Past exam of the mathematics course of the University of Cambridge / 2003 / iii / Paper 52 (Solitons and Instantons)

For a compactly supported variation $\varphi \mapsto \varphi +\epsilon \eta$, the first variation of the energy functional isIntegration by parts has no endpoint term. Since $\eta$ is arbitrary, the static Euler-Lagrange field equation isIntegrating the specified polynomial derivative gives $U(\varphi )={\varphi }^2/2-{\varphi }^4+{\varphi }^6/2+C$. The polynomial factors as a nonnegative square times ${\varphi }^2$; its global minimum is zero when $C=0$. ThereforeThese are the three constant scalar-field vacua. For a localized finite-energy kink, the two ends approach vacua and the derivative tends to zero. Multiplying the field equation by ${\varphi }^{\prime }$ gives the first integralA nonconstant kink must stay within one interval between neighboring vacua. A proposed direct connection from $-1$ to $1$ would reach $\varphi =0$ at some finite point, where the first integral forces ${\varphi }^{\prime }=0$. Uniqueness for the smooth second-order differential equation then forces the identically zero solution, a contradiction. This is the intermediate-vacuum obstruction to a kink; two elementary kinks can span the outer vacua only in an infinite-separation limit, not as one finite-width static kink.

There are consequently four oriented elementary sectors: $-1\to 0$, $0\to -1$, $0\to 1$ and $1\to 0$. If opposite orientations are grouped together, there are two kink families and their antikinks. Spatial reflection reverses orientation, and $\varphi \mapsto -\varphi$ exchanges the two adjacent intervals; both preserve the energy. All four therefore have the same rest mass, even though the central and outer vacua have different fluctuation curvatures.

For the sector $0\to 1$, define $W(\varphi )={\varphi }^2/2-{\varphi }^4/4$, so $U=(W^{\prime }{)}^2/2$ and $W^{\prime }=\varphi (1-{\varphi }^2)$. The Bogomolny bound follows by completing the square:It is saturated by the Bogomolny equationLet $y={\varphi }^2$ on this positive branch. Then $y^{\prime }=2y(1-y)$, so $\log [y/(1-y)]=2(x-x_0)$. Thus the explicit kink in a phi-six model isIt has the required limits and satisfies the first integral. The translated center $x_0$ is arbitrary. All oriented elementary solutions areFor $\eta =1$ the endpoints are $0\to \sigma$, and for $\eta =-1$ they are $\sigma \to 0$. Direct evaluation ${\int }_0^1\sqrt{2U(\varphi )}d\varphi ={\int }_0^1\varphi (1-{\varphi }^2)d\varphi =1/4$ independently gives their common mass.

In the usual smooth localized finite-energy class, the positive potential requires $|\varphi |\to 1$ at infinity and the gauge covariant derivative tends to zero there. On a sufficiently large circle write $\varphi \simeq e^{i\chi }$ locally. Then the tangential gauge potential approaches $d\chi$. The normalized Higgs phase defines a continuous map $S_{\infty }^1\to U(1)$ with integer winding numberThe second equality is Stokes theorem with the chosen orientation and $D_j={\partial }_j-i{a}_j$. Suitable decay makes the integrated covariant-derivative error vanish. Thus magnetic flux quantization of an Abelian Higgs vortex labels disconnected boundary sectors. This uses the standard continuous vacuum asymptotics, not a claim that bare integral finiteness alone supplies every pointwise boundary limit.

A nonzero winding requires zeros of the complex Higgs field somewhere inside: a nowhere-zero field on the whole disk would extend its normalized boundary map and make its winding zero. For isolated zeros the signed local indices sum to $N$. For arbitrary configurations, vortex-antivortex pairs may add zeros without changing $N$, so $|N|$ need not count all unsigned zeros.

The critical coupling relates the charge to a minimum energy. Set $j_i=\mathrm{Im}(\bar{\varphi }{D}_i\varphi )$. The covariant derivative commutator $[D_1,D_2]\varphi =-iB\varphi$ givesDiscarding its vanishing boundary integral and completing the magnetic square yieldsFor positive $N$, the Bogomolny vortex equations are $(D_1+i{D}_2)\varphi =0$ and $B=(1-|\varphi {|}^2)/2$. Reverse both signs for negative charge. HenceCritical-coupling vortex solutions attain this bound; their energy depends only on $|N|$. Positive-charge solutions have $N$ zeros counted with multiplicity, and their positions provide $2N$ real coordinates on the Abelian Higgs vortex moduli space. Coincident zeros are allowed. Position variations leave the static minimum energy unchanged, expressing the absence of static forces among same-sign Bogomolny vortices. The sign of $N$ reverses flux and orientation. Topology alone does not determine an arbitrary field's profile or full energy: the zero-charge vacuum has energy zero, while nonvacuum configurations in the same sector can have positive energy.

