We know that an analogue of the Hasse-Minkowski theorem does not hold for all cubic forms, e.g. because Selmer's cubic: $$ 3x^3 + 4y^3 + 5z^3 = 0 $$ has solutions over $\mathbb{R}$ and $\mathbb{Q}_p$ for all $p$, but no solutions over $\mathbb{Q}$.
My questions are:
- Can we find a (non-trivial) class of cubic forms where an analogue of the Hasse-Minkowski theorem does hold?
- Is there any intuition for why the local-global principle holds for quadratic but fails for cubic forms?
- Are there higher degree forms where the local-global principle holds again?
- Are questions like these addressed anywhere in the literature?
Many thanks.
By a theorem of Hilbert, Hasse principle holds for the affine cubic form $x\mapsto N(x)-a$ ( a rational number is globally a norm of a cyclic extension if and only if it is locally)
Note this is true if you replace $3$ by any integer, provided you restrict yourself to cyclic extensions. You can also replace $\mathbb{Q}$ ny a global field.
You can find a nice introduction to the definition here: http://homepages.warwick.ac.uk/staff/S.Siksek/arith/notes/brauermanin.pdf
For example, Bayer, Lee and Parimala computed the Brauer Manin obstruction for multinorm equations, and I'm pretty sure that we can use their work to construct counterexamples of arbitrary large degree. The full paper is available on Parimala's web page.