The pure water vapour and water–nitrogen continuum absorption in the 1000 and 2500 cm−1 atmospheric windows has been studied using a 2 m base-length White-type multi-pass cell coupled with a BOMEM DA3-002 Fourier transform infrared spectrometer. The measurements were carried out at the National Institute of Standards and Technology (NIST, Gaithersburg, MD) over the course of several years (2004, 2006–2007, 2009). New data on the H2O:N2 continuum in the 1000 cm−1 window are presented and summarized along with the other experimental results and the continuum model. The experimental data reported on the water vapour continuum in these atmospheric windows basically agree with the most reliable laboratory data from the other sources. The MT_CKD (MlawerTobinCloughKneizysDavies) continuum model significantly departs from the experimental data in both windows. The deviation observed includes the continuum magnitude, spectral behaviour and temperature dependence. In the 2500 cm−1 region, the model does not allow for the nitrogen fundamental collision-induced absorption (CIA) band intensity enhancement caused by H2O:N2 collisions and underestimates the actual absorption by over two orders of magnitude. The water vapour continuum interpretation as a typical CIA spectrum is reviewed and discussed.
The water–water as well as the water–nitrogen continua are significant contributors to the total absorption in the atmospheric windows. Accurate knowledge of the continuum absorption is important for remote-sensing applications and for modelling the Earth's radiative balance [1–4]. Despite many years of experimental and theoretical investigation, it still remains poorly understood and its quantification also requires significant improvement. The continuum interpretations are controversial and involve such mechanisms as collision-induced absorption (CIA) [5,6], far line wings absorption [7,8] and water-dimer absorption [9,10].
In the most used expression, the continuum absorption coefficient in cm−1 can be written as where Tcont is the transmittance, L is a path length in centimetres, wH2O is the water vapour density in molecules cm−3, pH2O and PN2 are, respectively, water vapour and nitrogen partial pressures in atm at 296 K. The binary absorption coefficients Cs(ν,Θ) and Cf(ν,Θ) are water–water (self-broadened) and water–nitrogen (nitrogen-broadened) continua in cm−1(atm×molecule cm−3)−1 and are dependent on wavenumber and temperature. We have recently reported measurements of the water vapour self-continuum Cs in the 1000 and 2500 cm−1 atmospheric windows [11,12], as well as measurements of the water–nitrogen continuum Cf in the 2500 cm−1 window . The data were obtained in the laboratory for high-resolution Fourier transform infrared (FTIR) spectroscopy at the National Institute of Standards and Technology (NIST, Gaithersburg, MD). These papers contain the detailed description of both the experimental and the digital data-processing procedures. The results reported basically agree with the most reliable experimental data from the other sources. Nevertheless, significant deviations between the laboratory data and the MT_CKD1 (Mlawer–Tobin–Clough–KneizysDavies) continuum model , widely used in atmospheric applications, were found and discussed. The deviations exist not only in the continuum absorption magnitude, but also in its spectral behaviour and temperature dependence, especially in the 2500 cm−1 atmospheric window. The H2O:N2 binary component of the nitrogen fundamental CIA band experimentally detected for the first time in Baranov  and predicted in the theoretical calculations of Brown & Tipping  accounts for the roughly two orders of magnitude deviation between our measurements and the model. Such a large difference is understandable and even expected because the MT_CKD model does not allow for the N2 fundamental band intensity enhancement resulting from interactions between water and nitrogen molecules.
The water–nitrogen continuum in the 10 μm window was the subject of many laboratory studies over the past several decades. It is much weaker than the water–water continuum and it is difficult to measure accurately. As was noted by Cormier et al. , the ratio γ=Cs/Cf, measured in different laboratories, varies approximately from 100 to 1000, and significantly deviates from the value γ≈2500 provided by the continuum model.
