Eliminating Luck and Chance in the Reactivation Process: A Systematic and Quantitative Study of the Thermal Reactivation of Activated Carbons

Increasing environmental concerns, stricter legal requirements, and a wide range of industrial applications have led to growing demand for activated carbon worldwide. The energy-intensive production of fresh activated carbon causes significant CO₂ emissions and contributes to global competition for renewable carbon-based raw materials. Although (thermal) reactivation of spent activated carbon can drastically reduce the demand for fresh material, the reactivation process itself is still mostly based on experience and empirical knowledge locked into activated carbon companies. Despite the vast number of papers published in the field, practically relevant, systematic, and quantitative knowledge on the thermal reactivation process is barely available. This paper presents a simple and robust methodology for the development of a predictive model for the production of reactivated carbon with a defined product quality under energetically optimized conditions. An exhausted activated carbon sample was subjected to 26 reactivation experiments in a specially designed laboratory rotary kiln, whereas the experiments were planned and evaluated with statistical design of experiments. The influence of the reactivation conditions (heating rate, heating time, H₂O/N₂ volume ratio, and CO₂/N₂ volume ratio) on the specific surface area, energy consumption, yield, and adsorption capacity for diatrizoic acid were evaluated. The BET surface of the reactivated carbons ranged between 590 m²/g and 769 m²/g, whereas the respective fresh carbon had a BET surface of 843 m²/g. The adsorption capacity for diatrizoic acid measured as the maximum solid phase concentration q_m derived from the Langmuir equation varied between 24.4 g/kg and 69.7 g/kg (fresh carbon: 59.6 g/kg). It was possible to describe the dependency of the quality criteria on different reactivation parameters using mathematical expressions, whereas the response surface methodology with nonlinear regression was applied to build the models. A reactivation experiment under statistically optimized conditions resulted in energy savings up to 65%, whereas the properties of the reactivated sample were close to the predicted values.