Past exam of the mathematics course of the University of Cambridge / 2012 / iii / Paper 55 (Applications of Differential Geometry to Physics)

Use the Lie bracket convention $[X,Y]=XY-YX$. If $E_{ij}$ denotes a matrix unit, then $E_{ij}{E}_{kl}={\delta }_{jk}{E}_{il}$. Here this givesThus, writing $[X_j,X_k]=f^i{}_{jk}{X}_i$, the nonzero structure constants of a Lie algebra areAll other entries vanish. This is the Lie algebra of the translation-dilation group of the plane: its two-dimensional translation ideal is abelian, and $X_0$ acts on that ideal as the identity. In group coordinates the multiplication and inverse areThese formulas also make the semidirect product structure explicit.

The Maurer-Cartan form is $g^{-1}dg={\lambda }^j{X}_j$. Direct matrix multiplication givesThese are left-invariant differential forms: for a constant group element $a$, $(ag{)}^{-1}d(ag)=g^{-1}dg$. Equivalently, left translation sends $(\rho ,x)$ to $(a\rho ,b+ax)$, so all three coordinate differentials and the denominator acquire the same factor $a>0$.

The left-invariant coframe of the translation-dilation group is pointwise linearly independent because $\rho >0$. Consequentlyis positive definite and unchanged under every left translation. It therefore defines the required left-invariant metric, with precisely the coordinate expression obtained from this sum. The notation $d{\rho }^2$ in that metric means $d\rho \otimes d\rho$, not the differential of ${\rho }^2$.

Taking the exterior derivative of the left-invariant differential forms yieldsFor example, $d({\rho }^{-1}d{x}^1)=-{\rho }^{-2}d\rho \wedge d{x}^1$. With the antisymmetric structure constants of a Lie algebra found above, the two terms with indices $(0,1)$ and $(1,0)$ combine asThe same computation holds for superscript $2$, and all superscript-$0$ terms vanish. Hence the Maurer-Cartan equation isThe factor $1/2$ avoids counting each antisymmetric pair twice. If coefficients were not required to be antisymmetric in $j,k$, their symmetric parts would be invisible to the wedge product of differential forms; the standard structure constants of a Lie algebra fix this ambiguity.

Take the frame dual to the left-invariant coframe of the translation-dilation group. It isIndeed ${\lambda }^i(L_j)={\delta }_j^i$. These are left-invariant vector fields, either because they are dual to invariant forms or because $L_j(g)=(d{L}_g{)}_e{X}_j$. For instance, the curve $g\exp (t{X}_0)$ differentiates to $\rho {\partial }_{\rho }$, while $g\exp (t{X}_1)$ differentiates to $\rho {\partial }_{x^1}$.

The coordinate Lie brackets satisfyFor the first identity only the derivative of the coefficient $\rho$ contributes. The linear map $\boxed{X_j⟼L_j}$ is a Lie algebra isomorphism: it preserves all basis brackets, is injective by independence of the frame, and is onto the three-dimensional space of left-invariant vector fields.

The topological degree measures an oriented net number of sheets of a map. Its simplest integer-valued setting is a map $f:M\to N$ between connected oriented closed manifolds of the same positive dimension $n$. The hypotheses matter: an integer sign count needs orientations; compactness prevents preimages from escaping; and the same dimension makes a regular fiber discrete. A fundamental class gives the definition for continuous maps:Here $H_n(M;\mathbb{Z})$ and $H_n(N;\mathbb{Z})$ are generated by their chosen fundamental classes. Reversing either orientation changes the sign of the degree of a map between oriented manifolds. Disconnected sources contribute a sum over components; disconnected targets require a degree for each target component rather than one universal integer.

For a smooth map, choose a regular value $y$. The Sard theorem guarantees plentiful choices. Each point of $f^{-1}(y)$ has an invertible derivative, so the inverse function theorem makes it an isolated point. The fiber is finite by compactness of $M$. An oriented coordinate chart defines $\mathrm{sgn}\det d{f}_x$, and the degree as a sum of local degrees becomesAn empty fiber contributes zero. The sign is independent of the chosen oriented charts. Thus the mapping degree counts sheets with signs, not simply the number of preimages: orientation-preserving and orientation-reversing sheets cancel.

