4.4
--
4.2
1.7
37%
Model-independent systematic uncertainty is based on sensitivity of
retrieved concentration to:
 Maximum pressure and temperature inhomogeneities along path
 Uncertainty in pathlength and pressure
 Baseline correction (10× larger contribution than other factors)
23
Demonstration: 2km Laser Path Above NIST
Cavity ringdown instrument calibrated to WMO scale on 30m tower
near center of path
24
Comparison with WMO-calibrated Sensor
Mole Fraction (ppm)
CO2
CH4
H2O
DCS
404.7 ± 0.8
Tower Sensor
397.6 ± 0.06
Difference
7.1 1.78%
1.878 ± 0.009 1.874 ± 0.002 0.004 0.20%
3223 ± 22
3168 ± 95
55 1.74%
CO2: Line mixing/speed dependent Voigt model
CH4, H2O: Hitran 2008 with Voigt model
25
0
5-minute Time-resolved Results
24:00
6/1/2013
12:00
24:00
6/2/2013
12:00
24:00
6/3/2013
12:00
b
1 day
(ppm)
420
400
2.0
(ppm)
CH4, dry
CO2, dry
440
1.9
Temp.
(C)
HDO
(ppm)
4
2
0
30
10
24:00
6/1/2013
12:00
24:00
6/2/2013
12:00
24:00
6/3/2013
12:00
 Excellent overall tracking between frequency comb measurement
and tower sampling device across multiple days
– 26 C temperature changes (9% full scale)
– 15.5 mbar pressure changes (2% full scale)
26
Summary
 Demonstrated dual-comb measurements of CO2, CH4, H2O, and HDO
over a 2-km open path at NIST
– First high resolution dual-comb spectroscopy outside of the laboratory
o 270 cm-1 of spectral coverage, 0.0033 cm-1 point spacing, 410-8 cm-1 resolution
– Robust against turbulence
– Eye-safe
– High SNR spectra during well-mixed conditions can reveal the quality
of the underlying absorption model
– Concentration comparisons with tower-mounted point sensor are
promising
 Future steps
– Many technical possibilities: longer range, multiple beams, Oxygen A-band,
2 micron sources, laboratory measurements, robust operation...
– High temperature laboratory H2O measurements for combustion underway
– What is useful to the HITRAN community?
27
Future Possibilities
Provide a means to accurately measure concentrations of multiple
greenhouse gases within and around sources (cities, wells, landfills…)
 CO2, CH4 (Methane), H2O, isotopes
 1-10 km length scales (between point & satellite-based sensors)
 Eye safe
Open-path Sensing Concept
Sensor Hub
Source
Retro-reflector
Beam
Sensor Hub
28
0
5-minute Time-resolved Results
24:00
6/1/2013
12:00
24:00
6/2/2013
12:00
24:00
6/3/2013
12:00
b
1 day
(ppm)
420
400
2.0
(ppm)
CH4, dry
CO2, dry
440
1.9
Temp.
(C)
HDO
(ppm)
4
2
0
30
10
24:00
6/1/2013
12:00
24:00
6/2/2013
12:00
24:00
6/3/2013
12:00
 Statistical uncertainty of 5-minute fits matches systematic
uncertainty
 Maintain link / data quality over multiple days
29
0
5-minute Time-resolved Results
24:00
6/1/2013
12:00
24:00
6/2/2013
12:00
24:00
6/3/2013
12:00
b
1 day
(ppm)
420
400
2.0
(ppm)
CH4, dry
CO2, dry
440
1.9
Temp.
(C)
HDO
(ppm)
4
2
0
30
10
24:00
6/1/2013
12:00
24:00
6/2/2013
12:00
24:00
6/3/2013
12:00
 Expect differences, particularly during stable conditions due to
different sampling paths
30