= Solution
Use the <Minkowski metric> $\operatorname{diag}(1,-1)$ and the angular field $u=\beta\phi$. The <Euler-Lagrange equation> becomes $u_{tt}-u_{xx}+m^2\sin u=0$. For the dimensionless <light-cone coordinates>
$$
\xi=\frac m2(x+t),\qquad \eta=\frac m2(x-t),
$$
this is $u_{\xi\eta}=\sin u$. This normalization keeps the coupling in the relation between the physical field and its angle, rather than silently setting $\beta=1$.
A <Bäcklund transformation> is a system of first-order differential relations that maps a solution to another solution. One convention for the <Sine-Gordon Bäcklund transformation> uses a nonzero parameter $a$ and defines $v=\mathcal B_a[u]$ by
$$
\boxed{\partial_\xi\frac{v+u}{2}=\frac1a\sin\frac{v-u}{2},\qquad
\partial_\eta\frac{v-u}{2}=a\sin\frac{v+u}{2}.}
$$
A compatible initial value or integration constant selects a particular transformed solution. Put $w=(v+u)/2$ and $d=(v-u)/2$. Differentiating gives $w_{\xi\eta}=\cos d\sin w$ and $d_{\xi\eta}=\cos w\sin d$. Their sum and difference yield
$$
v_{\xi\eta}=\sin(w+d)=\sin v,\qquad
u_{\xi\eta}=\sin(w-d)=\sin u.
$$
Thus compatibility of the first-order relations contains the field equations for both fields, and \b[the transformed physical field $v/\beta$ also solves <Sine-Gordon theory>].
The transformation provides a generating <conservation law>. On the branch close to the original field write $v=u+2d$. Its first equation is
$$
\sin d=a(u_\xi+d_\xi).
$$
It recursively determines a formal small-$a$ expansion of $d$ in local derivatives of $u$:
$$
d=a u_\xi+a^2u_{\xi\xi}
+a^3\left(u_{\xi\xi\xi}+\frac16u_\xi^3\right)+\cdots.
$$
At every subsequent order the new coefficient occurs linearly, so this recursion continues indefinitely. The second Bäcklund equation and the first imply the exact identity
$$
\boxed{\partial_\eta P(a)=\partial_\xi Q(a),\qquad
P(a)=\frac{1-\cos d}{a^2},\quad Q(a)=1-\cos(u+d).}
$$
Indeed $\partial_\eta\cos d=-a\sin d\sin(u+d)$ and $\partial_\xi\cos(u+d)=-a^{-1}\sin d\sin(u+d)$. Comparing powers of $a$ therefore gives local <conserved currents>. If $\partial_\eta P_j=\partial_\xi Q_j$, our coordinate convention gives
$$
\partial_t(P_j+Q_j)=\partial_x(P_j-Q_j),\qquad
\mathcal I_j=\int_{\mathbb R}(P_j+Q_j)dx.
$$
For localized fields approaching vacua at spatial infinity, the boundary flux vanishes and $\mathcal I_j$ is conserved. The formal expansion need not converge: each coefficient is a separately exact local <conservation law>.
For example $P_0=u_\xi^2/2$, $Q_0=1-\cos u$, a light-cone combination of <energy> and <momentum>. The next coefficient is a derivative improvement, so it contributes no independent charge under the same decay conditions. At order $a^2$, remove the improvement generated by $\partial_\xi(u_\xi u_{\xi\xi})$ and multiply by $-2$. One obtains the genuinely higher <conservation law>
$$
\boxed{\partial_\eta\left(u_{\xi\xi}^2-\frac14u_\xi^4\right)
=\partial_\xi\left(\cos u\,u_\xi^2\right).}
$$
It can also be checked directly using $u_{\xi\eta}=\sin u$. Continuing the recursion, and using the opposite light-cone construction, produces the <local conserved-charge hierarchy of sine-Gordon theory>, with infinitely many nontrivial higher-spin charges after derivative improvements are removed. This is the <Bäcklund generating current for sine-Gordon conserved charges>. The hierarchy is the characteristic field-theory form of <classical integrability>; an ordinary <energy> <conservation law> alone would not supply these constraints.
