I'm reading these Cornell lectures on solitons (link doesn't work right now, but it just worked yesterday), and I can't seem to prove what I thought would be a simple analysis exercise.
Namely, suppose you have the Lagrangian density
$$ \mathcal{L}= \frac{1}{2}(\partial_\mu \phi)^2 - U(\phi) $$
in 1+1 spacetime dimensions with $g_{\mu\nu} = (1,-1)$, and where $\phi$ is a real scalar field. In the highlighted text attached, the author claims that (i'm paraphrasing)
In order for the integral $$ E[\phi] = \int_{-\infty}^\infty \frac{1}{2}\phi^2 + \frac{1}{2}\left(\frac{\partial \phi}{\partial x}\right)^2 + U(\phi)$$ to be finite, then $U(\phi)$ must approach a minimum $\phi_i$ such that $U(\phi_i)=0$ as $x\to \pm \infty$.
Question: Is this true? If so why?
My attempt: If the minimum is at $\phi_i=0$, then the proof is trivial. So suppose that $\phi_i \neq 0$. For simplicity consider only the case that $\lim_{x\to\infty}\phi(x) = \phi_i$. Then, for large values of $x$ we have that
\begin{align} \frac{1}{2}\phi^2 + \frac{1}{2}\left(\frac{\partial \phi}{\partial x}\right)^2 + U(\phi) &\sim \frac{1}{2}\phi_i^2 + U(\phi_i) \qquad \qquad (\text{large $x$})\\ &=\frac{1}{2}\phi_i^2 \end{align}
It follows then that the integrand converges to a nonzero value asymptotically, and since the integrand is positive definite we have that $E[\phi]\to \infty$ as $x\to \infty$. This is in contradiction with the image attached.
Question 2: Is this just a manifestation of the "total energy" divergence that plagues all field theories and really we should be looking at energy differences?
Comments: I am posting this on Pysics SE and not Math SE, because I think the actual answer to this question is an underlying assumption in field theories that I may be missing, not so much a mathematical error.