For the given radial ansatz, $a_{\theta }$ is the coordinate one-form coefficient: the gauge potential is $a(r)d\theta$, and the orthonormal azimuthal component is $a(r)/r$. Thus $B=a^{\prime }(r)/r$, consistently with the printed energy. Since $h(\infty )=1$, the phase $e^{i\theta }$ has winding one; equivalently $a(0)=0$, $a(\infty )=1$ givesFor the trial functions all energy is inside $0<r<r_0$. Their four contributions to $E/\pi$ areTherefore the compact-core variational estimate for an Abelian Higgs vortex isWriting $s=r_0^2>0$, the derivative is $-2/s^2+1/24$, and the energy diverges at both ends of the $s$ range. The unique global minimum isThe magnetic and potential energies agree at this optimum, as required by radial scale balance. The piecewise fields are continuous with square-integrable first derivatives; their derivative jumps do not create an additional surface energy. They are trial fields, not exact solutions. Their optimized energy is an upper-bound estimate, about 24.4% above the exact critical one-vortex energy $\pi$.

A Derrick scaling variation changes a configuration's size without changing its topological sector. Use $U_L(\mathbf{x})=U(\mathbf{x}/L)$, so $L>1$ expands the configuration. An integrated nonnegative term homogeneous in $s$ spatial derivatives scales as $L^{d-s}{E}_s$. For a scalar gradient-plus-potential energy, for example,A static solution is a stationary point of the full energy functional, so it must in particular satisfy $E^{\prime }(1)=0$. In this example the Derrick virial identity is $(d-2)E_2+d{E}_0=0$. If both energies are nonnegative and the coefficients cannot balance, a nontrivial localized static solution is impossible. If balance is possible, it fixes relations among energy contributions and may select a characteristic size. A negative $E^{\prime \prime }(1)$ proves an instability under size changes; a positive second derivative gives stability only against that one deformation. Passing the scale test is a necessary condition, not an existence theorem or a proof of full stability. It does not directly rule out time-dependent or charge-constrained solitons outside the static variational problem.

For the specified model, each spatial current is anti-Hermitian: differentiating $U{U}^{†}=1$ gives $R_i^{†}=-R_i$. Thus $-\mathrm{Tr}(R_i^2)$ is nonnegative, and the static energy of the nonlinear sigma model isThe boundary condition $U\to 1_2$ identifies spatial infinity to one point, so one-point compactification turns the field into a based map $S^d\to SU(2)$. By SU(2) as the three-sphere, its sectors are classified by the homotopy group ${\pi }_d(S^3)$. In particular,There is no corresponding topological protection in one or two spatial dimensions. In three dimensions there is an integer degree of a map between oriented manifolds, the usual topological baryon number in the Skyrme model when the same field is interpreted in that theory. With a compatible orientation convention it can be writtenConjugation from the left current to the right current preserves the trace in this formula. Higher dimensions are governed by their corresponding homotopy groups, which need not all be trivial; no assumption of their triviality is needed in the scaling obstruction below.

For the pure sigma model, $E(L)=L^{d-2}{E}_2$, so stationarity requires $(d-2)E_2=0$. In every $d\ne 2$ a finite-energy nonconstant static solution is excluded by Derrick scaling. In $d=1$ spreading lowers its energy; in $d>2$ shrinking lowers it. In particular, a degree-carrying three-dimensional sigma configuration can collapse toward zero size and energy within its sector, with a singular limiting field. The topological charge prevents smooth unwinding, but not this singular collapse.

In $d=2$, the sigma energy is scale invariant. The scaling identity supplies no obstruction, but ${\pi }_2(S^3)=0$ gives no topological soliton sector either. One must not infer nonexistence of every nonconstant stationary harmonic map from these two facts: scale-invariant sigma models can have non-topological harmonic-map saddles, for example using a totally geodesic two-sphere in the target three-sphere. The conclusion is absence of a topologically protected sigma soliton here, with no positive scale-restoring force supplied by the quadratic term alone.