The same transformation constructs solutions rather than only currents. Starting from the vacuum $u_0=0$, its two first-order equations integrate to
$$
\boxed{u_a=4\arctan E_a,\qquad E_a=\exp(\xi/a+a\eta+\delta_a).}
$$
For $a=e^\rho>0$ the exponent is $m\cosh\rho\,x-m\sinh\rho\,t+\delta_a$. This is a <Sine-Gordon kink> with velocity $\tanh\rho$, center set by $\delta_a$ and classical <rest mass> $8m/\beta^2$. Negative parameters can supply the corresponding opposite-orientation seeds.
The allowed <Bianchi permutability for sine-Gordon Bäcklund transformations> gives the two-step field algebraically. For vacuum seed and $a\ne b$,
$$
\boxed{\tan\frac{u_{a,b}}4
=\frac{a+b}{b-a}\frac{E_b-E_a}{1+E_aE_b}.}
$$
Branches of the inverse tangent must be continued smoothly; its principal value alone does not specify the vacuum labels of a multi-kink field. All angles here are $u=\beta\phi$. In the physical-field version each field difference in the superposition formula carries $\beta$; the formula printed without it implicitly uses the $\beta=1$ angular-field convention.
To display two real scattering solutions, take $0<v<1$, set $\Gamma=(1-v^2)^{-1/2}$, and put $X=m\Gamma x$, $Y=m\Gamma vt$. Choosing $a=e^\rho$, $b=e^{-\rho}$ with $v=\tanh\rho$ and zero phase constants yields the <Sine-Gordon kink-antikink scattering solution>
$$
\boxed{u_{K\bar K}(x,t)=-4\arctan\frac{\sinh Y}{v\cosh X}.}
$$
Choosing instead $b=-e^{-\rho}$ yields the <Sine-Gordon two-kink solution>
$$
\boxed{u_{KK}(x,t)=4\arctan\frac{v\sinh X}{\cosh Y}.}
$$
The latter has net angular winding $4\pi$, while the former has zero net winding. At large positive or negative time they separate into localized <kinks> with velocities $\pm v$. Repeated commuting transformations give general multi-soliton fields. Continuing the kink-antikink velocity to an imaginary value also gives a real <Sine-Gordon breather>:
$$
u_B(x,t)=4\arctan\left[\frac{\sqrt{m^2-\omega^2}}{\omega}
\frac{\sin\omega t}{\cosh(\sqrt{m^2-\omega^2}\,x)}\right],\qquad0<\omega<m,
$$
up to the irrelevant overall field sign and translations.
The scattering interpretation ties the construction to the conserved hierarchy. In an exact multi-soliton sector, the incoming species and <rapidities> reappear after the collision, up to permutation: the solitons change positions, not their asymptotic shapes or velocities, and the collision emits no radiation. In the two-kink example the large-time centers obey $|X|=|Y|-\log v+o(1)$, so a right-moving trajectory acquires a shift $-2\log v/(m\Gamma)$. These shifts encode the interaction even though the collision is elastic. The commuting construction makes the net displacement in a many-soliton collision the sum of its pairwise displacements, independent of how the collisions are ordered; this is <pairwise additivity of soliton shifts> and the classical counterpart of <factorized scattering>.
Conservation of the entire hierarchy is much stronger than conservation of <energy> and <momentum>: its independent rapidity-weighted sums constrain the whole asymptotic soliton data and rule out particle production in this sector. Generic initial fields may also contain radiation scattering data, so this statement is about the exact soliton collisions, not a claim that every initial field is a pure soliton. \b[The Bäcklund map unifies the conserved hierarchy, explicit soliton construction and elastic, pairwise scattering picture of <classical integrability>.]
Back to article page
