There is a long standing and quite robust result in the literature, originating with Christophe Chamley and Ken Judd, that physical capital should not be taxed. Larry Jones, Rodolfo Manuelli and Peter Rossi extend this reasoning to human capital. These are very strong results that are not borne by the data, thus either the models are missing something, or economists still have a lot of convincing to do.
Christoph Braun challenges the last result on a technicality. Jones, Manuelli and Rossi assumed that human capital exhibits constant returns to scale in its production function, that is, a doubling of current human capital doubles ceteribus paribus future human capital. Braun take the opposite extreme: current human capital has no impact on future human capital, as the latter is only dependent on the time dedicated to education. The first assumption was very convenient, because it made human capital disappear from key equations, but this also drives the no-tax result. The latter makes it simpler than an intermediate assumption (decreasing returns to scale), but give a very different result from the original paper: in particular, the return to human capital does not vanish from the taxation equation and thus should be taxed. However, the accumulation of human capital is encouraged through the tax deductibility of tuition. And in the end, physical capital is still not taxed.