The recent literature on direct synthesis of H2O2 is reviewed with respect to two important directions. The best catalysts to date are assessed with a view to their composition, e.g., alloys, and regarding the role of the support material and solvent used. Moreover, a number of novel reactor concepts that have been proposed in literature for direct synthesis of H2O2 are analyzed with a focus on their potential for practical implementation. These include: (i) microchannel process technology for safe mixing and conversion of the reactants in undiluted gaseous state, (ii) dense palladium-based membranes acting as a means to prevent gas-phase contact of the reactants while providing a catalytically active interface, (iii) asymmetrically porous catalytic membranes operated in wetted state as an interphase contactor where one reactant is introduced in a controlled way from the gas phase into the liquid reaction medium, (iv) microreactors for multiphase operation with improved mass transfer efficiency compared to conventional reactor technologies, and (v) electrocatalytic synthesis, both in electrolysis mode and in fuel cell mode, which also promise improved safety by avoiding gas-phase mixing of hydrogen and oxygen and increased selectivity due to improved reaction control. Concerning the first aim, an acid-pretreated 5 wt.% Pd/C catalyst appears to be the best choice for methanol as solvent whereas a Pd-Au exchanged Cs-containing heteropolyacid catalyst is the current benchmark for water as solvent. With proper acid pre-treatment of the Pd/C catalyst or acid function incorporated into the support (heteropoly acids), no halides and no inorganic acids are needed in the solvent to prevent hydrogen peroxide decomposition. However, the H2O2 hydrogenation activity in pure water is still rather high which limits the selectivity. Among the different reactor concepts, microchannel process technology seems to offer key advantages in terms of high productivity, selectivity and process safety. The electrocatalytic approach also looks promising. However, major achievements are still needed, e.g., successful scale-up of these concepts to relevant throughputs confirming the good performance obtained in lab scale. (C) 2014 Elsevier B.V. All rights reserved.