Formation Energy is All You Need

A Pourbaix diagram maps which species of an element (metal, ion, oxide or hydroxide) is thermodynamically stable in water as a function of potential and pH, at fixed temperature, pressure and composition. Constructing one requires the Gibbs energies of formation of the participating phases and species. Obtaining reliable values, however, remains difficult: competing phases and species can differ by only a few kilojoules per mole, while published compilations may disagree and are not always freely reusable. Density functional theory (DFT) offers an alternative, but its errors can exceed the energy differences that determine phase stability. We introduce pourbase, an open, source-linked thermodynamic database with a Python package and desktop application for constructing Pourbaix diagrams. Individual reported values, their provenance and a configurable selection policy are accessible and configurable to users even while the database expands as agent-guided literature searches collect new data. We show that replacing a single formation energy can substantially change a predicted stability region and that established thermodynamic compilations can yield conflicting oxide stability boundaries. By making thermodynamic data accessible, traceable, and revisable, pourbase enables reproducible Pourbaix diagrams built directly from experimental evidence, as well as a platform for benchmarking computed values, as we demonstrate for the Materials Project, against experimental values. Ultimately, accurate phase stability predictions need not depend on increasingly complex models, but on getting the formation energies right.