gpt 6 astra ultra attempt on Math-313

Marked Length Spectrum Rigidity Conjecture

Definitions and mathematical statement

Let $M$ be a closed manifold and let $g,h$ be negatively curved Riemannian metrics. Each nontrivial free homotopy class $[\gamma]$ of loops has a unique closed geodesic representative for each metric. Its length defines the marked length spectrum $\ell_g([\gamma])$. The conjecture asserts \[ \ell_g([\gamma])=\ell_h([\gamma])\text{ for every }[\gamma] \quad\Longrightarrow\quad \exists f:M\to M\text{ isotopic to }\mathrm{id},\quad f^*h=g. \] The marking identifies classes in the same manifold, not just an unordered collection of numerical lengths. The isotopy requirement is part of the supplied target and should not be dropped in favour of an arbitrary isometry. Negative curvature refers to sectional curvature in every tangent two-plane.

Formulation and concept references: [S1].

Short English statement

Do the lengths of closed geodesics, labelled by their loop classes, determine a negatively curved metric up to an isometry isotopic to the identity?

Research attempt

Literal catalogue formulation disproved by an attributed high-dimensional construction; standard homotopy-rigidity question unresolved here. Scope and a formulation obstruction. The word "isotopic" in the supplied statement changes the problem in high dimensions. This section gives a counterexample to that literal conclusion, conditional only on explicitly identified established topology theorems. It does not disprove the usual rigidity question with "homotopic to the identity" in the conclusion. In the examples below an isometry homotopic to the identity exists by construction. The original statement above is retained so that this difference remains visible.

The external input is the following consequence of Farrell--Jones [E1]: for a closed connected real hyperbolic manifold $M$ of dimension $m>10$, the group $\pi_0\operatorname{Diff}(M)$ of smooth isotopy classes is infinite. The reference computes the infinite group $\pi_0\operatorname{Top}(M)$ in Corollary 10.16 and discusses the stable splitting from diffeomorphisms in equation $10.28$. Here compact support imposes no restriction because $M$ is closed. Butt [E2], Lemma 2.43 and Remark 2.44, explicitly records both the finite quotient by diffeomorphisms homotopic to the identity and this infinite quotient by diffeomorphisms isotopic to the identity. We use only the real hyperbolic case of the older theorem; no extension to all negatively curved metrics is needed. Closed real hyperbolic manifolds exist in these dimensions. These topology and existence inputs are cited results, not proved by the elementary argument that follows.

An abstract pullback obstruction. Let $(M,g)$ be any closed Riemannian manifold for which the following hold:

1. the isometry group $I(g)$ is finite;
2. infinitely many smooth isotopy classes contain diffeomorphisms homotopic to the identity.

Choose a diffeomorphism $\phi$ homotopic to the identity whose isotopy class is not the class of any member of $I(g)$. There are only finitely many forbidden classes, so the choice is possible. Put $h=\phi^*g$. Pullback preserves smoothness, sectional curvature, and compactness; in particular, if $g$ has constant curvature $-1$, so does $h$.

For any piecewise smooth loop $c$, \[ L_h(c)=L_g(\phi\circ c). \] Since $\phi$ is homotopic to the identity, the loops $c$ and $\phi\circ c$ have the same free homotopy class. Taking infima over that class, and using the bijection of its loops induced by $\phi$, gives \[ \ell_h([c])=\ell_g([c]). \] Negative curvature ensures these infima are the lengths of the specified unique geodesic representatives. Thus the marking is preserved, not merely the unmarked set of lengths.

Suppose the supplied conclusion held: there were $f$ smoothly isotopic to the identity with $f^*h=g$. Pullback reverses the order of composition, so \[ g=f^*(\phi^*g)=(\phi\circ f)^*g. \] Consequently $\phi\circ f\in I(g)$. An isotopy from $f$ to the identity, composed on the left with $\phi$, is an isotopy from $\phi\circ f$ to $\phi$. This contradicts the choice of $\phi$. Notice that selecting an arbitrary nontrivial isotopy class in the homotopy kernel would not by itself suffice: one must exclude the classes realized by isometries. The finiteness/infinity argument makes that exclusion explicit.