A differential-form argument both proves that this count is independent of $y$ and relates it to the topological definition. Around a regular value, take a small neighborhood $V$ whose full preimage is a disjoint union of neighborhoods mapped diffeomorphically onto $V$. Compactness rules out additional preimages approaching $V$ from elsewhere. Choose a smooth top-degree form $\eta$ supported in $V$, with ${\int }_N\eta =1$. The change of variables formula givesAlthough $\eta$ need not be everywhere positive, it represents the same normalized top cohomology class as any normalized volume form. The top de Rham cohomology of a compact connected oriented manifold says that two top forms with equal integrals differ by $d\beta$. Their pullback integrals agree, because the Generalized Stokes theorem gives ${\int }_{M}d(f^{\ast }\beta )=0$. Therefore the degree by integration of a pullback volume form isfor every smooth top form $\eta$. Pairing de Rham cohomology with fundamental classes identifies this integer with the homological definition. This proof also shows that every regular value gives the same signed sum.

The homotopy invariance of mapping degree makes it stable under continuous deformation. Homologically this follows from homotopic maps inducing the same map on homology. For a smooth homotopy $F:M\times [0,1]\to N$, a direct proof uses $d\eta =0$ and the Generalized Stokes theorem on the cylinder to getwith the cylinder orientation chosen to give the indicated boundary signs. Smooth approximation extends the geometric description to continuous maps, while the fundamental class definition already applies without differentiability. The multiplicativity of mapping degree follows from functoriality and givesA constant map has degree zero when $n>0$. If the mapping degree is nonzero, the map is onto: a point outside its image is a regular value with empty preimage. An orientation-preserving diffeomorphism has degree $1$, and an orientation-reversing one has degree $-1$. An orientation-preserving $k$-sheeted covering map has degree $k$; the signs of sheets must be included for other orientation conventions.

Examples show both the strength and the limits of the invariant. On the circle, $e^{i\theta }\mapsto e^{ik\theta }$ has degree $k$, including negative integers and zero; it is the winding number. On the sphere $S^n$, an ambient reflection has degree $-1$, and the antipodal map has degree $(-1{)}^{n+1}$, because the determinant of $-I$ on the ambient ${\mathbb{R}}^{n+1}$ is $(-1{)}^{n+1}$. On the torus, an integer matrix $A$ induces a map of degree $\det A$, obtained by pulling back the constant top form. However, degree does not classify general manifold maps: the identity and an integer shear on the two-torus both have degree one but induce different maps on $H_1$, so are not homotopic.

The mapping degree gives useful obstruction and existence arguments. There is no retraction of a closed ball onto its boundary sphere: such a retraction would make the degree-one identity of the boundary null-homotopic, since it would extend across a contractible ball, contradicting homotopy invariance and the degree-zero constant map. This is the degree obstruction underlying the Brouwer fixed-point theorem. Also an everywhere nonzero tangent vector field on an even-dimensional sphere would, after normalization, produce the homotopy $\cos (\pi t)x+\sin (\pi t)v(x)$ from the identity to the antipodal map. Orthogonality makes every value a unit vector, but the endpoint degrees are $1$ and $-1$, impossible. This proves the corresponding Hairy ball theorem obstruction.

There are appropriate extensions rather than an unrestricted integer degree for every pair of manifolds. For a proper map between connected oriented manifolds without boundary of equal dimension, the regular-value count remains finite and is invariant; compactly supported top forms or locally finite homology replace the compact fundamental-class description. Without orientations, degree modulo two counts preimages modulo two and uses mod-two fundamental classes. For manifolds with boundary one uses relative mapping degree for maps of pairs, keeping the boundary away from the chosen target value. Maps of unequal dimensions, or nonproper maps where preimages can escape to infinity, do not in general admit this same signed sheet-count invariant. The central principle is an integer signed count stable under homotopy, with its hypotheses explicitly preserved.

