ARTICLE IN PRESS

Contents lists available at ScienceDirect

Journal of Quantitative Spectroscopy &Radiative Transfer

Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–1140

0022-40

doi:10.1

� Cor

E-m

journal homepage: www.elsevier.com/locate/jqsrt

Temperature dependence of the absorption spectrum ofCH4 by high resolution spectroscopy at 81 K: (I) The regionof the 2n3 band at 1.66mm

Le Wang, Samir Kassi, Alain Campargue �

Laboratoire de Spectrometrie Physique (associated with CNRS, UMR 5588), Universite Joseph Fourier de Grenoble, B.P. 87,

38402 Saint-Martin-d’H�eres Cedex, France

a r t i c l e i n f o

Article history:

Received 18 September 2009

Received in revised form

21 October 2009

Accepted 26 October 2009

Keywords:

Methane

CH4

Near infrared

Absorption spectroscopy

HITRAN

GOSAT

Titan

73/$ - see front matter & 2009 Elsevier Ltd. A

016/j.jqsrt.2009.10.019

responding author. Tel.: +33 476514319; fax:

ail address: Alain.Campargue@ujf-grenoble.fr

a b s t r a c t

In a recent contribution, (Gao B, Kassi S, Campargue A. Empirical low energy values for

methane transitions in the 5852–6181 cm�1 region by absorption spectroscopy at 81 K. J

Mol Spectrosc 2009;253:55–63.), the low energy values of methane transitions between

1.71 and 1.62mm were derived from the variation of the line intensities between 296

and 81 K. The line intensities at 81 K were retrieved from the high resolution absorption

spectrum of methane recorded at liquid nitrogen temperature by direct absorption

spectroscopy using a cryogenic cell and a series of distributed feed back (DFB) diode

lasers. For the line intensities at 296 K, the values provided by the HITRAN database

were used. As a consequence of the relatively high intensity cut off (4�10�24

cm/molecule) of the HITRAN line list in the considered region, the lower energy values

were derived for only 845 of the 2187 transitions measured at 81 K. In the present work,

our line list was extended by the retrieval of many weak line intensities leading to a set

of 3251 transitions. The minimum value of the measured line intensities (at 81 K) is on

the order of 10�26 cm/molecule. In relation with the project ‘‘Greenhouse Gases

Observing Satellite’’ (GOSAT), a much more complete line list for CH4 at 296 K has

become available (intensity cut off of 4�10�26 cm/molecule). By applying the two

temperature method to our line intensities at 81 K and GOSAT intensities at 296 K, the

lower energy values of 2297 transitions could be derived. These transitions represent

99.1% and 90.8% of the total absorbance in the region, at 81 and 296 K respectively. This

line list provided as Supplementary Material allows then accounting for the

temperature dependence of CH4 absorption below 300 K. The investigated spectral

range is dominated by the 2n3 band near 6005 cm�1 which is of particular interest for

atmospheric retrievals. The factor 2 narrowing of the Doppler linewidth from room

temperature down to 81 K has allowed the resolution of a number of 2n3 multiplets and

improving the line intensity retrievals. A detailed comparison with GOSAT and HITRAN

line lists has revealed a number of possible improvements.

& 2009 Elsevier Ltd. All rights reserved.

1. Introduction

The 2n3 absorption band of methane near 1.66mmhas a particular interest for atmospheric retrievals.

ll rights reserved.

+33 4 76 63 54 95.

(A. Campargue).

SCIAMACHY on board of ENVISAT satellite uses a micro-window corresponding to strong Q and R multiplets tomonitor methane with very high sensitivity for the entireatmospheric column [1,2]. The Greenhouse gases Obser-ving SATellite (GOSAT) launched last January uses thesame 2n3 band to measure CH4 concentration in the nearsurface layers [3]. Systematic biases in the methaneconcentration may be introduced by an insufficient

ARTICLE IN PRESS

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–1140 1131

characterization of the CH4 spectrum from laboratoryspectra [4,5]. This characterization is made difficult by (i)the blending of the 2n3 multiplets at room temperatureeven at Doppler resolution, (ii) collisional and line mixingeffects [6,7] and (iii) overlapping with numerous weakerlines which are poorly characterized. The highly con-gested background of weak transitions is due to thesuperposition of other bands of the high energy part of thetetradecad (14 levels and 60 sublevels in interaction)which are not yet theoretically modeled.

In the tetradecad region (5500–6150 cm�1), theHITRAN database in its last version [8] reproduces theempirical line positions and intensities obtained byMargolis 20 years ago [8,9]. The knowledge of the lowerenergy levels is necessary to calculate methane absorp-tion at different temperatures, line intensities beingproportional to the Boltzmann factor. This is why, inabsence of theoretical interpretation (except for theassigned 2n3 multiplets), the lower state energies pro-vided in HITRAN are those obtained by Margolis from thetemperature dependence of line intensities measured atroom [8] and reduced (180–220 K) [9] temperatures.