Checking the two hypotheses for the hyperbolic example. Write \[ G=\operatorname{Diff}(M)/\operatorname{Diff}^{\rm iso}_0(M),\qquad H=\operatorname{Diff}^{\rm hom}_0(M)/\operatorname{Diff}^{\rm iso}_0(M). \] The first denominator consists of diffeomorphisms joined to the identity by a smooth isotopy; the superscript "hom" means only homotopic as maps. These are normal subgroups. The quotient $G/H$ is finite by the cited homotopy-class finiteness theorem. If $H$ were finite, the finite union of its cosets would make $G$ finite. Farrell--Jones gives the opposite, so $H$ is infinite. This supplies the second hypothesis without identifying an explicit coordinate formula for one of the exotic diffeomorphisms.

For completeness, finiteness of $I(g)$ follows from standard compact-isometry group theory and a short curvature calculation. The isometry group of a closed Riemannian manifold is a compact Lie group. An infinitesimal isometry is a Killing field $X$. The integrated Killing-field identity is \[ \int_M|\nabla X|^2\,dV=\int_M\operatorname{Ric}(X,X)\,dV. \] For a real hyperbolic $m$-manifold, the right side is $-(m-1)\int_M|X|^2\,dV$. Both sides can agree only when $X=0$. The compact Lie group therefore has zero-dimensional Lie algebra, is discrete, and hence is finite. This argument uses the customary smooth isometry-group theorem and the Bochner identity, neither a claim about arbitrary homeomorphisms nor an assertion that every homotopy-trivial diffeomorphism is an isometry.

It follows that a closed real hyperbolic manifold in any dimension greater than ten supplies a pair $(g,\phi^*g)$ satisfying the hypothesis and violating the isotopy conclusion. The inverse $\phi^{-1}$ is nevertheless homotopic to the identity and obeys $(\phi^{-1})^*h=g$. Therefore the very same example satisfies the standard homotopy version of rigidity.

What the remaining geometric rigidity question asks. The obstruction is not a failure to reconstruct lengths accurately. Lengths labelled by free homotopy classes are invariant under the entire group $\operatorname{Diff}^{\rm hom}_0(M)$, so such data cannot distinguish its different isotopy components. One cannot repair the proof by extracting more closed-orbit lengths from exactly the same marking. A formulation using a homotopy-trivial isometry, or an additional restriction making the two relevant quotients agree, removes this particular obstruction. Neither modification is silently substituted for the inherited statement here.

There are useful elementary checks on potential arguments for the geometric version. If $h=a^2g$ with $a>0$, then every length is multiplied by $a$. On a closed negatively curved manifold there is a nontrivial closed geodesic, so equal marked lengths force $a=1$. Thus rescaling cannot furnish a competing counterexample. Equality of volume alone would be substantially weaker than equality of every marked length and cannot replace it.

A second check concerns pointwise ordered metrics. Suppose $h\geq g$ as quadratic forms and their marked lengths agree. If $c$ is the closed $h$-geodesic in a nontrivial class, then \[ \ell_g([c])\leq L_g(c)\leq L_h(c)=\ell_h([c])=\ell_g([c]). \] Both inequalities are equalities. The continuous nonnegative function $|\dot c|_h-|\dot c|_g$ has integral zero, so it vanishes at every point of this orbit. Using the standard density of periodic geodesic-flow directions for a closed negatively curved metric, it follows by continuity that $(h-g)(v,v)=0$ on every tangent direction. Polarization gives $h=g$. Here periodic-direction density is an external dynamical theorem, and the pointwise ordering is an extra hypothesis. The pullback counterexamples need not satisfy it; an arbitrary pair of metrics cannot be ordered by adding a constant without changing its length spectrum.