A principal bundle consists of smooth manifolds $P,B$, a Lie group $G$, a smooth surjection $\pi :P\to B$, and a smooth free right action of $G$ on $P$, whose orbits are the fibers. Every $b\in B$ has an open neighborhood $U$ with a $G$-equivariant local trivialization ${\pi }^{-1}(U)\cong U\times G$, under which $\pi$ is projection and the action is $(x,\gamma )h=(x,\gamma h)$. Equivalently a local section $s:U\to P$ writes every fiber point uniquely as $p=s(x)\gamma$. On overlaps, sections differ by smooth transition functions $s_{\beta }=s_{\alpha }{u}_{\alpha \beta }$ satisfying the cocycle identity.

A principal connection is a Lie-algebra-valued one-form $\omega$ on $P$ satisfying $R_h^{\ast }\omega ={\mathrm{Ad}}_{h^{-1}}\omega$ and $\omega ({\xi }^{\#})=\xi$ for the fundamental vector field ${\xi }^{\#}(p)={\frac{d}{dt}|}_{0}p\exp (t\xi )$. Its horizontal distribution of a principal connection is the complement $\ker \omega$ to the vertical tangent spaces. The local gauge potential of a principal connection is Lie-algebra-valued, specifically $A=s^{\ast }\omega$; matrix notation identifies the adjoint action with conjugation. The displayed connection expression is understood on each coordinate trivialization, not as a choice of a global section.

Choose local sections related by $s^{\prime }=su$, where $u:U\to G$. Keeping the same point of the principal bundle requires $s\gamma =s^{\prime }{\gamma }^{\prime }$, so ${\gamma }^{\prime }=u^{-1}\gamma$. For reconstruction of a principal connection from local gauge potentials, the local principal connection form transforms asUsing $d(u^{-1})=-u^{-1}(du)u^{-1}$, computeandThe inhomogeneous terms cancel. Thus $\boxed{{\omega }^{\prime }=\omega }$: both local expressions define the same one-form on the total space. This is compatibility across trivializations, whereas $R_h^{\ast }\omega ={\mathrm{Ad}}_{h^{-1}}\omega$ describes the separate right-action equivariance of a principal connection. In particular a change of trivialization is not an assertion that the local gauge potential of a principal connection $A$ itself is unchanged.

Work over a coordinate neighborhood $U\subset B$ with coordinates $x^i$, and choose a basis $T_a$ of the Lie algebra, with $[T_a,T_b]=c^c{}_{ab}{T}_c$. Put $A=A_i^a{T}_{a}d{x}^i$. For each $T_a$ let $R_a$ be its right-invariant vector field on the fiber group. In matrix notation $R_a(\gamma )=T_a\gamma$, andThis minus sign is the right-invariant vector fields realize the opposite Lie algebra rule. It is essential here. Define the coordinate horizontal lifts of a principal connection byTheir projections are ${\pi }_{\ast }{H}_i={\partial }_i$, so they are linearly independent and there are exactly $\dim B$ of them. Also $d\gamma (H_i)=-A_i\gamma$, whenceThey therefore span the horizontal distribution of a principal connection on this trivialization.

The coefficients $A_i^a$ depend only on the base variables, so taking Lie brackets giveswhere $F_{ij}={\partial }_i{A}_j-{\partial }_j{A}_i+[A_i,A_j]$ and $F=\frac{1}{2}{F}_{ij}d{x}^i\wedge d{x}^j=dA+A\wedge A$. Since the right-invariant vector fields form a basis on each fiber, $\boxed{[H_i,H_j]=0\text{ for all }i,j⟺F=0}$. This is the local coordinate version of vanishing curvature of a principal connection and, by the Frobenius theorem, integrability of the horizontal distribution.