The 10 Torr vapour pressure available for methane atliquid nitrogen temperature (LNT) makes possible torecord spectra at a much lower temperature i.e. intemperature conditions approaching those existing inthe atmosphere of Titan, for instance. This increasesconsiderably the variation of the line intensities fromtheir room temperature (RT) values and allows a moreprecise determination of the lower state energies (inparticular those originating from low J levels whichdominate the spectrum at LNT). In a recent contribution[11], we applied this two temperature method to the5850–6180 cm�1 region using the HITRAN line intensitiesat RT and the line intensities retrieved from spectrarecorded at LNT with a new cryogenic cell and a seriesof distributed feed back (DFB) laser diodes. In our region,Margolis’s line list (as reproduced in HITRAN) provideswith 1300 transitions while we could measure 2187transitions at LNT [11]. Using the wavenumber agreementas criterion to associate RT and LNT line intensities,the low J values of 845 transitions were determined [11].1342 transitions measured in our LNT spectrum were leftwithout lower state energy determination for lackof the corresponding RT line strength values in theHITRAN database. This is a consequence of thehigh intensity cut off of the HITRAN line list inthe considered region (4�10�24 cm/molecule [8]) whichis one much higher than our detectivity limit at LNT(3�10�26 cm/molecule [11]).

In relation with the ‘‘Greenhouse Gases ObservingSatellite’’ (GOSAT) project [3], a much more complete linelist has become available for methane at room tempera-ture [12]. This line list was constructed from FTS spectrarecorded with a coolable white-type cell in a large varietyof temperature and pressure conditions [13]. Most of therecordings were performed with a 20 m pathlengthand temperature values ranged between 240 and296 K [13,14]. The intensity cut off of the GOSAT line list(about 4�10�26 cm/molecule) is two orders of magnitudelower than the HITRAN cut off leading to a considerable

extension of the dataset: in our region (5852–6181 cm�1),the GOSAT line list provides with 4878 transitionscompared to 1300 in HITRAN.

The present contribution takes advantage of this newRT line list to extend the empirical determination of thelower state energies of the transitions observed in our LNTspectra. In addition, we have performed a new line by linefitting in order to extract the spectroscopic parameters ofall the weak lines detected in our LNT spectra. It allowedincreasing the LNT dataset from 2145 to 3123 transitions.As a result of the extension of the measurements both atRT and LNT, we could increase from 845 to more than2200 the number of determinations of lower stateenergies from the temperature variation of the lineintensities.

After a brief description of the experiment details andof the line list construction (Section 2), the resultsobtained by applying the two temperature method willbe presented (Section 3). We will then discuss a detailedcomparison of our results with the HITRAN [8] and GOSAT[12] line lists in particular for the strong multiplets of the2n3 band (Section 3).

2. Experiment

Most of the spectra presently analysed are thoseobtained in Refs. [11,15] by direct absorption spectro-scopy using a cryogenic cooled down to LNT. Briefly, thecell is based on an original design with no externalvacuum jacket, the low pressure CH4 sample constitutingitself a good thermal insulation. In the absence of aninternal pair of windows, the gas sample fills both theinside of the cryostat and the thermal insulation volume(see Ref. [15]). The absorption path length is 141.8 cm andthe gas pressure is continuously measured during therecordings (MKS Baratron, 10 Torr range). The wholeinvestigated region was recorded with a sample pressureof 9.35 Torr (at LNT). As the strong multiplets of the 2n3

band were saturated at this pressure, additional spectrawere recorded with a 0.75 Torr pressure.

A series of 13 DFB fibred laser diodes was used aslight sources allowing for a continuous coverage ofthe 5852–6181 cm�1 range, except for an inaccessible8.7 cm�1 gap between 6124.1 and 6132.8 cm�1. The DFBlaser diode was tuned over about 30 cm�1 by a slowtemperature scan from �10 to 60 1C within 11 minutesThe individual spectra were linearized using an etalonsignal, and calibrated independently by statisticallymatching the observed spectral line positions to theHITRAN line positions (at 296 K) [8]. The standarddeviation error of the differences between our linepositions and HITRAN values was minimized leading torms values on the order of 1�10�3 cm�1. More detailsabout the spectra acquisition and the cell design can befound in Ref. [15]. This reference includes a movieshowing the evolution of the spectrum near 6096 cm�1,recorded during the cell cool-down to 77 K. The steadystate was achieved about one hour after the filling of thecryostat with liquid nitrogen. A gas temperature value of8171 K was determined from the Doppler profile of

ARTICLE IN PRESS

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–11401132

several tens of well isolated lines, the error bar corre-sponding to one standard deviation [15]. This temperaturevalue is what we refer to as ‘‘liquid nitrogen temperature’’(LNT). Note that the temperature gradient over the 1 cmdistance between the ends of the cryostat and the opticalwindows was found to have a negligible impact on theobserved line profile.

Fig. 1. Comparison of the spectra of methane recorded at liquid nitrogen t

6062 cm�1 with the GOSAT line list [12]. The RT and LNT spectra were recorded

temperature variation of the line intensities are indicated.

Fig. 2. An example of simulation of the CH4 spectrum recorded at LNT. From top

spectrum resulting from the line fitting procedure (a Voigt profile was affected

spectra.