2. Water–nitrogen continuum absorption in the 1000 cm−1 atmospheric window
The experimental set-up used in our investigation includes the 2 m base-length White-type multi-pass cell with a maximum path length of about 100 m. The cell is coupled to a BOMEM DA3-002 FTIR spectrometer . With the BaF2 windows installed in the cell, the maximum total gas pressure cannot be higher than 506 kPa (5 atm). The water vapour partial pressures may exceed 27 kPa (200 torr) depending on the temperature of the cell. Under the typical experimental conditions of Θ=326 K, pH2O∼12 kPa (90 torr), the transmittance is reduced by approximately 2 per cent after adding into the cell 405 kPa (4 atm) of pure nitrogen. Measurements of such weak absorption cannot provide very accurate results; however, they may be interesting and helpful because of the above mentioned strong disagreements in the data from different sources.
Sixteen spectra recorded at a resolution of 0.1 cm−1 with an MCT detector over the spectral range 800–3500 cm−1 were used to experimentally estimate the water–nitrogen continuum at a temperature of 326 K. In contrast to the experimental procedure described in Baranov , a base line for these spectra was taken with the cell filled with pure nitrogen at the same pressure as in a sample. Figure 1a shows the continuum absorption coefficients in the microwindow at 943 cm−1, and in the middle of the window at 2475 cm−1 (figure 1b).
The data are plotted against the water vapour partial pressure in order to show clearly their scatter and the absence of a trend with the pressure. To determine these absorption coefficients Cf, we used the values Cs, reported in Baranov et al.  and Baranov & Lafferty . The data in figure 1a were obtained after spectral processing and smoothing of initial spectra according to the description given in Baranov et al. . As can be seen from the figure, the averaged values of the water–nitrogen continuum are not very accurate; their standard deviations (s.d.) exceed 50 and 20 per cent in the 1000 and 2500 cm−1 windows, respectively. It is important to note here that the value Cf(2475)=(4.5±1.1)×10−25 cm−1(atm×molecule cm−3)−1 agrees reasonably with the more precise value (3.9±0.6)×10−25 cm−1(atm×molecule cm−3)−1 reported in Baranov . Very good ‘base line’ stability in our measurements  allows us to conclude the absence of significant systematic error in reported data.
Figure 2 shows the spectral behaviour of the water–nitrogen continuum in the 1000 cm−1 window and compares our results2 with the models and with the data from the other sources. The roughness (noise) of the continuum curve obtained is not larger than the statistical error (1 s.d.) and seems to originate from the non-ideality of the instrument and the data-processing method. Note that we obtain a very small increase in absorption after adding nitrogen to a pure water vapour sample. In the majority of laboratory measurements of the water–nitrogen continuum in the 1000 cm−1 window, a tunable CO2-laser was used as a radiation source, and the P(20) transition at 944.2 cm−1 was the most popular in those investigations. It can be seen from the figure that the measurements of Nordstrom et al. , Loper et al.  and Cormier et al.  agree with each other and provide the foreign continuum value of about 6×10−25 cm−1(atm×molecule cm−3)−1, which is two times larger than our result. In contrast, the data of Hinderling et al. , obtained with a CO2-laser and a spectrophone, are in good agreement with our measurements. The data of Peterson et al.  are not easy to comment upon. These authors reported measurements with both a long path cell and spectrophone techniques at 26 CO2-laser transitions. Some of the results seem to have significant local line contribution to the absorption, as it is in the case of measurement on the R(20) 975.93 and R(14) 971.93 cm−1 transitions (see table 1 from Peterson et al. ). In some other cases, the water vapour continuum absorption coefficients obtained with the opto-acoustic detector significantly deviate from measurements with the multi-pass cell. To analyse the data, we have selected 11 laser transitions located far from water vapour absorption lines, and which were used in both sets of measurements. First, it is interesting to note significant disagreement between water–nitrogen continuum absorption coefficients, obtained for these transitions with a spectrophone and with a multi-pass cell technique. The data for Cf deviate on average by about 50 per cent. Only six measurements with a multi-pass cell, shown in figure 2 by black squares, fall lower than the value of 1×10−24 cm−1(atm×molecule cm−3)−1. Those values are scattered between 4×10−25 and 8.5×10−25 cm−1(atm × molecule cm−3)−1. The data obtained with a spectrophone (grey squares) fall systematically lower than and scatter within 2.2×10−26 and 7.3×10−25 cm−1(atm×molecule m3)−1. We presume that the uncertainty of the data by Peterson et al.  is comparable to, or even larger than, estimated errors in the data from this study.
