---
title: Llarull's Scalar Curvature Rigidity Theorem
url: https://www.emergentmind.com/topics/llarull-s-scalar-curvature-rigidity-theorem
type: topic
---

# Llarull's Scalar Curvature Rigidity Theorem

Llarull’s scalar curvature rigidity theorem is a comparison theorem for positive scalar curvature that characterizes the round sphere by combining a lower scalar-curvature bound with a metric-comparison map to the unit sphere. In its classical sphere formulation, if \(g\) is a smooth metric on \(S^n\) satisfying \(g\ge g_0\) and \(\mathrm{Scal}_g\ge n(n-1)\), where \(g_0\) is the standard round metric of sectional curvature \(1\), then \(g\) is isometric to \(g_0\). In its more flexible form, if \((M^n,g)\) is a closed spin manifold and \(f\colon M\to S^n\) is area-non-increasing with \(\deg(f)\neq0\), then the lower bound \(\mathrm{scal}_g\ge n(n-1)\) forces \(f\) to be a Riemannian covering, and in degree \(\pm1\) an isometry onto the round sphere. The theorem is a prototype of scalar-curvature extremality, and many later developments treat stability, low regularity, singular metrics, and alternative model spaces [2310.14412, 2206.11796].

## 1. Classical statement and equivalent formulations

The unit round sphere \((S^n,g_0)\) has scalar curvature \(n(n-1)\). In the formulation most directly tied to the sphere itself, a smooth metric \(g\) on \(S^n\) is said to dominate \(g_0\), written \(g\ge g_0\), if \(g_x(v,v)\ge g_{0,x}(v,v)\) for every \(x\in S^n\) and every tangent vector \(v\in T_xS^n\). Llarull’s theorem then states that if \(n\ge2\), \(g\) is smooth, \(\mathrm{Scal}_g(x)\ge n(n-1)\) for all \(x\in S^n\), and \(g\ge g_0\) everywhere, then \(g\) is isometric to the round metric \(g_0\) [2405.19724].

A more invariant version replaces the identity map on the sphere by a map from an arbitrary closed spin manifold. Let \((M^n,g)\) be closed and spin, let \(f\colon M\to S^n\) be smooth, and assume that \(f\) is area-non-increasing, meaning \(|\Lambda^2d_xf|\le1\) at each point, or equivalently that the \(2\)-dimensional Jacobians are non-expanding. If \(\deg(f)\neq0\) and \(\mathrm{scal}_g\ge n(n-1)\), then Llarull’s conclusion is that \(g=f^*g_0\) and \(f\) is a Riemannian covering; when \(\deg(f)=\pm1\), \(f\) is an isometry [2310.14412].

These formulations encode the same rigidity principle. The identity map \(\mathrm{id}\colon (S^n,g)\to(S^n,g_0)\) is \(1\)-Lipschitz precisely when \(g\ge g_0\), and in dimension \(n\ge4\) the hypothesis can also be expressed by the a.e. condition \(|\Lambda^2df|\le1\). The theorem therefore asserts that one cannot simultaneously keep the round sphere as a metric lower bound and raise scalar curvature to at least the round value, except in the trivial case where the metric is already round [2206.11796].

## 2. Spinorial mechanism and the rigidity proof

The standard proof is spinorial. One twists the Dirac operator on \(M\) by the pull-back of the spinor bundle of the round sphere. In the smooth setting the Bochner–Lichnerowicz–Weitzenböck formula takes the form
\[
D^2=\nabla^*\nabla+\tfrac14\,\mathrm{Scal}_g
\]
for the untwisted operator, and the twisted operator satisfies an analogous Schrödinger–Lichnerowicz formula with an additional curvature term coming from the pull-back bundle. The scalar-curvature lower bound enters through the \(\tfrac14\,\mathrm{Scal}_g\) term, while the area-non-increasing hypothesis bounds the twisting curvature from below [2405.19724, 2206.11796].

