Ortho- to para-Hydrogen Conversion Kinetics: New Experimental Data and an Industrial Perspective

Hydrogen is a remarkable molecule with applications ranging from refining and petrochemicals to medicine, to space travel and the energy transition. It consists of two spin isomers, namely orthohydrogen (ortho-H₂) and parahydrogen (para-H₂), that are separated in energy by only 1.455 kJ/mol. Chemically, these isomers are indistinguishable, yet each isomer has its own unique physical properties including thermal conductivity, optical behavior, and specific heat capacity. These physical traits are important during hydrogen liquefaction because the para-H₂ form is more stable at cryogenic temperatures (i.e., T < 77 K). In the context of energy transition, the production and supply of liquid hydrogen requires the application of a catalyst in liquefaction plants to induce hydrogen isomer interconversion via the process known colloquially as “spin-flipping.” The same catalyst can also improve brightness in neutron spallation sources, or enable the parahydrogen induced hyperpolarization (PHIP) technique used in magnetic imaging. Although the current preferred catalyst for these applications is a class of iron oxide materials, only a limited set of catalyst performance data is available. A confluence of this data was measured more than half a century ago, often using catalyst samples synthesized at lab-scale, and/or derived under reaction conditions irrelevant for practical application. Consequently, the few widely cited kinetic models for this system were developed by fitting data against an even smaller subset of this data limited by temperature or pressure conditions, all from the 1950s/1960s. It is reasonable to question if these models have predictive capabilities that are relevant for the current-day design using today’s commercial catalyst. Our work compares three of these kinetic models against new Ionex Type O–P Catalyst conversion data, spanning >300 data points across a broad and industrially relevant temperature–pressure window, and four published experimental data sets. The new data were measured in a custom-built cryostat, as part of a joint program between Quantum Technology Corporation and Shell. The data were analyzed using a custom Python script. A Langmuir–Hinshelwood model used by Donaubauer, and a similar model by Zhuzhgov/Buyanov showed nonphysical behavior under certain experimental conditions. These models are therefore unsuitable for a design of any real-world application. However, our work finds that three of the literature kinetic models do perform reasonably well in predicting the para-H₂ outlet concentration of the experiments conducted in this study, provided that the reaction system remains in the gas phase. The two best performing models are the first-order model by Donaubauer et al. (R² = 0.98), and the model by Wilhelmsen et al. if the model parameters are modified. Thus, our conclusion is that these models do offer relevant predictive capabilities for current-day process design, and using a commercial catalyst.