For comparison, the RT spectra in the same regionwere recorded with the same experimental setup but they were not exploited for line intensityretrieval. Fig. 1 shows a comparison between the RT andLNT spectra around 6062 cm�1. The cooling inducesstrong changes on the line intensities distribu-tion leading to dramatic changes in the appearance

emperature (upper panel) and room temperature (middle panel) near

at 9.41 and 9.40 Torr respectively. The round J values obtained from the

to bottom: (a) experimental spectrum at LNT (P=9.39 Torr), (b) simulated

to each line), and (c) residuals between the simulated and experimental

ARTICLE IN PRESS

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–1140 1133

of the LNT spectrum compared to the RT spectrum.The stick spectrum corresponding to the GOSAT linelist (at 296 K) included in Fig. 1 shows an excellentagreement with our RT spectrum. We note that thenoise level of our RT spectrum (noise equivalentabsorption of about 10�6 cm�1) corresponds roughly tothe GOSAT intensity cut off of 4�10�26 cm/molecule at10 Torr pressure.

3. Line intensity retrieval

The pressure self broadening of the n3 and 2n3 bands at81 K has a value on the order of 0.20 cm�1/atm (HWHM)[16]. Assuming the same value for the vibrational bandscontributing to our spectrum, it leads to a 2.8�10�3 cm�1

HWHM at 10.0 Torr which is small but significantcompared to the Doppler width (HWHM 4.8�10�3 cm�1

at 81 K). We therefore adopted a Voigt function of the

Fig. 3. Overview spectrum of methane between 5852 and 6181 cm�1 recorded

line list at 296 K [12]. The full circles highlight the 2297 pairs of transitions fo

wavenumber for the line profile. Note that the DFB linewidth (1–5 MHz) is much smaller than the Dopplerbroadening [175 MHz (HWHM) at LNT] and is thereforeneglected.

The line intensity, Sv0(cm/molecule), of a rovibrational

transition centred at n0, was obtained from the integratedline absorbance, Iv0

(cm–2/molecule):

Iv0ðTÞ ¼

Zline

avl dv¼

Zline

lnI0ðvÞ

IðvÞ

� �dv¼ Sv0

ðTÞNl ð1Þ

where I0ðvÞ=IðvÞ is the ratio of the incident intensity to thetransmitted intensity, l the absorption pathlength in cm, nthe wavenumber in cm�1, a(n) the absorption coefficientin cm�1, and N the molecular concentration in molecule/cm3 obtained from the measured pressure value: P=NkT.

The first (manual) step of the line intensities retrievalconsisted in the determination of spectral sections ofoverlapping or nearby transitions that could be treated

at liquid nitrogen temperature (this work) and as provided in the GOSAT

r which the lower energy values could be derived.

ARTICLE IN PRESS

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–11401134

independently. For each of these spectral sections, thelocal baseline was assumed to be a quadratic function ofthe wavenumber. Spectral lines were fitted with a Voigtfunction profile with the HWHM of the Gaussian compo-nent fixed to its theoretical value. Apart from the localbaseline, three parameters (line centre, integrated absor-bance, HWHM of the Lorentzian component) weredetermined using an interactive least square multi-linefitting program based on the Levenberg–Marquardtalgorithm. For each piece of spectrum, an average valueof the Lorentzian component of the Voigt profile wasobtained by fitting about 20 well isolated lines. In the caseof blended overlapping lines or low signal to noise ratios,this parameter could be constrained to its average value.

Fig. 2 shows an example of comparison between themeasured and fitted spectra. A typical value of 2�10�4 isachieved for the rms deviation of the experimental andsimulated absorbances. This rms value corresponds to aminimum line strength value of 10�26 cm/molecule at LNT.

The complete LNT line list was obtained by mergingthe line lists corresponding to the different DFB laserdiodes, each extending over about 30 cm�1. In the casewhere the coverage of two diodes overlapped, the centreand intensities of the lines corresponding to these over-lapping regions were averaged. The final linelist consistsof 3251 lines with intensity values ranging from1.19�10�26 to 4.53�10�21 cm/molecule for methane innatural abundance at 81 K.

4. Determination of the lower state energy

Independently of any rovibrational assignment, thelow energy value, E, EEB0J (J+1), and then the value of theangular momentum J can be deduced from the intensityvalues of a given transition, Sv0

ðTÞ and Sv0ðT0Þ, at two

temperatures, T and T0. This method has been successfully

Fig. 4. Percentages of lines (left hand) and absorbance (right hand) correspon

do0.002 cm�1, 0.002odo0.003 cm�1 and 0.003odo0.005 cm�1 (d is the dif

applied to methane in a number of studies [17–19]including Margolis’ study [10] and our recent study ofthe icosad region [20].

Taking into account the partition function, the lowerenergy can be calculated from the following equation (seeRef. [11] for more details):

lnSv0ðTÞT3=2

Sv0ðT0ÞT

3=20

!¼ � E

1

kT0�

1

kT

� �ð2Þ

where T0 and T are the RT and the LNT, respectively. Asmentioned in the Introduction, we adopted GOSAT values[12] for the line intensities at RT. An overview comparisonof our LNT line list and the GOSAT line list (at 296 K) ispresented in Fig. 3.