The topological input is the nonvanishing of an index. In even dimensions, the twisted Dirac operator has nonzero index, explicitly \(2\,\deg(f)\) in the low-regularity formulation of Cecchini–Hanke–Schick, so there exists a nonzero harmonic spinor \(\psi\). The integral Schrödinger–Lichnerowicz identity then yields
\[
\bigl(D\psi,D\psi\bigr)
=\|\nabla\psi\|^2
+\tfrac14\,\bigl\langle\!\langle\Scal(g),|\psi|^2\bigr\rangle\!\bigr\rangle
+\bigl(\mathcal R^E\psi,\psi\bigr),
\]
and at differentiability points of \(f\) the twist curvature satisfies
\[
\langle\mathcal R^E_x\omega,\omega\rangle
\ge -\tfrac14\sum_{i\ne j}\mu_i\mu_j\,|\omega|^2
\ge -\tfrac14\,n(n-1)\,|\omega|^2,
\]
where the \(\mu_i\) are the singular values of \(df\circ b_g\). Under \(\mathrm{Scal}(g)\ge n(n-1)\) and \(|\Lambda^2df|\le1\), all nonnegative terms are forced to vanish, so \(\nabla\psi=0\) and the equality case implies that \(d_xf\) is an orientation-preserving isometry at almost every point. Reshetnyak theory for quasiregular maps then upgrades the a.e. statement to a global metric isometry [2206.11796].

Odd dimensions require a different topological package because the ordinary Fredholm index vanishes. One route is to pass to \(M\times S^1_r\) and then slice, as in the stability paper; another is to construct a one-parameter family of self-adjoint twisted Dirac operators \(\{D_s\}_{s\in[0,1]}\) and compute its spectral flow. In the odd-dimensional spectral-flow proof, Getzler’s theorem identifies
\[
\mathrm{sf}\{D_s\}=-\,\deg(f),
\]
while the Schrödinger–Lichnerowicz estimate shows that strict inequality in the curvature comparison would make every \(D_s\) invertible, forcing zero spectral flow. The contradiction leaves only the equality case, from which local and then global isometry follow [2306.06906].

## 3. Equality, hypotheses, and common points of confusion

The theorem’s hypotheses are structurally coupled. The spin assumption is built into the classical Dirac-operator proof and into many extensions. In the map version, the nonzero degree assumption is not merely technical: it is the device that produces nontrivial index or nontrivial spectral flow, hence the harmonic spinor that detects rigidity. The area-non-increasing condition is equally essential on the analytic side, because it is what controls the curvature endomorphism in the twisted Weitzenböck formula [2310.14412, 2306.06906].

A recurring misconception is that surjectivity of \(f\colon M\to S^n\) might replace \(\deg(f)\neq0\). That replacement fails for scalar-curvature rigidity in dimensions \(n\ge3\). For every \(S_0>0\) and every \(\lambda>0\), if \(M^n\) admits some metric of positive scalar curvature and \(n\ge3\), there exists a metric \(g_M\) with \(\mathrm{scal}_{g_M}\ge S_0\) and a smooth surjective \(\lambda\)-Lipschitz map \(f\colon (M,g_M)\to (N,g_N)\). Taking \(N=S^n\), \(S_0=n(n-1)\), and \(\lambda=1\) shows that a surjective \(1\)-Lipschitz map with \(\mathrm{scal}_g\ge n(n-1)\) and \(\deg(f)=0\) need not be rigid. By contrast, for Ricci curvature the surjective analogue is valid in all dimensions, and in dimension \(2\) surjectivity suffices in the scalar-curvature setting as well [2507.05459].

Another subtle point is the relationship between equality and global isometry. The spinorial argument first yields pointwise saturation of the curvature estimate and hence pointwise isometry of differentials. The global conclusion then uses topology: nonzero degree, quasiregularity without branch points, or simple connectedness of the target imply that the map is a covering, and on the sphere this becomes a global isometry in degree \(\pm1\) [2206.11796, 2310.14412].

