Validation tests of the W2020 energy levels of water vapor

A decade ago, a task group of the International Union of Pure and Applied Chemistry performed an ex-haustive collection and review of measured transitions, applied the MARVEL procedure, and derived rec-ommended empirical energy levels for nine major water isotopologues. Very recently, using an improved methodology, the sets of empirical energy levels of H-2 O-16, H-2 O-18 and H-2 O-17 were updated, leading to the so-called W2020 energy levels and transition wavenumbers [Furtenbacher et al. J. Phys. Chem. Ref. Data 49 (2020) 043103; 10.1063/5.0030680]. Here we present validation tests of the W2020 line list of H-2 O-16 against spectra recorded by cavity ring down spectroscopy (CRDS) referenced to a frequency comb (FC), newly obtained in the 8040-8630 cm(-1) region. The recorded spectra are found in excellent agreement with previous high-quality studies available in the literature. While these literature sources were all incorporated in the transition database used to derive the W2020 energy levels, the direct superposition of the FC-CRDS spectra to the W2020 line list of H-2 O-16 shows a number of large disagreements. Cases where deviations largely exceed the W2020 claimed uncertainty on the transition frequencies are noted. In the considered spectral region, the resulting W2020 list is thus less accurate than some of the published original sources used to derive the W2020 energy levels. We conclude that the sophisticated global procedure and algorithm elaborated to identify and adequately weight inaccurate line positions among the large W2020 transition database do not always prevent less accurate data from "spoiling" higher quality data sources. The W2020 list of H-2 O-16 is also compared to newly recorded CRDS spectra in the 12970-13200 cm(-1) region (corresponding to the region of the A-band of O-2 ), where previous observations were very scarce. As in the previous region, substantial position deviations are evidenced, and in many cases, the W2020 error bars appear to be strongly underestimated. (c) 2021 Elsevier Ltd. All rights reserved.