The coincidence of the line centres is the only criterionused to associate the transitions of the RT and LNT linelists. Consequently, the obtained results depend on thequality of the wavenumber calibration of the spectra andon the precision on the line centre determinations. Ina first step, we considered as identical RT and LNTtransitions when the difference, d, of their line centresdiffered by less than 0.002 cm�1. This value correspondsto one-fifth and tenth of the Doppler width (FWHM) at 81and 296 K, respectively. It takes into account the un-certainties on the centre values in the RT and LNT linelists. 1862 pairs of transitions were found in coincidenceaccording to this criterion. They represent 58% and 39% ofthe total number of lines detected at LNT and RTtemperatures, respectively but 94% and 77% of the totalabsorbance in the studied region. A careful examination ofthe line lists and spectra shows that the do0.002 cm�1

criterion is too strict and that a number of additional pairsof lines correspond undoubtedly to the same transitionsand should be associated. Such situations are due to a lackof precision in the centre values obtained from the lineprofile fitting and concern mainly blended lines or lines

ding to different association criteria of the RT and LNT line intensities:

ference between the RT and LNT line centres).

ARTICLE IN PRESS

Fig. 5. Upper panels. Histograms of the obtained J values with a step

interval of 0.5 both in terms of line numbers and of intensities at RT and

LNT. A 70% of the obtained J values fall in a 70.25 interval around

integer values. It corresponds to 88% of the absorbance at 81 K. Lower

panels. Scattered graph of the RT and LNT line intensities [12] versus the

lower J values obtained from the strengths of the transitions at 81 and

296 K between 5852 and 6181 cm�1.

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–1140 1135

which appear very weak either at RT (low J values) or LNT(high J values). Due to the larger Doppler broadening andhigher rotational congestion, the spectral congestion ismore important in the RT spectrum but the LNT spectrumis not free of strongly blended absorption features.

We then decided to ‘‘relax’’ the coincidence criterionup to do0.003 cm�1 and undertook a careful line by lineexamination of the spectra and line lists in order toassociate the maximum number of identical transitionswithout increasing the number of accidental coincidences.The additional absorbance of the lines corresponding to0.002odo0.003 cm�1 represents 3.9% of the total LNTabsorbance. The spectrum analysis showed that somepairs of lines with 0.003odo0.005 cm�1 interval shouldalso be associated. They correspond to very weak linesclose to the noise level in the LNT spectrum and representonly 1.3% of the total LNT absorbance (but 7.0% of the totalnumber of lines). Fig. 4 summarizes the variation of thepercentage of associated lines both in terms of absorbanceand of number of lines, according to the chosen criterion.The 2297 pairs of associated lines have been highlightedon the overview plot (Fig. 3). They represent 99.14% and90.78% of the total absorbance at 81 and 296 Krespectively. Nevertheless, about half of the transitionsincluded in GOSAT list remain without LNT partners(Fig. 4). This is mainly due to the high number of linescorresponding to high J values in the RT spectrum whichfall below our detection limit at LNT, as a consequence ofthe considerable depletion of the population of the highenergy levels. For instance, the intensity of a transitionfrom a J=11 level is decreased by a factor of 1075 whencooling down to 81 K.

Using Eq. (2), the E values were empirically determinedand the corresponding J values were calculated asthe positive root of the E=B0J(J+1) equation (withB0=5.241 cm�1). As example, the obtained rounded J

values have been indicated on the RT and LNT spectradisplayed in Fig. 1.

We present in Fig. 5 a scattered graph showing thedistribution of the RT and LNT line intensities of theassociated transitions as a function of the obtained J

values. As expected the obtained J values show a clearpropensity to be close to integer values. Fig. 5 includes thehistogram both in terms of line numbers and lineintensities. We calculated that 70% of the J values fallwithin a 70.25 interval around integer values. Thecorresponding lines represent 88% of the absorbance inagreement with the fact that the uncertainty on the lineintensities values (and then on the J value) is larger for theweaker lines. The histograms show that the quality of theJ determination degrades both for the J=0–2 values andfor the higher J values (4 9) which correspond tothe weakest transitions in the RT and LNT spectra,respectively. The marked dissymmetry of the RT andLNT panels reflects the considerable intensity variation ofthe transitions corresponding to the low and high J values.

The complete list of the 3251 transitions measured atLNT is provided as Supplementary Material. For the 2297transitions with J determination, the E and J values of thelower state are given together with the GOSAT lineintensities at RT [12]. A sample of this line list is

reproduced in Table 1. As larger uncertainties on theline intensities are expected for the transitions showinglarger difference between their RT and LNT centres, wehave specifically marked the J values obtained fromtransitions with 0.002odo0.003 cm�1 and 0.003odo0.005 cm�1 which are believed to be less reliable.

