While thousands of exoplanets have been confirmed, the known properties about individual discoveries remain sparse and depend on detection technique. To utilize more than a small section of the exoplanet dataset, tools need to be developed to estimate missing values based on the known measurements. Here, we demonstrate the use of a neural network that models the density of planets in a space of six properties that is then used to impute a probability distribution for missing values. Our results focus on planetary mass which neither the radial velocity nor transit techniques for planet identification can provide alone. The neural network can impute mass across the four orders of magnitude in the exoplanet archive, and return a distribution of masses for each planet that can inform about trends in the underlying dataset. The average error on this mass estimate from a radial velocity detection is a factor of 1.5 of the observed value, and 2.7 for a transit observation. The mass of Proxima Centauri b found by this method is $1.6^{\rm +0.46}_{\rm -0.36}$ M$_\oplus$, where the upper and lower bounds are derived from the root mean square deviation from the log mass probability distribution. The network can similarly impute the other potentially missing properties, and we use this to predict planet radius for radial velocity measurements, with an average error of a factor 1.4 of the observed value. The ability of neural networks to search for patterns in multidimensional data means that such techniques have the potential to greatly expand the use of the exoplanet catalogue.

The hydrogen isotopic (D/H) ratio reflects the global cycling and evolution of water on Earth as it fractionates through planetary processes. We model the water cycle taking seafloor hydrothermal alteration, chemical alteration of continental crust, slab subduction, hydrogen escape from the early Earth, and degassing at mid-ocean ridges, hot spots, and arcs into account. The differences in D/H ratios between present-day oceans, oceanic and continental crust, and mantle are thought to reflect isotopic fractionation through seafloor alteration, chemical alteration, and slab dehydration. However, if the speed of plate tectonics has been nearly constant through out Earth's history, the degassing and regassing rates are too small to reach the present-day D/H ratios. We show that (a) hydrogen escape from reduced early atmosphere, (b) secular net regassing, or (c) faster plate tectonics on early Earth is needed to reproduce the present-day D/H ratios of the water reservoirs. The low D/H ratio of Archean seawater at 3.8 Ga has previously been interpreted as a signature of (a) hydrogen escape, but we find it can also be explained either by (b) secular net regassing or by (c) faster plate tectonics on early Earth. The rates of hydrogen escape from early Earth and secular regassing on present-day Earth are constrained to be lower than 2.1 x 10^11 kg/yr and 3.9 x 10^11 kg/yr. Consequently, the volume of water in the present-day mantle could result entirely from the regassing through Earth's history. In that case, the volume of initial oceans could be 2 to 3 times larger than that of current Earth. We suggest that, in addition to the D/H ratio of Archean seawater, identifying the D/H ratios of both seawater and mantle throughout Earth's history would allow to distinguish these evolutionary scenarios.

We have found many Earth-sized worlds but we have no way of determining if their surfaces are Earth-like. This makes it impossible to quantitatively compare habitability, and pretending we can risks damaging the field.