Our data appear to be in very good agreement with the data of Burch & Alt . We also note that the statistical errors of the data are quite comparable. The MT_CKD continuum model in the middle of the window goes through the lower border of scatter of the data by Peterson et al. . It agrees with our data here within experimental error, but does not reproduce correctly the continuum spectral behaviour. It underestimates the water–nitrogen continuum at the window boundaries by roughly twofold. The empirical continuum model developed by Aref'ev  provides estimates that are significantly higher than our data, and fall on the upper border of the experimental data scatter. We also note the inadequate spectral behaviour of this model in the high-frequency part of the window.
The water–nitrogen continuum temperature dependence remains an open question. Our quick attempt to measure the continuum at the higher temperature of 339 K did not lead us to a reliable conclusion on its temperature dependence because of the large experimental uncertainties of the data. Recently in Baranov , we reported the lack of a significant temperature dependence of the water–nitrogen continuum in the 2500 cm−1 spectral region at temperatures from 326 to 363 K. Analysis of the data by Loper et al.  also shows the impossibility of detecting systematic and reliable changes of the continuum at temperatures of 300 and 283 K. The empirical model by Aref'ev  does not take into account the temperature dependence of the water–nitrogen continuum. The MT_CKD model predicts very weak temperature dependence of the absorption coefficient of the order of 0.05%K−1, but, in the scale used in figure 2, it is not possible to see deviations in the plotted curves at 326 and 296 K. The good agreement between our results at 326 K and the data of Burch & Alt  at 296 K also indicates the absence of strong temperature dependence of the H2O:N2 continuum. The measurements by Cormier et al.  employing an advanced high-sensitivity cavity ring-down spectroscopy (CRDS) technique with a CO2-laser operating at P(20) 944.2 cm−1 transition provide the result dCf/dΘ=−0.9%K−1, which is about 20 times larger than the prediction of the model. In Baranov et al. , the suggestion was made that the data of Cormier et al.  may underestimate the self-continuum by about 15 per cent. Note that in experiments with H2O:N2 mixtures only, when both continuum absorption coefficients Cs and Cf are derived from the same set of measurements, a systematic error in the absorption coefficient Cs may lead to an opposite systematic error in the absorption coefficient Cf. Perhaps this is why the value Cf(944.2 cm−1)=5.8×10−25 cm−1(atm × molecule cm−3)−1 3 is about two times larger than our data and the data of Hinderling et al. . Despite this disagreement in the continuum values, the relative continuum temperature trend from Cormier et al.  should be more exact. Perhaps a more reliable verdict on the water–nitrogen continuum temperature dependence in the 1000 cm−1 window could be given on the basis of measurements over a much larger temperature range than that shown in fig. 7 of Cormier et al. , because almost all the data points overlap inside their error bars.
3. Summary of measurements of the water–water and water–nitrogen continuum in the 1000 and 2500 cm−1 atmospheric windows
Figures 3 and 4 summarize our results on the water–water and water–nitrogen continuum in the 1000 and 2500 cm−1 atmospheric windows. The small black dots show the spectral behaviour of the continuum, while the lesser number of white circles shows the statistical uncertainty of the data at selected intervals. It is important to note also that, for reasons of consistency, we show our results on the self-continuum in the 2500 cm−1 region in figure 3 with error bars equal to 1 s.d., not 2, as this was reported in our original paper . The data are compared with experimental results from other laboratories and with the continuum model.