The coordinate qualification is necessary: horizontal lifts of arbitrary noncommuting base fields need not commute even for a flat principal connection. Nor does flatness guarantee a globally defined coordinate frame or a global horizontal section; holonomy of a connection can obstruct the latter. If the requested frame were read globally, the trivial flat bundle $S^2\times G$ would already be a counterexample: restricting a global horizontal frame to the identity-fiber section would trivialize $T{S}^2$, contrary to the Hairy ball theorem. The construction on each coordinate trivialization supplies the intended result without that global claim.

Fix the convention ${\iota }_{X_f}\omega =-df$ for a Hamiltonian vector field. Nondegeneracy of the symplectic form gives one unique smooth $X_f$ for each smooth real-valued function $f$. Define the Poisson bracket byIn coordinates with $\omega =\sum d{q}^i\wedge d{p}_i$, this is $\sum (f_{q^i}{g}_{p_i}-f_{p_i}{g}_{q^i})$. Thus the sign convention is explicit and agrees with the usual coordinate Poisson bracket. By Cartan's magic formula, ${\mathcal{L}}_{X_f}\omega =d{\iota }_{X_f}\omega +{\iota }_{X_f}d\omega =0$. The contraction–Lie derivative commutator identity now givesHence the Hamiltonian Lie algebra homomorphism is $\boxed{\Phi (f)=X_f,[X_f,X_g]=X_{\{f,g\}}}$. It is linear and onto the space of Hamiltonian vector fields by definition. For completeness, the Jacobi identity for the bracket on functions follows from closure of $\omega$: evaluating $d\omega =0$ on $X_f,X_g,X_h$ and using the displayed commutator relation gives the cyclic Jacobi sum zero. Thus this really is a map of Lie algebras, not just a bracket-preserving notation.

Its kernel is determined by nondegeneracy:On a connected manifold these are precisely the real constants; on a disconnected manifold the constant may differ on each component. Therefore the quotient by these functions is isomorphic to the Lie algebra of Hamiltonian vector fields. The printed “homeomorphis” is interpreted as homomorphism: the map before taking this quotient is not an isomorphism because of its nontrivial kernel. The alternative convention ${\iota }_{X_f}\omega =df$ requires a corresponding bracket sign change to retain this homomorphism.

Differentiate the circle group action at angle zero. The resulting real vector field isIn real coordinates $z=x+iy$, this is $K=-y{\partial }_x+x{\partial }_y$; the complex coefficients are conjugates, so it is indeed real. Its integral curves satisfy $\dot{z}=iz$ and hence $z(t)=e^{it}z(0)$, with period $2\pi$ away from the fixed points.

The affine coordinate does not cover the whole complex projective line, so check the other chart. With $w=1/z$, the same vector field is $K=-iw{\partial }_w+i\bar{w}{\partial }_{\bar{w}}$, smooth at $w=0$. It vanishes at $z=0$ and $w=0$, the two fixed poles. It therefore generates the stated global rotation of $S^2$, not merely a flow in one chart.

Set $s=|z{|}^2$. Contraction of the symplectic form with the rotation vector field givesWith the convention fixed in the unheaded solution, ${\iota }_K\omega =-dH$, so the Hamiltonian function isIts derivative is $dH=ds/(1+s{)}^2$. In the chart $w=1/z$, $H=1/(1+|w{|}^2)+C$, so it extends smoothly across infinity; the additive constant is the kernel freedom already identified. Thus $K=X_H$ globally, proving that the action is a Hamiltonian group action. With $C=0$, the moment map for rotation of the complex projective line takes values in $[0,1]$, reaching its endpoints at the two fixed poles.

One can check the normalization in polar coordinates: $\omega =2r(1+r^2{)}^{-2}dr\wedge d\varphi$ has total symplectic area $2\pi$, not $4\pi$, and $H=r^2/(1+r^2)$. Equivalently, in a polar angle with $r=\tan (\vartheta /2)$, $H=(1-\cos \vartheta )/2$. This explains the half-height factor that would be lost by replacing the printed form with the unscaled round-sphere area form. Under the opposite convention ${\iota }_{X_H}\omega =dH$, the Hamiltonian is the negative of this function, up to a constant.