The GOSAT line list includes lower state J values forabout half of the transitions [12]. These J values weredetermined either on the basis of a theoretical model orby applying the same two temperature method to FTSspectra recorded down to 240 K [13]. 1862 of our 2297 J

determinations can be compared to GOSAT J values. For

ARTICLE IN PRESS

Table 1Wavenumbers and strengths of the absorption lines of methane recorded at 81 K near 6040 cm�1.

This work (81 K) GOSAT (296 K) [12]

ISO Line centre

(cm�1)

Line intensity

(cm/mol)

Line centre

(cm�1)

Line intensity

(cm/mol)

Jlow Elow (cm�1) Jlow Note

6038.9964 3.08E�25 6038.9970 2.80E�25 6 143.278 4.75

6039.0179 4.02E�25 6039.0202 2.05E�23 11 bl

6039.0207 5.63E�25 6 bl

6039.0274 5.96E�26 6039.0277 1.93E�23 11 598.599 10.20

6039.0863 3.11E�25 6039.0866 1.82E�24 7 287.632 6.93

6039.1418 8.05E�25

62 6039.1572 2.80E�23 6039.1565 2.12E�23 4 129.154 4.49

62 6039.1790 1.64E�23 6039.1796 9.58E�24 4 109.023 4.09

62 6039.1943 1.07E�23 6039.1943 5.14E�24 4 93.804 3.76

6039.2137 3.95E�25

62 6039.2338 1.61E�23 6039.2340 8.16E�24 4 98.006 3.85

6039.2899 1.26E�25 6039.2901 7.03E�24 9 462.535 8.91

6039.3628 1.05E�24 6039.3635 7.45E�24 7 302.467 7.11

6039.3756 2.69E�25

6039.3899 1.59E�25 6039.3872 9.18E�24 9 465.145 8.93 *

6039.4010 5.12E�25 6039.4012 4.64E�24 7 321.470 7.35

6039.5096 3.36E�25 6039.5096 2.06E�23 9 469.677 8.98

6039.6431 7.57E�26

6039.6578 5.02E�24 6039.6580 3.52E�23 7 301.591 7.10

6039.6885 4.63E�26

6039.7049 5.37E�24 6039.7057 1.79E�24 2 65.676 3.08

6039.7459 2.19E�25 6039.7420 2.77E�23 10 525.708 9.53 **

6039.7583 2.12E�24 6039.7583 5.97E�24 5 230.797 6.16

6039.7722 2.11E�25

6039.7955 8.35E�26

6039.8169 2.66E�25 6039.8167 6.49E�25 5 219.801 6.00

6039.8770 6.48E�25 6039.8769 7.42E�25 5 161.172 5.07

6039.8941 2.56E�24 6039.8946 4.98E�25 2 23.652 1.68

6039.9238 1.01E�24 6039.9244 1.88E�24 6 198.850 5.68

6039.9649 4.16E�26 6039.9630 3.28E�25 8 310.879 7.22

6040.0038 3.48E�25 6040.0041 8.05E�24 8 394.072 8.19

6040.0379 3.12E�25

6040.0513 4.28E�26 6040.0491 4.88E�25 339.282 7.56 *

6040.1427 5.01E�25 6040.1428 1.03E�23 8 384.876 8.08

6040.2470 1.65E�24 6040.2484 5.01E�24 6 236.858 6.24

6040.2687 4.43E�26

6040.3843 5.05E�25 6040.3843 3.04E�23 9 468.337 8.97

6040.4055 4.94E�25 6040.4055 3.05E�23 9 470.080 8.98

6040.5046 3.67E�24 6040.5048 9.98E�24 6 228.182 6.12

6040.5262 2.37E�24 6040.5265 6.44E�24 6 228.162 6.12

6040.7426 6.63E�25 6040.7426 4.81E�24 7 304.372 7.14

6040.7796 6.10E�26 6040.7800 3.00E�25 7 274.101 6.75

6040.8260 6.81E�25 6040.8261 5.22E�24 7 308.434 7.19

6040.8573 1.07E�24 6040.8558 1.15E�24 5 156.328 4.98

6040.9202 4.04E�26

6040.9480 2.77E�24 6040.9479 5.70E�23 8 385.142 8.09

6040.9898 3.30E�24 6040.9895 3.30E�23 8 329.259 7.44

6041.0064 1.08E�24 6041.0062 2.15E�23 8 382.788 8.06

The low energy E and J values were obtained for the transitions whose centres coincide with the GOSAT line positions at 296 K [12]. This Table is a small

section of the list of 3251 transitions attached as Supplementary Material.

Notes: The label ‘‘62’’ in the first column corresponds to lines due to 13CH4 or strongly blended with a 13CH4 transition.

The ‘‘*’’ and ‘‘**’’ symbols in the last column mark the lines whose RT and LNT line centres differ by 0.002odo0.003 cm�1 and 0.003odo0.005 cm�1,

respectively. ‘‘bl’’ indicates that several RT (or LNT) components are found in coincidence with a single line in the LNT (or RT) list, which renders the

determination of the corresponding lower J value.

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–11401136

80.4% of them, GOSAT J values and our rounded J valuescoincide while a difference of 71 is found for 14.2% (seeFig. 6). Note that we could derive the J values of 437transitions provided without J values in the GOSAT list.