It can be seen from figure 4 that our results on the water–nitrogen continuum agree very well with the data of Burch & Alt  in both the 1000 and 2500 cm−1 regions, although the measurements were conducted at different temperatures. The Burch and Gryvnak  measurements of the self-continuum in the 1000 cm−1 window also agree with our results, extrapolated to temperatures in these measurements (see figs 8 and 9 from Baranov et al. ). However, in contrast to that, the Burch & Alt  self-continuum in the 2500 cm−1 region (figure 3) appears to be about a factor of three lower than our experimental results at near room temperature and about fourfold different at 326 K. Note that the data of Watkins et al.  (grey diamonds) and the most recent high-resolution laboratory data of Ptashnik et al.  (grey circles with error bars) tend to support our measurements. For the sake of clarity, we reproduce here only the limited number of the data points from table 3 of Ptashnik et al. . The fact that there is strong disagreement between Burch's and the other measurements of the self-continuum magnitude and temperature dependence around 2500 cm−1 remains unexplainable and requires additional experimental validation. The use of the CRDS technique  combined with a tunable DF-laser as a radiation source  seems to be a very promising method for the data validation.
Figures 3 and 4 demonstrate the existence of significant disagreement between the experimental results and the MT_CKD model in both windows and for both the water–water and the water–nitrogen continua. The disagreement concerns the continuum magnitude, spectral behaviour and temperature dependence. The most significant deviation of about two orders of magnitude appears in the middle of the 2500 cm−1 window for the water–nitrogen continuum. It results from the presence of the nitrogen fundamental band component induced by collisions between water and nitrogen molecules . Evidently, the MT_CKD model does not allow for this type of induced absorption, which has been predicted in the theoretical calculations of Brown & Tipping . The recent measurements of Ptashnik et al.  (thick solid grey line) agree well with our results in magnitude, although some difference in the continuum spectral behaviour exists.
4. On the continuum interpretation
In the great majority of publications, the continuum is interpreted as the cumulative absorption of far wings of lines composing H2O pure rotational and ro-vibrational bands, or as a superposition of broad absorption bands of water vapour dimer. But we would like to draw the reader's attention to the other point of view on the continuum origin.
In 2001, Bauer & Godon  proposed that CIA processes may be involved in the continuum formation. The authors have shown that binary absorption coefficients BX−X of several dipole-less particles X (X=Ar, CH4, N2, C2H6, C2H4 and CO2) measured in the millimetre wave spectral region strongly correlate with the binary absorption coefficients BX−H2O measured in X–H2O mixtures. Absorption in pure X gases is caused by roto-translational (translational for Ar) collision-induced bands . Absorption in X–H2O mixtures (after subtracting the pure H2O and pure X-gas contributions) is a ‘buffer-gas-broadened’ continuum (like the water–nitrogen continuum Cf). Therefore, the observed correlation leads to the assumption that ‘buffer-gas-broadened’ H2O continua at least partially are also ‘collision-induced spectra’. Later, Clough et al.  suggested a dual origination of the continuum as CIA and far line wings absorption, and implemented this idea in MT_CKD_1.0 and later versions of the model.
It is interesting to note that the self- and water–nitrogen continuum profiles are shaped like typical CIA spectra (CIA/dimer in the case of CO2) of the other molecules. This is illustrated in figure 5, which shows the continuum profiles (plotted on the linear Y -scale) together with the CIA/dimer spectra of carbon dioxide (except for the allowed ν2 fundamental band) [29,30] and the CIA spectra of nitrogen [31,32]. We assume that the compared spectra have the same physical origin. As seen from the figure, the self-induced (figure 5a) and nitrogen-induced (figure 5c) water vapour continua consist of two broad features, which can be associated with roto-translational and fundamental collision-induced bands. Carbon dioxide CIA/dimer spectra (figure 5b) also demonstrate strong negative temperature dependence of the intensity, as does the pure water vapour continuum. On the contrary, both the water–nitrogen continuum and pure nitrogen CIA band intensity have much weaker temperature dependence [32,33]. A huge difference in the magnitudes of the binary absorption coefficient of the continuum and other CIA spectra may arise from different properties of intermolecular potential surfaces and different mechanisms of collisional induction, which are dipole–dipole for water vapour, dipole–quadrupole for a H2O+N2 mixture and quadrupole–quadrupole for pure CO2 and N2. Other processes, such as breaking or acceleration of rotational motion in colliding molecules (rotational relaxation), have also been taken into account. For a dipolar molecule, such as H2O, those processes may lead to enforced acts of absorption and/or emission.