## 4. Quantitative stability and Gromov’s spherical stability problem

Llarull’s theorem is a gap theorem at the threshold \(n(n-1)\). A natural question, emphasized by Gromov, is whether a small scalar-curvature deficit forces an approximate spherical geometry. The all-dimensional stability theorem answers this affirmatively. If
\[
R[g]\ge n(n-1)-\varepsilon
\]
under the same spin, degree, and area-non-increasing hypotheses, then \(M\) admits a decomposition
\[
M=\mathcal G\cup\mathcal B_1\cup\mathcal B_2
\]
into a good set and two bad sets, with
\[
|g-f^*g_0|_g\le\sqrt{\varepsilon}\quad\text{on }\mathcal G,
\]
and
\[
\bigl|\partial\mathcal B_1\bigr|_g+\bigl|\mathcal B_2\bigr|_g+\bigl|S^n\setminus f(\mathcal G)\bigr|_{g_0}
\le C(n)\,N\,\sqrt\varepsilon,
\]
where \(N=\max_p\#(f^{-1}(p))\) over regular \(p\in S^n\). When \(f=\mathrm{Id}\colon S^n\to S^n\), the statement is slightly stronger and no \(N\)-factor appears. In this precise sense the manifold is \(C^0\)-close to a finite number of spheres outside a small bad set, and the paper states that this completely solves Gromov’s spherical stability problem [2310.14412].

The proof remains spinorial but becomes quantitative. Its main ingredients are a twisted harmonic spinor, an integral inequality of Llarull type, a decomposition into good and bad sets defined through the singular values of \(df\), coarea and level-set arguments applied to \(|\phi|\), and Poincaré or Cheeger estimates on the sphere. A Sobolev-refined version replaces the Poincaré/Cheeger step by an \(L^{\alpha/2}\)-bound on the negative part of \(R-n(n-1)\) together with a Sobolev constant hypothesis, yielding an explicit estimate for the bad set volume while still giving \(C^0\)-closeness off that set [2310.14412].

A distinct three-dimensional stability theory predates the full-dimensional result. For sequences of Riemannian \(3\)-spheres whose distance functions are bounded below by the unit sphere’s, with uniformly bounded Cheeger isoperimetric constant and scalar curvatures tending to \(6\), the metrics converge to the round \(3\)-sphere in the volume preserving Sormani–Wenger Intrinsic Flat sense. That argument uses the spacetime-harmonic-function proof of Llarull’s theorem in dimension three rather than the spinorial proof [2305.18567].

## 5. Low-regularity, punctured, and singular versions

A major development is that smoothness of the metric and map is not essential to rigidity. Cecchini–Hanke–Schick proved that if \(M\) is closed, smooth, connected, spin, and even-dimensional, if \(g\in W^{1,p}\) with \(p>n\) has distributional scalar curvature in the sense of Lee–LeFloch satisfying \(\Scal(g)\ge n(n-1)\), and if \(f\colon (M,d_g)\to S^n\) is Lipschitz of nonzero degree, then \(f\) is a metric isometry. For \(n\ge4\) the map need only be area-nonincreasing a.e.; for \(n=2\) one assumes \(f\) is \(1\)-Lipschitz. In particular, \(f\) is bijective, \(C^\infty\), and \(f^*g_0=g\) [2206.11796].

There is also a punctured \(L^\infty\) theory. For \(n>3\), let \(M\) be a closed spin manifold, let \(S\subset M\) be finite, and let \(g\) be an \(L^\infty\) metric on \(M\) that is smooth on \(M\setminus S\). If \(f\colon M\to S^n\) is a Lipschitz map of nonzero degree, smooth on \(M\setminus S\), with \(\mathrm{Scal}_g(x)\ge n(n-1)\) on \(M\setminus S\), and if \(f\) is area non-increasing there, then \(f\) is a distance isometry and \(f^*g_{\mathrm{round}}=g\) on \(M\setminus S\). In the identity-map corollary for \(M=S^n\), the conclusion is \(g=g_{\mathrm{round}}\) on all of \(S^n\) [2405.19724].