For the Skyrme model, the added commutator term gives the nonnegative static energyAnti-Hermiticity of the commutators fixes its sign. The target and boundary condition are unchanged, so the topological classification is unchanged. The Derrick dimension test for the Skyrme model instead becomesIn $d=3$, the opposite exponents permit a balance:For an arbitrary reference shape with both energies nonzero, its preferred scale is $L_{\ast }=(E_4/E_2{)}^{1/2}$ and its optimized energy is $2\sqrt{E_2{E}_4}$. Expansion costs quadratic-gradient energy and contraction costs quartic-gradient energy. Integer-degree Skyrmions are therefore topologically possible and evade the pure sigma collapse. This mechanism permits finite-size stable solitons, although existence and stability in all field directions require more than the scaling argument itself.

The other dimensions behave differently. In $d=1$, all spatial commutators vanish identically, so the added term changes nothing and nonconstant static finite-energy fields remain excluded. In $d=2$, stationarity forces $E_4=0$; any configuration with nonzero commutator energy can lower it by spreading. In fact a smooth finite-energy static field of this particular potential-free model must be constant. To prove this commuting-current obstruction in a two-dimensional Skyrme model, $E_4=0$ forces $[R_1,R_2]=0$ pointwise. The quartic first variation then vanishes, leaving the sigma equation ${\partial }_i{R}_i=0$. The right-current Maurer-Cartan equation isDivergence-free and curl-free currents have componentwise $\Delta R_i=0$. Their components belong to L2 space by finite $E_2$. An entire square-integrable harmonic function on the plane vanishes: the mean-value formula and Cauchy–Schwarz bound its value by $(\pi R^2{)}^{-1/2}$ times its full L2 norm, tending to zero as the disk radius $R$ grows. Thus $R_i=0$, and the boundary condition fixes $U=1_2$. This additional argument uses the vanishing commutator; it does not apply to the pure sigma model's general noncommuting currents.

In $d=4$, the virial relation gives $2E_2=0$, hence a constant field and zero $E_4$. In $d>4$ both coefficients are positive and the same conclusion follows. Only three spatial dimensions allow a nontrivial finite-size balance of these two positive derivative energies. Higher-dimensional topological classes, where present, do not evade this obstruction.

An instanton is a localized finite-action classical solution of a Euclidean field theory. Its role differs from that of a static spatial classical field-theory soliton: an instanton is localized in Euclidean spacetime and contributes to quantum transition amplitudes, rather than persisting as a stationary real-time particle. Wick rotation changes the oscillatory weight $e^{iS/\hslash }$ into $e^{-S_E/\hslash }$ when the continuation and boundary conditions are valid. The Euclidean path integral is consequently dominated semiclassically by stationary Euclidean configurations. Perturbation theory around one vacuum misses other saddles whose weights have the nonanalytic form $e^{-S_I/\hslash }$.

A useful first example is quantum tunnelling between two degenerate minima in quantum mechanics. For coordinate $q$ with mass $m$, the Euclidean action iswith the vacuum energy subtracted from $V$. Its Euler-Lagrange equation is $m\ddot{q}=V^{\prime }(q)$, equivalent to motion in the inverted potential $-V$. A finite-action trajectory joining zero-energy minima has conserved Euclidean energy $m{\dot{q}}^2/2-V=0$. A forward instanton therefore satisfies $\dot{q}=\sqrt{2V/m}$, while an anti-instanton has the opposite orientation. Completing the square gives the action boundwhich these zero-energy paths saturate. This is the Euclidean counterpart of the first-order kink argument, with Euclidean time replacing the spatial coordinate.

For the quartic double-well instanton, chooseThe first-order equation in the interval $-a<q<a$ is $\dot{q}=\omega (a^2-q^2)/(2a)$. Separation givesThe center ${\tau }_0$ is arbitrary. Its derivative ${\dot{q}}_I$ is an instanton translation zero mode: differentiating the Euclidean field equation shows that it lies in the kernel of the quadratic fluctuation operator $-m{\partial }_{\tau }^2+V^{\prime \prime }(q_I)$. A naive Gaussian determinant therefore has a zero eigenvalue. Replace its integration by the collective-coordinate integral over ${\tau }_0$, and take the determinant only over nonzero modes. The resulting instanton fluctuation prefactor combines that Jacobian with the determinant ratio against the vacuum. The action controls the exponential suppression; the prefactor is essential for a quantitative rate or splitting.