We examined in details the cases of disagreementbetween our J values and GOSAT J values. For a number oflines, a Note has been added in the line list explaining the

possible origin of the discrepancy. In most of them,the disagreement can be explained by the blending of thelines which has led to inaccurate line intensity values. Inprinciple, GOSAT J assignments are supposed to be morereliable when supported by a theoretical model but insome cases our observations are found in clear contra-diction with GOSAT assignments. As an example, the

ARTICLE IN PRESS

Fig. 6. Histogram of the differences between the J values obtained from

the temperature dependence of their line intensities (this work) and

those of the GOSAT list [12].

Fig. 7. Comparison of the spectra of methane recorded at liquid nitrogen

temperature (upper panel) and room temperature (middle panel) in the

region of the P(10) multiplet of the 2n3 band near 5891 cm�1, with the

GOSAT line list at 296 K [12]. The room temperature and liquid nitrogen

temperature spectra were both recorded at 9.39 Torr respectively. The

round J values obtained from the variation of the line intensities are

indicated in the two upper panels while GOSAT J values are included in

the lower panel. Note the line at 5891.457 cm�1 (marked with an arrow)

which does not follow the temperature variation of the nearby J=10

transitions.

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–1140 1137

spectrum in the region of the P(10) multiplet of the 2n3

band is presented in Fig. 7. By cooling down to 81 K, lineintensities from J=10 lower levels are divided by a factor240 (Eq. (2)). This is indeed observed in Fig. 7 for some ofthe components assigned to J=10 in GOSAT list but theline at 5891.457 cm�1 remains strong in the LNT spectrumand does not follow the temperature dependence of aJ=10 line. From the ratio of GOSAT intensity and our LNTintensity, we calculated a J value of 7.1 for this lineindicating that the main contribution to the intensity atLNT is not a J=10 transition. It is interesting to note that inthe LNT spectrum displayed in Fig. 7, two lines (markedwith *) are observed with a relatively high intensity(on the order of 2�10�23 cm/molecule) but vanish in theRT spectrum. We could not derive their lower state energyvalue but they correspond obviously to very low J values(0 or 1). This example illustrates the completeness of ourline list at LNT compared to temperature extrapolationfrom line lists at RT such as GOSAT line list.

As the 12CH4 and 13CH4 isotopologues of methane havepractically the same ground state rotational constants(5.241 cm�1), the J determination from the line intensityratio is valid for the two species and the isotopologuecannot be discriminated from the spectra of methane innatural abundance. Nevertheless, by comparison with

spectra of 13C enriched methane that we recordedseparately at LNT, the 13CH4 transitions could be identi-fied. They are indicated in the line list attached asSupplementary Material.

5. Discussion of the 2m3 multiplets

As mentioned above, the 2n3 band is a strong band ofparticular interest for atmospheric retrievals. It representsabout 75% of the total absorbance in the region. This bandexhibits a regular P, Q and R structure with P(J), Q(J) andR(J) transitions split into several tetrahedral components.At room temperature, these multiplets are stronglyblended even if the resolution is limited to the Dopplerbroadening. Recently, in relation with the analysis of thespectra provided by SCIAMACHY, Frankenberg et al.measured the N2 pressure broadening and shift of 2n3

multiplets recorded by FTS at room temperature [21]. Intheir multispectrum treatment, the line positions of thecomponents were constrained to the values obtained by

ARTICLE IN PRESS

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–11401138

Margolis as given in HITRAN 2008 [8]. Similar and moreextensive measurements of pressure broadening andshifting parameters have also been performed in thesame region in the frame of GOSAT project [13].

Compared to spectra recorded at room temperature, ourrecordings at LNT benefit of the narrowing by a factor of twoof the Doppler width (HWHM 4.8�10�3 cm�1 at 81 K). Inaddition, the DFB laser diodes do not contribute to theobserved line profile which is an advantage for the lineintensity retrieval, compared to FTS spectra. We havecompared in details the structure of the 2n3 multiplets asobserved in our LNT spectra and as provided in the GOSATand HITRAN lists. Overall, GOSAT multiplets are found insignificantly better agreement with our observations thanHITRAN multiplets. As examples, Fig. 8 shows the comparison

Fig. 8. Comparison of the structure of the P(5), Q(9), Q(6) and R(4) multiplets o

GOSAT [12] line lists at 296 K. The RT and LNT relative intensities of a given m

state J value and are then unchanged by cooling.

for the P(5), Q(9), Q(6) and R(4) multiplets where importantdiscrepancies are noted, mainly with HITRAN.

The 2n3 multiplets as provided by GOSAT [12] are theresults of the line profile analysis of several spectracombined with a theoretical modelling while our line listwas constructed from the line profile fitting of our LNTspectrum only. This is why, in a few cases, we could notresolve in our LNT spectra, components calculated withcoinciding line centre in GOSAT list. In such problematic(and seldom) situations, the underlying theoretical modelis a clear advantage to constrain the line profile fitting.