Singular rigidity extends further to cone-like singularities in odd dimensions. If \(N^{2k+1}\) is compact, connected, oriented, spin, with finitely many cone-like singularities, if \(\scal_G\ge (2k+1)2k\) as a distribution, and if \(f\colon (N,G)\to(S^{2k+1},g_0)\) is Lipschitz, a.e. area-nonincreasing, and of nonzero degree, then \(f\) is a smooth Riemannian isometry away from the singular points. The result goes on to state that \(N\) must be diffeomorphic to a punctured sphere and that \(f\) extends to a global isometry onto \(S^{2k+1}\). Analytically, this combines twisted Dirac operators on singular manifolds, abstract cone operators, and spectral flow [2605.00151].

## 6. Generalizations and surrounding rigidity theory

Llarull’s theorem has become a model for a broader scalar-curvature extremality program. One direction generalizes the target. Goette–Semmelmann’s area-rigidity theorem, as reformulated through higher mapping degree, replaces the sphere by a closed connected manifold \(N\) with nonnegative curvature operator, strict pinching \(0<2\,\mathrm{Ric}_N<\mathrm{scal}_N\), and \(\chi(N)\neq0\). If \(f\colon M\to N\) is area nonincreasing and the higher mapping degree \(\deg_{hi}(f)\) is nonzero, then \(\mathrm{scal}_M\ge \mathrm{scal}_N\circ f\) forces equality and, under the stated pinching, \(f\) becomes a Riemannian submersion. This includes examples such as \(\mathrm{pr}_1\colon S^{2n}\times T^k\to S^{2n}\) [2402.05834].

Another direction weakens or reshapes the curvature hypothesis. A spectral Llarull theorem replaces \(R_g\ge n(n-1)\) on \(S^n\) by the lower bound
\[
\lambda_1\!\left(-\Delta_g+\frac{n-2}{4(n-1)}R_g\right)\ge n,
\]
still under \(g\ge \bar g\), and concludes \(g=\bar g\). The same paper treats warped-product models using spinor and spacetime-harmonic-function methods [2310.13329]. Listing’s earlier improvement on even-dimensional spheres shows that if \(g\) is a metric on \(S^{2m}\) with
\[
\mathrm{scal}_g\ge (2m-1)\,\mathrm{tr}_g(g_0),
\]
then \(g\) is a constant multiple of the round metric \(g_0\) [1007.1832].

Further extensions preserve the Llarull pattern while changing the model space. Products \(S^{n-m}\times\mathbb T^m\) satisfy Llarull-type rigidity under a degree-nonzero map whose spherical projection is area non-increasing, together with the additional scalar inequality involving a function \(\psi\); when \(m=0\), the theorem recovers the sphere case [2511.04407]. In odd dimensions, the spectral-flow method also yields rigidity when the sphere is replaced by a smooth strictly convex closed hypersurface in \(\mathbb R^{n+1}\) [2306.06906]. A later family-index approach treats products of convex hypersurfaces and, in the presence of circle factors, gives the corresponding optimal splitting theorem [2606.15710]. Parallel developments include Llarull-type theorems for three- and four-dimensional bands via warped \(\mu\)-bubbles [2602.21813], rigidity for domains in three-dimensional spherical warped products [2503.04025], and weighted versions in which the ordinary scalar curvature is replaced by \(P\)-scalar curvature [2511.12517].

Taken together, these results show that Llarull’s theorem is not an isolated sphere theorem but a central rigidity paradigm. Its characteristic structure is the same across settings: a curvature lower bound at a model threshold, a non-expanding map of nontrivial topological type, a Dirac or variational identity with a sharp equality case, and a conclusion that forces the comparison map to be an isometry or a rigid splitting.

Source: https://www.emergentmind.com/topics/llarull-s-scalar-curvature-rigidity-theorem