Understanding JWST water spectra: What can thermochemical models tell us about the (cold) water in protoplanetary disks?

¹ Leiden Observatory, Leiden University, 2300 RA Leiden, The Netherlands
² Max-Planck Institut für Extraterrestrische Physik (MPE), Giessenbachstr. 1, 85748 Garching, Germany

^★ Corresponding author: vlasblom@strw.leidenuniv.nl

Received: 4 June 2025
Accepted: 28 August 2025

Abstract

Context. Rotational H₂O spectra as observed with JWST/MIRI trace a wide range of excitation conditions and, thereby, provide a good probe of the temperature and column density structure of the inner disk. H₂O emission can also be influenced by dynamical processes in the disk. In particular, dust grains can drift inward and their icy mantles sublimate once they cross the snow lines, thus enriching the inner regions in, for instance, H₂O vapor. Recent work has found that this process may leave an imprint in the H₂O spectrum in the form of excess flux in the cold, low-E_up H₂O lines.

Aims. To interpret JWST spectra, local thermodynamic equilibrium (LTE) slab models are commonly used to determine the temperature, column density, and emitting region that is traced by the observed emission. In this work, we aim to test the accuracy of several common retrieval techniques on full 2D thermochemical disk models, to derive the underlying 2D distribution. Moreover, we investigate the cold H₂O emission that has been proposed as a signature of drift, to gain further insights into the underlying radial and vertical distribution of H₂O.

Methods. We present two sets of Dust And LInes (DALI) thermochemical models, one in which the abundances are set by the chemical network, and the other in which the abundances are parameterized. We ran several commonly used retrieval techniques on the generated synthetic spectra and investigated how the retrieved temperature and column density compare to our models.

Results. Single-temperature slab retrievals mainly trace the warm (~500 K) H₂O reservoir, whereas a three-component fit is able to better trace the full temperature gradient in the IR emitting region. Retrieved temperatures tend to underestimate the true temperature of the emitting layer due to non-LTE effects such as sub-thermal excitation. The retrieved column density traces close to the mid-IR dust τ = 1 surface. We arrive at the same conclusion when performing this analysis for CO₂ emission and find that ¹³CO₂ emission retrieves a lower temperature than ¹²CO₂ due to it tracing deeper into the disk. Additionally, we find that our fiducial parameterized model predicts a very strong flux in the cold H₂O lines, but only when the H₂O abundance in the upper layers is high. The fiducial model with the full chemistry, by contrast, does not.

Conclusions. We find that the strength of the cold H₂O emission is directly linked to the H₂O abundance above the snow surface at large radii (>1 au). This implies that sources in which the excess cold H₂O flux is detected likely have a high H₂O abundance in this region (≳10⁻⁵) – higher than what is predicted by the chemical network. This discrepancy is most likely caused by the absence of dust transport processes in our models, further strengthening the theory that this emission may be a signature of radial drift and vertical mixing.