Set $\phi (x) = u(x)+iv(x)$, $x=(x_1,\ldots,x_N)$, a $T$-periodic function in $H^1_\text{loc}(\mathbb{R}^N)$, that is $\phi (x) = \phi (x_1 + T ,\ldots, X_N + T)$ for all $x$ and where $u = \operatorname{Re} \phi$ and $v = \operatorname{Im} \phi$. I wonder if $$ \int_{[0,T]^N}\frac{u_{x_1}^2}{2} + \frac{v_{x_1}}{\sqrt{2}}(u-1)^2\ dx > 0$$ for functions satisfying $$ \int_{[0,T]^N} v_{x_1}(u-1)\ dx = \varepsilon > 0$$ for some small and fixed $\varepsilon > 0$. Any idea or help to show this would be appreciate! I am stucked. Thank you in advance.

EDIT: maybe proving $$ \int_{[0,T]^N}\frac{|\nabla u|^2}{2} + \frac{v_{x_1}}{\sqrt{2}}(u-1)^2\ dx > 0$$ under the same constraint is easier?


1 Answer 1


Take $N=1$, $T=1$, and let $u=1/2$. Let $v(x) =1/2 - x$ for $x\in \mathbb{Z}+[h,1-h]$ for some small positive number $h$. By taking $h$ small enough, you can obviously make sure

$$ \int_{[0,1]} v'(u-1) = \int_h^{1-h} (-1)(1/2-1) = \dfrac{1}{2}(1-2h) $$

is positive. However,

$$ \int_{[0,1]} v'(u-1)^2 = \int_h^{1-h}(-1)(1/2-1)^2 = -\dfrac{1}{4}(1-2h) $$

is then negative (and equal to $\sqrt{2}$ times your first integral).

You definitely seem to need additional hypotheses on $u$ and $v$ to have chance of anything like what you propose being true (perhaps you missed some?)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.