= Solution
Choose boundary states with nonzero overlap with the lowest-energy state in the vacuum sector $Q=0$ and in the one-<kink> <topological sector> $Q=1$. In a finite spatial box, a fixed <scalar field configuration eigenstate> may be used formally, with temporal endpoints equal to the chosen configuration. More generally, smear the endpoints with wavefunctionals $\Psi_Q$. Their <unitary time evolution> kernels are
$$
\begin{aligned}
K_Q(T)&=\langle\Psi_Q|e^{-i\widehat HT}|\Psi_Q\rangle\\
&=\int\mathcal D\phi_f\,\mathcal D\phi_i\;\Psi_Q^*[\phi_f]\Psi_Q[\phi_i]\int_{\phi(0)=\phi_i}^{\phi(mT)=\phi_f}\mathcal D\phi\;e^{iS[\phi]}.
\end{aligned}
$$
The inner <scalar field path integral> remains in the chosen <topological sector>, with $\phi(+\infty)-\phi(-\infty)=2\pi Q$. A zero-total-<momentum> projection can be included to select the rest state; alternatively, the translational prefactor does not change the large-time exponential. After a <Wick rotation> to physical Euclidean time $\tau$, the <energy eigenstate> expansion gives $K_Q^E(\tau)\sim C_Qe^{-E_Q\tau}$. Thus the exact <vacuum-subtracted soliton mass> is
$$
\boxed{M=-\lim_{L\to\infty}\lim_{\tau\to\infty}\frac1\tau\log\frac{K_1^E(\tau)}{K_0^E(\tau)}.}
$$
Equivalently it is $iT^{-1}\log(K_1/K_0)$ at large time with a damping prescription. The ratio subtracts the <vacuum energy>; the <Hamiltonian operator> and <action> here are the fully regulated and renormalized ones, not merely their classical approximations.
With the dimensionless coordinates of this paper, the correctly normalized classical <action> is
$$
S[\phi]=\frac1{\beta^2}\int_0^{mT}dt\int dx\left[\frac12\dot\phi^2-\frac12(\phi')^2+\cos\phi-1\right].
$$
For the static <Sine-Gordon kink>, $\phi_K'=2\operatorname{sech}x$ and $1-\cos\phi_K=2\operatorname{sech}^2x$, so $M_{\rm cl}=8m/\beta^2$. Write $\eta=\delta\phi$. Expanding and integrating by parts gives
$$
\boxed{S[\phi_K+\eta]=-\frac{8mT}{\beta^2}+\frac1{2\beta^2}\int_0^{mT}dt\int dx\;\eta(-\partial_t^2-\Delta_x)\eta+O(\eta^3/\beta^2),\quad \Delta_x=-\partial_x^2+\cos\phi_K=-\partial_x^2+1-2\operatorname{sech}^2x.}
$$
The first variation is $\beta^{-2}\int(\phi_K''-\sin\phi_K)\eta$, plus boundary terms. It vanishes because the <kink> satisfies the <Euler-Lagrange field equation> and the fluctuations have the prescribed temporal endpoints and admissible spatial boundary behavior. This is the <principle of stationary action>, not a symmetry assumption about $\eta$.
The printed expansion omits $\beta^{-2}$ despite the stated <Lagrangian density>. \b[Its displayed form is the expansion of $\beta^2S$; it is not the physical $S$ at arbitrary coupling.] Alternatively, writing $\eta=\beta\chi$ puts the quadratic term in canonical normalization, while leaving the classical term $-8mT/\beta^2$ unchanged. This normalization repair does not change $\Delta_x$ or the physical fluctuation frequencies.
The <Sine-Gordon kink fluctuation operator> has the useful factorization
$$
A=\partial_x+\tanh x,\qquad \Delta_x=A^\dagger A,\qquad AA^\dagger=-\partial_x^2+1.
$$
It is nonnegative, and $A\psi_0=0$ gives the normalized <translational zero mode of a sine-Gordon kink>:
$$
\boxed{\omega_0^2=0,\qquad \psi_0(x)=\frac{\operatorname{sech}x}{\sqrt2}\ \propto\ \phi_K'(x).}
$$
A displacement $a$ changes $\phi_K(x-a)$ by $-a\phi_K'(x)$. Thus the <zero mode in field theory> is the position <collective coordinate> of the <kink>, reflecting <translation invariance>. It has no restoring force and no oscillator <zero-point energy>. Integrate that <collective coordinate> separately rather than inserting a zero factor into the Gaussian <functional determinant>.
