Acetone undergoes aldol self-condensation in the presence of acids or bases:
which makes drying agents such as 3A/4A molecular sieves, barium oxide, potassium hydroxide and carbonate (basic), copper(II) sulfate or aluminium(III) oxide (acidic) less effective and promotes formation of mesityl oxide to a significant degree. Boron(III) oxide is considered as one of the most appropriate desiccants for acetone [1, pp. 3967–3968]:
As with $\ce{Me2SO},$ the root of the difficulty is the acidic α protons, which in this case compounds the drying problem not only by inflating apparent water content by exchange process but also by providing a pathway to self-condensation through enol intermediates. This facet of acetone chemistry makes the choice of a successful desiccant a delicate process. As Table IV shows, mild siccatives such as calcium sulfate are inept; more potent desiccants such as molecular sieves exhibit a short initial drying action but thereafter actually cause disastrous increases in water content by displacement of the condensation equilibrium. This interpretation was confirmed for molecular sieves and other basic desiccants such as barium oxide by gas chromatographic analysis which demonstrated the presence of mesityl oxide in the dried solvent (see Table IV).
In summary, while both cupric sulfate and 3A molecular sieves are clearly at least useful preliminary desiccants, the agent par excellence for acetone is powdered boric anhydride.
However, $\ce{B2O3}$ is a Lewis acid itself and reacts with water to produce boric acid, which should build up upon drying the solvent. Why doesn't $\ce{B2O3}$ appear to be promoting aldol condensation, at least in the case of acetone?
Reference
- Burfield, D. R.; Smithers, R. H. Desiccant Efficiency in Solvent Drying. 3. Dipolar Aprotic Solvents. J. Org. Chem. 1978, 43 (20), 3966–3968. DOI: 10.1021/jo00414a038.