When $S_I/\hslash \gg 1$, crossings are rare and their typical separations greatly exceed their widths. In the dilute instanton gas, ordered center integrals for $n$ well-separated crossings over Euclidean duration $\mathcal{T}$ produce ${\mathcal{T}}^n/n!$. If a single-crossing rate is $\kappa =K{e}^{-S_I/\hslash }$, a symmetric two-well system has even-crossing and odd-crossing sums proportional to $\cosh (\kappa \mathcal{T})$ and $\sinh (\kappa \mathcal{T})$. The spectral interpretation gives levels $E_{\mathrm{well}}\mp \hslash \kappa$ andThis is a real tunnelling level splitting. A Euclidean bounce instead describes false vacuum decay: it leaves and returns to a metastable vacuum, and a characteristic negative fluctuation mode supplies the imaginary part interpreted as decay. A bounce must not be confused with a stable double-well instanton merely because both use Euclidean saddles.

The central gauge-theory example is a Yang-Mills instanton on Euclidean four-space. State conventions carefully: take anti-Hermitian generators $T_a=-i{\sigma }_a/2$, with $\mathrm{Tr}(T_a{T}_b)=-{\delta }_{ab}/2$, connection $A=A_{\mu }d{x}^{\mu }$, and gauge curvature $F=dA+A\wedge A$. With coupling $g$, the positive Yang-Mills action isThe Hodge star operator squares to one on Euclidean two-forms. Under the usual finite-action boundary conditions, the connection approaches a pure gauge at infinity, defining a map $S_{\infty }^3\to SU(2)\simeq S^3$. Its winding degree is the integer instanton numberThis is unchanged by smooth deformations preserving the boundary sector. Geometrically this is, up to the stated sign convention, the second Chern number of the compactified bundle. Different sign conventions for trace, curvature or orientation must be changed together.

Use the positive norm $\Vert F{\Vert }^2=-\int \mathrm{Tr}(F\wedge \ast F)$. Decomposition into self-dual and anti-self-dual curvature givesThe Yang-Mills instanton Bogomolny bound is saturated by $F=\ast F$ for positive $k$, or $F=-\ast F$ for negative $k$. These first-order equations imply the full Yang-Mills equations: the Bianchi identity $D_{A}F=0$ becomes $D_A\ast F=0$ when $\ast F=\pm F$. Thus the topological minimum is a genuine classical saddle, not just an arbitrary field with integer charge. Self-duality alone is a local differential condition; finite action is an additional requirement.

An explicit unit-charge example is the BPST instanton. Let $y=x-x_0$, $\rho >0$, and choose the self-dual symbols ${\eta }_{ij}^a={ϵ}_{aij}$, ${\eta }_{i4}^a={\delta }_{ai}$ and ${\eta }_{4i}^a=-{\delta }_{ai}$ for $i,j=1,2,3$, using orientation $d{x}^1\wedge d{x}^2\wedge d{x}^3\wedge d{x}^4$. In regular gauge,The curvature is self-dual. Since ${\sum }_{a,\mu ,\nu }({\eta }_{\mu \nu }^a{)}^2=12$, its squared component curvature is $192{\rho }^4/(y^2+{\rho }^2{)}^4$. Radial integration yieldsso $k=1$. Replacing self-dual symbols by anti-self-dual ones produces the opposite charge with the same positive action.

The center has four coordinates and $\rho$ is an instanton size modulus. Classical four-dimensional Yang-Mills action is scale invariant, so varying $\rho$ changes the action-density width but not the total action. If gauge transformations must tend to the identity at infinity, three global gauge orientations also remain, giving eight parameters for the one-instanton framed instanton moduli space. If those global orientations are quotiented out, the unframed one-instanton space has five parameters. A zero-size limit is a singular concentration of curvature, not another smooth instanton. Higher-charge solutions likewise possess instanton moduli spaces; their collective-coordinate integrals replace the corresponding zero-mode determinants.

In a Euclidean-time description, topologically distinct gauge vacua are separated by a change of Chern-Simons number; an instanton interpolates between them, with the change equal to $k$ in compatible orientation conventions. Theta-weighted instanton sectors give $Z(\vartheta )={\sum }_k{e}^{ik\vartheta }{Z}_k$, connecting topology to the vacuum angle. With fermions, the gauge-field topology also controls fermionic zero modes and chiral selection rules through the chiral anomaly. These phenomena explain why instantons encode physics beyond small fluctuations around a single vacuum.

Instantons link Euclidean classical solutions, topological sectors and exponentially suppressed quantum effects. Their interpretation requires the boundary conditions, normalization, fluctuation spectrum and measure over collective coordinates. A dilute approximation is controlled only when its individual-action suppression and separation assumptions hold; classical scale moduli and their quantum corrections can make size integrals delicate rather than automatically finite.