[proofplan] We separate the compact shrinking and noncompact steady alternatives, applying the classification theorems to a representative soliton time-slice metric rather than to the whole ancient interval. In the compact case, the gradient shrinking soliton equation and positive two-dimensional curvature force constant positive Gaussian curvature, and the only smooth compact spherical quotients are $S^2$ and $\mathbb{RP}^2$. In the noncompact steady case, Hamilton's classification theorem for complete two-dimensional steady gradient Ricci solitons with bounded positive curvature identifies the metric with Hamilton's cigar soliton up to scaling. Finally, the soliton structure converts the representative model metric into the corresponding ancient Ricci-flow solution by scaling and pullback along the soliton diffeomorphisms. [/proofplan] [step:Apply compact shrinking soliton rigidity to force constant curvature] Assume first that $M$ is compact and that the soliton is shrinking. Let $f: M \to \mathbb{R}$ be a smooth potential function for the gradient Ricci soliton structure, so that for each time slice the metric satisfies the shrinking soliton equation \begin{align*} \operatorname{Ric}_{g} + \operatorname{Hess}_g f = \lambda g \end{align*} for a constant $\lambda > 0$, where $g$ denotes a representative time-slice metric and $\operatorname{Hess}_g f$ is the Hessian with respect to $g$. In dimension two, positive scalar curvature is equivalent to positive Gaussian curvature because $S_g = 2K_g$, where $S_g$ is scalar curvature and $K_g$ is Gaussian curvature. Thus $M$ is compact, two-dimensional, smooth, complete, and has positive Gaussian curvature, so all hypotheses of the [compact surface shrinking-soliton rigidity theorem](https://doi.org/10.1090/conm/071/954419) are satisfied. The theorem states that every compact two-dimensional gradient shrinking Ricci soliton with positive curvature has constant positive Gaussian curvature. Hence $g$ has constant positive Gaussian curvature. A compact complete two-dimensional Riemannian manifold with constant positive Gaussian curvature is a quotient of the round sphere by a finite group of round isometries acting freely. A finite subgroup of the isometry group of $S^2$ acting freely is either trivial or generated by the antipodal map; any non-antipodal rotation has fixed points and would produce an orbifold point rather than a smooth surface. Therefore the compact shrinking model is the round shrinking sphere when the [quotient group](/page/Quotient%20Group) is trivial, and otherwise the constant-curvature quotient $\mathbb{RP}^2$. [guided] We first isolate exactly which structure is being used. The phrase “gradient shrinking Ricci soliton” means that there is a smooth map $f: M \to \mathbb{R}$ and a constant $\lambda > 0$ such that a time-slice metric $g$ satisfies \begin{align*} \operatorname{Ric}_{g} + \operatorname{Hess}_g f = \lambda g. \end{align*} Here $\operatorname{Ric}_{g}$ is the Ricci tensor of $g$, and $\operatorname{Hess}_g f$ is the Hessian of $f$ computed using the Levi-Civita connection of $g$. We apply the [compact surface shrinking-soliton rigidity theorem](https://doi.org/10.1090/conm/071/954419). Its hypotheses are: the manifold is compact, the dimension is two, the metric is a gradient shrinking Ricci soliton, and the curvature is positive. Its conclusion is that such a surface soliton has constant positive Gaussian curvature. These hypotheses are exactly the assumptions in the compact shrinking branch of the theorem: $M$ is compact, $M$ has dimension $2$, the flow is a gradient Ricci soliton in the shrinking case, and the curvature is bounded positive curvature. If the curvature hypothesis is stated as positivity of scalar curvature, then in dimension two this gives positive Gaussian curvature because $S_g = 2K_g$. The theorem therefore gives that the Gaussian curvature of $g$ is constant and positive. Once constant positive curvature is known, the geometric classification is standard: a complete connected surface of constant positive curvature is a spherical space form. Equivalently, it is the quotient of the round sphere by a finite group of round isometries acting freely. In dimension two the free finite possibilities are restricted: every nontrivial orientation-preserving finite isometry of $S^2$ is a rotation and has fixed points, while the antipodal map is the only fixed-point-free nontrivial possibility on a smooth quotient. Hence the compact smooth quotient is either $S^2$ or $\mathbb{RP}^2$ with the induced constant-curvature metric. [/guided] [/step] [step:Verify Hamilton's hypotheses in the noncompact steady case] Assume now that $M$ is noncompact and the soliton is steady. Let $f: M \to \mathbb{R}$ be a smooth potential function for the steady gradient soliton structure, so that a representative time-slice metric $g$ satisfies \begin{align*} \operatorname{Ric}_{g} + \operatorname{Hess}_g f = 0. \end{align*} The theorem hypotheses give that $(M,g)$ is complete, two-dimensional, noncompact, has positive curvature, and has bounded curvature. These are precisely the assumptions required by [Hamilton's classification of complete two-dimensional steady gradient Ricci solitons](https://doi.org/10.1090/conm/071/954419). [guided] We now check the noncompact branch in the same explicit way. The steady gradient Ricci soliton structure means that there is a smooth map $f: M \to \mathbb{R}$ such that the representative time-slice metric $g$ satisfies \begin{align*} \operatorname{Ric}_{g} + \operatorname{Hess}_g f = 0. \end{align*} Hamilton's classification theorem for complete two-dimensional steady gradient Ricci solitons requires a smooth complete noncompact surface, a steady gradient soliton equation, bounded curvature, and positive curvature. These conditions are supplied by the present hypotheses: completeness is assumed for the ancient Ricci flow and hence for each time-slice metric, $M$ is noncompact in this branch, the displayed equation is exactly the steady gradient soliton equation, and bounded positive curvature is part of the theorem statement. Therefore Hamilton's classification applies to $(M,g,f)$ and identifies $(M,g)$ with Hamilton's cigar metric after multiplying the metric by a positive constant and pulling back by a diffeomorphism. The ancient-time assumption is not an additional input to this classification step; it ensures that the soliton-generated Ricci flow exists on the ancient interval under discussion. [/guided] [/step] [step:Identify the noncompact steady metric with the cigar soliton] By [Hamilton's classification of complete two-dimensional steady gradient Ricci solitons](https://doi.org/10.1090/conm/071/954419) applied to $(M,g,f)$, there exist a scaling constant $a > 0$ and a diffeomorphism $\Phi: M \to \mathbb{R}^2$ such that \begin{align*} g = \Phi^*(a g_{\mathrm{cig}}), \end{align*} where $g_{\mathrm{cig}}$ denotes Hamilton's cigar metric on $\mathbb{R}^2$. Thus the time-slice metric is Hamilton's cigar soliton up to scaling and pullback by a diffeomorphism. [/step] [step:Recover the Ricci-flow model from the soliton metric] For a gradient steady Ricci soliton satisfying $\operatorname{Ric}_{g} + \operatorname{Hess}_g f = 0$, the Ricci flow generated by the soliton metric evolves only by pullback along the one-parameter family of diffeomorphisms generated by the vector field $-\nabla f$, with the opposite sign convention corresponding to reversing the parametrisation of the same pullback family. Thus the ancient-time-interval hypothesis is used here only to regard this soliton-generated family as the ancient flow in the statement; the metric classification itself was carried out on the representative time slice. Therefore the noncompact steady ancient flow with representative metric $g$ is the cigar Ricci flow model up to the same scaling and pullback by diffeomorphisms. Combining this with the compact shrinking classification proves both alternatives in the statement. [/step]