CIA bands of dipole-less molecules are well known and quite well studied experimentally (see Frommhold ). The fact that molecular collisions produce absorption in regions of forbidden transitions is very important. It follows from this fact that this induced absorption exists as an excessive natural component of any molecular absorption band profile, and this component should not be attributed to allowed transitions. Note also that any theoretical approach for molecular absorption band shape, based on the analysis of properties of an autocorrelation function of dipole moment , should provide a whole absorption spectrum, including its CIA ‘sub-band’.
A possible role of the stable water vapour dimer in the continuum formation has to be expected similar to that observed in carbon dioxide [30,35]. The right-hand side of figure 5b shows that the carbon dioxide spectrum profile in the region of the CO2 Fermi dyad (ν1, 2ν2) is a superposition of the stable dimer features ‘sitting’ atop much broader ‘pedestals’, which are the CIA bands. The recent experimental investigations of the continuum by Paynter et al.  demonstrate that the observed continuum absorption profiles within the water vapour ro-vibrational bands in the mid- and near-infrared regions also can be interpreted as some ‘pedestals’ superimposed by quite distinct features attributed in Paynter et al.  presumably to ‘stable and metastable’ water vapour dimers (see fig. 6 in Paynter et al. ). Within those bands, the water dimer contribution to the continuum may be significant. However, in atmospheric windows far away from dimer band centres this contribution should be expected to be negligible in comparison with absorption by far wings of water vapour CIA bands.
The experimental data presented here and those recently reported [11–13] on the water vapour continuum in the 1000 and 2500 cm−1 atmospheric windows basically agree with the most reliable laboratory data from the other sources. The exception is the water vapour self-continuum in the 4 μm spectral region, where our results significantly deviate from the measurements of Burch & Alt . Additional independent experimental investigations are necessary for the final validation of the 2500 cm−1 self-continuum at near-room and lower temperatures.
The MT_CKD continuum model significantly deviates from the experimental data in both the 10 and 4 μm atmospheric windows. The observed deviation includes the continuum magnitude, spectral behaviour as well as the temperature dependence. In the 4 μm region, the model does not allow for the nitrogen fundamental CIA band intensity enhancement caused by H2O:N2 collisions and strongly underestimates the observed absorption. Note that the similar case of an absorption enhancement was described recently in Baranov et al. , when adding CO2 to a sample of pure O2 significantly increases absorption in the region of the oxygen fundamental CIA band.
Avoiding discussion on the use of different terms, such as CIA, ‘interaction-induced absorption’  or ‘bimolecular absorption’ , we note rather the physical similarity of the spectral features illustrated in figure 5. The water vapour continuum absorption in the atmospheric windows is certainly the absorption by far wings of these features. But it is hard to believe that far wings of water dimer bands, presumably detected in recent works by Paynter et al.  and Ptashnik et al. , may be predominant in comparison with far wings of water vapour CIA bands.
The authors acknowledge support from the Upper Atmospheric Research Programme of NASA. We thank also the anonymous referees for many useful suggestions. Yu.B. would like to thank the NIST Physics Laboratory Management for the opportunity to carry out the reported research. He also acknowledges partial financial support from the Russian Foundation for Basic Research through grant 10-05-93105.
One contribution of 17 to a Theo Murphy Meeting Issue ‘Water in the gas phase’.
↵1 Here and below we mean the MT_CKD_2.5 version of the continuum model.
↵2 The data presented here are available in digital form from the author Yu.B.
- This journal is © 2012 The Royal Society