Apart from some limited but significant deviations in theirstructure (see Fig. 8), a systematic wavenumber shift wasidentified for the 2n3 multiplets of the GOSAT list. Thedifferences between all the line centres in common in GOSAT

f the 2n3 band as observed at LNT and as provided in the HITRAN [8] and

ultiplet can directly be compared are they correspond to the same lower

ARTICLE IN PRESS

Fig. 9. Differences between GOSAT [12] and HITRAN [8] line positions

for methane between 5850 and 6185 cm�1. The values corresponding to

the strong multiplets of the 2n3 band have been highlighted.

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–1140 1139

and HITRAN lists (Fig. 9) have an average value of�0.33�10�3 cm�1 which is reasonable considering thatHITRAN line positions were used to calibrate GOSATspectra. Nevertheless, the values relative to the 2n3

multiplets (highlighted in Fig. 9) deviate significantly fromthis average value. Apart the dispersion due to theuncertainty of the line centre determinations (see above),the centres of GOSAT multiplets appear to be underestimatedby about 1.5�10�3 cm�1 on average. This observationprobably indicates that GOSAT list combined different FTSspectra to obtain the spectroscopic parameters of the strong2n3 multiplets and those of the weaker lines. Note thatGOSAT wavenumber calibration is based on HITRAN valuesand that our spectra were calibrated using GOSAT list(excluding the 2n3 multiplets). In other words, thecalibration of the different line lists are all based onMargolis line centre values which have not been confirmedby an independent study.

6. Conclusion

The positions and strengths at 81 K of 3251 transitionswere obtained from the analysis of the spectrum ofmethane recorded by direct absorption spectroscopybetween 5852 and 6181 cm�1. On the basis of linepositions coincidence, 2297 transitions were found incommon between our LNT line list and a very recent linelist at 296 K constructed in relation of the GOSAT project[12]. Empirical J values of the lower level of these 2297transitions could be determined from the variation of theline intensities between 81 and 296 K. These transitionsrepresent 99.1% and 90.8% of the absorbance at 81 and296 K, respectively. The line list provided as Supplemen-

tary Material allows accounting accurately for the tem-perature dependence of methane absorption at lowtemperature. It includes in particular 954 transitionswhich are absent in GOSAT line list (as they are veryweak at RT) but of significant intensities at LNT. Inaddition, we could determine 437 lower energy levels fortransitions which are provided without assignment inGOSAT line list.

The higher spectral resolution obtained by coolingdown to 81 K has allowed to evidence some inaccuraciesin the spectroscopic parameters of the 2n3 multiplets asprovided in the HITRAN database [8]. Compared toHITRAN line list based on Margolis results [9,10], GOSATline list at 296 K represents an important improvementboth in terms of sensitivity (the intensity cut off has beenlowered by two orders of magnitude) and in terms ofaccuracy. But the wavenumber calibration of the strong2n3 multiplets in GOSAT list is inconsistent with thecalibration of the weaker lines, GOSAT values beingunderestimated by 1.5�10�3 cm�1 on average.

The improvement of the knowledge of methaneabsorption at low temperature will be valuable for theanalysis of the spectra of Titan (atmospheric temperaturearound 90 K) provided by the Cassini–Huygens mission. Inabsence of a satisfactory theoretical model, the twotemperature methods applied to spectra recorded at 81and 296 K has proven to be an efficient way to obtain thetemperature dependence of methane absorption. We haveapplied this method in the high energy region of thetetradecad (this work) and in the icosad regions [20]. Inthese two high absorbing regions, methane spectra at 81 Kwere recorded by direct absorption spectroscopy but aprecise characterization of the methane spectrum in thelow opacity windows between the tetradecad and icosadregions requires higher sensitivity techniques. We haverecently coupled the cryogenic cell used in the presentwork with the CW-cavity ring down spectroscopy(CW-CRDS) technique in order to characterize the CH4

absorption spectrum at RT [22] and LNT [23] in the1.58mm transparency window. This transparency windowis of particular interest for the correct determination ofthe surface reflectivity of Titan (with implications on itscomposition). The sensitivity achieved by CW-CRDS at81 K (amin�3�10�10 cm�1) lowers by more than threeorders of magnitude that of the present recordings. It willallow for the detection of lines with intensity smaller than10�28 cm/molecule. By combining the results obtained bydirect absorption and CW-CRDS spectroscopies, we hopeto account for the temperature dependence of methaneabsorption over the whole 1.26–1.70mm range.

Acknowledgements

We thank A.V. Nikitin and V.I. Perevalov for commu-nicating the GOSAT line list of methane near 1.6mmbefore publication. This work is part of the ANR project‘‘CH4@Titan’’ (ref: BLAN08-2_321467). The support of theGroupement de Recherche International SAMIA bet-ween CNRS (France), RFBR (Russia) and CAS (China) isacknowledged.

ARTICLE IN PRESS

L. Wang et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 111 (2010) 1130–11401140

Appendix A. Supplementary material

Supplementary data associated with this article canbe found in the online version at doi:10.1016/j.jqsrt.2009.10.019.

