[step:Verify continuity, nonnegativity, and convexity of the integrand]
Define the map $L:[0,1]\times\mathbb R\times\mathbb R\to[0,\infty)$ by
\begin{align*}
L(x,z,\xi)=(z^3-x)^2|\xi|^6.
\end{align*}
The maps $(x,z,\xi)\mapsto z^3-x$, $t\mapsto t^2$, and $\xi\mapsto |\xi|^6$
are continuous, and products of continuous real-valued functions are continuous. Hence $L$ is continuous. Since both factors $(z^3-x)^2$ and $|\xi|^6$ are nonnegative, $L(x,z,\xi)\ge 0$ for every $(x,z,\xi)\in[0,1]\times\mathbb R\times\mathbb R$.
It remains to check convexity in $\xi$. Define the function $q:\mathbb R\to[0,\infty)$ by
\begin{align*}
q(\xi)=|\xi|^6=\xi^6.
\end{align*}
The function $q$
is $C^2$ on $\mathbb R$ and satisfies
\begin{align*}
q''(\xi)=30\xi^4\ge 0
\end{align*}
for every $\xi\in\mathbb R$. Therefore $q'$ is nondecreasing on $\mathbb R$. If $\xi_1,\xi_2\in\mathbb R$ with $\xi_1<\xi_2$ and $\lambda\in[0,1]$, applying the [Mean Value Theorem](/theorems/631) to $q$ on $[\xi_1,(1-\lambda)\xi_1+\lambda\xi_2]$ and on $[(1-\lambda)\xi_1+\lambda\xi_2,\xi_2]$, together with the monotonicity of $q'$, gives
\begin{align*}
q((1-\lambda)\xi_1+\lambda\xi_2)\le (1-\lambda)q(\xi_1)+\lambda q(\xi_2).
\end{align*}
Thus $q$ is convex. For fixed $(x,z)\in[0,1]\times\mathbb R$, the coefficient $(z^3-x)^2$ is nonnegative, so the map
\begin{align*}
\xi\mapsto L(x,z,\xi)=(z^3-x)^2q(\xi)
\end{align*}
is a nonnegative scalar multiple of a [convex function](/page/Convex%20Function). Hence $L$ is convex in the gradient variable $\xi$.
[/step]