\Image[/past-exam-of-the-mathematics-course-of-the-university-of-cambridge/2015/iii/paper-50-fluctuation.png]
{title=Sine-Gordon kink fluctuation potential and normalized translational zero mode}
{height=420}
In the <Gaussian fluctuation approximation>, the formal oscillator contribution to the <one-loop soliton mass correction>, before adding any <counterterm>, is
$$
\boxed{\Delta M_{\rm osc}(L)=\frac m2\left(\sum_n\omega_n-\sum_n\omega_n^{(0)}\right),\qquad \omega_n=\sqrt{\lambda_n(\Delta_x)},\quad \omega_n^{(0)}=\sqrt{\lambda_n(-\partial_x^2+1)}.}
$$
Use a common regulator for the two sums, include all discrete modes, and treat the translation mode as above. The vacuum sum is essential: subtracting only classical vacuum energy would leave an extensive oscillator energy. A periodic fluctuation and its derivative are matched at the two ends of the large box; the one-<kink> background lies in the twisted <topological sector>, and its infinite-line profile is accurate up to exponentially small boundary corrections.
For a continuum <scattering wavefunction>, equality of its two asymptotic values gives the <periodic-box phase-shift quantization>
$$
e^{ik_nL+i\delta(k_n)}=1,\qquad k_nL+\delta(k_n)=2\pi n,\qquad k_n^{(0)}=\frac{2\pi n}{L}.
$$
Away from the threshold, expand at a matched mode number:
$$
k_n-k_n^{(0)}=-\frac{\delta(k_n^{(0)})}{L}+O(L^{-2}),\qquad \omega_n-\omega_n^{(0)}=-\frac{\delta(k_n^{(0)})}{L}\frac{k_n^{(0)}}{\sqrt{1+(k_n^{(0)})^2}}+O(L^{-2}).
$$
Since consecutive free wave numbers are separated by $2\pi/L$, replacing the continuum-mode sum by an <integral> proves the displayed continuum contribution:
$$
\boxed{\Delta M_{\rm cont}\simeq-\frac m2\int_{-\Lambda}^{\Lambda}\frac{dk}{2\pi}\,\delta(k)\frac{k}{\sqrt{k^2+1}}.}
$$
The <ultraviolet cutoff> is retained until the <counterterm> is added.
There is a finite threshold issue if this expression is identified with the complete oscillator correction. It can be settled directly using the factorization: a continuum <eigenfunction> is $f_k\propto(\tanh x-ik)e^{ikx}$. Its transmission phase obeys
$$
e^{i\delta(k)}=\frac{1-ik}{-1-ik}=\frac{k+i}{k-i},\qquad \delta(k)=2\arctan(1/k)\quad(k\ne0),
$$
with the odd phase branch that tends to zero at large $|k|$. For $k>0$, the continuum roots have labels $n=1,2,\ldots$; there is no periodic continuum root at $k=0$, since the limiting eigenfunction $\tanh x$ has opposite signs at the two ends. The <bound state> at $\omega_0=0$ replaces the free $k=0$ oscillator with $\omega_0^{(0)}=1$. Consequently, in <mode-number regularization of soliton masses>,
$$
\boxed{\Delta M_{\rm osc}\simeq-\frac m2+\Delta M_{\rm cont}.}
$$
\b[The PDF's continuum-only formula misses this finite $-m/2$ under this standard phase and mode-counting convention.] It has the correct logarithmic <ultraviolet divergence>, but the missing term is not suppressed by large $L$. Changing the phase branch requires changing the mode labels and endpoint terms consistently; it cannot erase a physical mode from the formal spectrum sum.
At high momentum, $\delta(k)=2/k+O(k^{-3})$, so both expressions have divergent part $-(m/\pi)\log\Lambda$. The canonical field $\varphi=\phi/\beta$ has a quartic interaction with coupling $-m^2\beta^2$. Its vacuum <tadpole diagram> shifts the squared <mass> by $-m^2\beta^2 J(\Lambda)/4$, where
$$
J(\Lambda)=\int_{-\Lambda}^{\Lambda}\frac{dk}{2\pi\sqrt{k^2+1}}=\frac{\operatorname{arsinh}\Lambda}{\pi}.
$$
The <Sine-Gordon vacuum tadpole counterterm> has $\delta m^2=+m^2\beta^2J/4$ and adds potential energy density $\delta m^2(1-\cos\phi)/\beta^2$. Since $\int(1-\cos\phi_K)dx=4$, its vacuum-subtracted <kink> energy is
$$
\boxed{\Delta M_{\rm ct}=\frac{4\delta m^2}{m\beta^2}=mJ(\Lambda)\sim\frac m\pi\log(2\Lambda),}
$$
which cancels the logarithmic <ultraviolet divergence>. Finite parts require a specified <renormalization condition>. As a consistency check, with this tadpole subtraction and matched mode-number cutoff, integration by parts yields
$$
\Delta M_{\rm osc}+\Delta M_{\rm ct}=-\frac{m}{2\pi}\,\delta(\Lambda)\sqrt{\Lambda^2+1}\longrightarrow-\frac m\pi.
$$
The complete <semiclassical soliton mass> is then $8m/\beta^2-m/\pi+O(m\beta^2)$ in that convention. This last finite result uses the bound-mode term and is additional to the requested ultraviolet cancellation.
Back to article page