References

[1] Buchwitz M, De BR, Noel S, Burrows JP, Bovensmann H, Schneising,O, et al. Atmospheric carbon gases retrieved from SCIAMACHY byWFM-DOAS: version 0.5 CO and CH4 and impact of calibra-tion improvements on CO2 retrieval. Atmos Chem Phys 2006;6:2727–51.

[2] Schneising O, Buchwitz M, Burrows JP, Bovensmann H, BergamaschiP, Peters W. Three years of greenhouse gas column-averaged dry airmole fractions retrieved from satellite—part 2: methane. AtmosChem Phys Dis 2008;8:8273–326.

[3] /http://gosat.nies.go.jpS.[4] Frankenberg C, Meirink JF, Bergamaschi P, Goede APH, Heimann M,

Korner, S, et al. Satellite chartography of atmospheric methanefrom SCIAMACHY on board ENVISAT: analysis of the years2003 and 2004. J Geophys Res 2006;111:D07303. doi:10.1029/2005JD006235.

[5] Bergamaschi P, Frankenberg C, Meirink JF, Krol M, Den-tener F,Wagner, T, et al. Satellite chartography of atmospheric methanefrom SCIAMACHY on board ENVISAT: 2. Evaluation based oninverse model simulations. J Geophys Res 2007;112:D02304.doi:10.1029/2006JD007268.

[6] Mondelain D, Payan S, Deng W, Camy-Peyret C, Hurtmans D, MantzAW. Measurement of the temperature dependence of line mixingand pressure broadening parameters between 296 and 90 K in then3 band of 12CH4 and their influence on atmospheric methaneretrievals. J Mol Spectrosc 2007;244:130–7.

[7] Predoi-Cross A, Brawley-Tremblay M, Brown LR, Malathy Devi V,Benner DC. Multispectrum analysis of 12CH4 from 4100 to 4635cm–1: II. Air-broadening coefficients (widths and shifts). J MolSpectrosc 2006;236:201–15.

[8] Rothman LS, Gordon IE, Barbe A, Benner DC, Bernath, PF, et al. TheHITRAN 2008 molecular spectroscopic database. JQSRT 2009;110:533–72.

[9] Margolis JS. Measured line positions and strengths of methane from5500 to 6180 cm�1. Appl Opt 1988;27:4038–51.

[10] Margolis JS. Empirical values of the ground-state energies formethane transitions between 5500 and 6150 cm�1. Appl Opt1990;29:2295–302.

[11] Gao B, Kassi S, Campargue A. Empirical low energy values formethane transitions in the 5852–6181 cm�1 region by absorptionspectroscopy at 81 K. J Mol Spectrosc 2009;253:55–63.

[12] Nikitin AV, Lyulin OM, Mikhailenko SN, Perevalov VI, Filippov NN,Grigoriev IM, et al. To be published.

[13] Lyulin OM, Nikitin AV, Perevalov VI, Morino I, Yokota T, Kumazawa,R, et al. Measurements of N2- and O2-broadening and shiftingparameters of methane spectral lines in the 5550–6236 cm–1

region. JQSRT 2009;110:654–68.[14] Nikitin AV, Mikhailenko S, Morino I, Yokota T, Kumazawa R,

Watanabe T. Isotopic substitution shifts in methane and vibrationalband assignment in the 5560–6200 cm–1 region. JQSRT 2009;110:964–73.

[15] Kassi S, Gao B, Romanini D, Campargue A. The near infrared(1.30–1.70mm) absorption spectrum of methane down to 77 K.Phys Chem Chem Phys 2008;10:4410–9.

[16] Menard-Bourcin F, Menard J, Boursier C. Temperature dependenceof rotational relaxation of methane in the 2n3 vibrational state byself- and nitrogen-collisions and comparison with line broadeningmeasurements. J Mol Spectrosc 2007;242:55–63.

[17] Pierre G, Hilico JC, Debergh C. The region of the 3n3 band ofmethane. J Mol Spectrosc 1980;82:379–93.

[18] Tsukamoto T, Sasada H. Extended assignments of the 3n1+n3 bandof methane. J Chem Phys 1995;102:5126–40.

[19] Varanasi P. Intensity measurements in the 2n3-band of 13CH4 at lowtemperatures. JQSRT 1981;25:319–23.

[20] Sciamma-O’Brien E, Kassi S, Gao B, Campargue A. Experimental lowenergy values of CH4 transitions near 1.33mm by absorptionspectroscopy at 81 K. JQSRT 2009;110:951–63.

[21] Frankenberg C, Warneke T, Butz A, Aben I, Hase F, Spietz, P, et al.Pressure broadening in the 2n3 band of methane and its implicationon atmospheric retrievals. Atmos Chem Phys 2008;8:5061–75.

[22] Liu AW, Kassi S, Campargue A. High sensitivity CW-cavity ringdown spectroscopy of CH4 in the 1.55mm transparency window.Chem Phys Lett 2007;447:16–20.

[23] Kassi S, Romanini D, Campargue A. Mode by Mode CW-CRDS at80 K: application to the 1.58mm transparency window of CH4.Chem Phys Lett 2009;477:17–21.