How can I derive Hicksian demand, when I get from FOC only px/py=1/3.

The problem is to minimize pxx + pyy s.t. x+3y>U

If I plug the price relation into the budget constraint I=pxx+pyy I get the income in the demand function, but then this is Marshallian demand. Plugging the relation in expenditure function, obtained from the indirect utility function also doesn't lead to the Hicksian demand(, that I obtained via Shephard's lemma and equals hx=U+x+3y).