Linearization and why it does not remove the topology. For a differentiable family $g_t=g+tq+O(t^2)$ of negative-curvature metrics, the first variation of the length in a fixed free homotopy class is \[ \left.\frac{d}{dt}\right|_{t=0}\ell_{g_t}([c]) =\frac12\int_c q(\dot c,\dot c)\,ds, \] with $c$ parametrized at unit $g$-speed. The curve-variation term vanishes because $c$ is a closed geodesic. If $q=\mathcal L_Vg$, then $q(\dot c,\dot c)=2\frac{d}{ds}g(V,\dot c)$, whose integral is zero. This proves the unavoidable infinitesimal gauge kernel directly. It does not show that this is the entire kernel, and even an injectivity theorem after fixing this gauge would address a neighborhood of a metric, not all components of the diffeomorphism group. The exotic pullbacks above are a global isotopy-component obstruction, so an infinitesimal inverse function theorem cannot by itself establish the supplied universal claim.

Verification and outcome. The offline diagnostic checks pullback composition on positive definite matrices and an elementary finite quotient example illustrating the coset argument. These are algebraic sanity checks only; the infinitude assertion comes from the cited topology, not a finite computation. The source audit checked the original Farrell--Jones text and the explicit homotopy/isotopy warning in Butt. On those established inputs, the literal formulation is disproved in high dimensions. The general marked-length-spectrum rigidity problem with a homotopy conclusion remains outside what this argument proves.

Assessment of the claimed solution

The argument above is mathematically sound as a counterexample to the literal formulation stated in this document, namely the formulation requiring an isometry that is smoothly isotopic to the identity. It should not, however, be interpreted as a solution or disproof of the standard Marked Length Spectrum Rigidity Conjecture.

The essential distinction is between being homotopic to the identity and being isotopic to the identity. Suppose that $(M,g)$ is a closed real hyperbolic manifold of dimension greater than ten for which there exists a diffeomorphism \[ \phi:M\to M \] that is homotopic to the identity but whose smooth isotopy class is not represented by any element of the finite isometry group $I(g)$. The cited high-dimensional topology results of Farrell--Jones [E1], together with the homotopy/isotopy distinction recorded explicitly by Butt [E2], provide the required supply of such isotopy classes.

Set \[ h=\phi^*g. \] Because $\phi$ is homotopic to the identity, it preserves every free homotopy class of loops. Hence, for every nontrivial free homotopy class $[\gamma]$, \[ \ell_h([\gamma]) = \inf_{c\in[\gamma]} L_h(c) = \inf_{c\in[\gamma]} L_g(\phi\circ c) = \inf_{c\in[\gamma]} L_g(c) = \ell_g([\gamma]). \] Thus $g$ and $h$ have identical marked length spectra.

Now suppose that there were a diffeomorphism $f$ smoothly isotopic to the identity such that \[ f^*h=g. \] Then \[ g = f^*(\phi^*g) = (\phi\circ f)^*g, \] so $\phi\circ f\in I(g)$. Since $f$ is isotopic to the identity, $\phi\circ f$ is isotopic to $\phi$. This contradicts the choice of $\phi$, whose isotopy class was assumed not to contain any $g$-isometry. Therefore the conclusion requiring an isometry isotopic to the identity fails.

However, this does not produce two nonisometric metrics. Indeed, \[ (\phi^{-1})^*h = (\phi^{-1})^*(\phi^*g) = g, \] and $\phi^{-1}$ is homotopic to the identity. Hence the two metrics are isometric by an isometry preserving the marking up to homotopy.

This distinction is important because the usual marked length spectrum rigidity problem is formulated in terms of a marking-preserving isometry, equivalently an isometry homotopic to the identity, rather than necessarily isotopic to the identity. The AIM problem list [S1] asks whether equality of the marked length spectra forces the metrics to be isometric, while [E2] makes the relevant homotopy formulation explicit.

Consequently, the result established above should be described more precisely as follows:

In dimensions greater than ten, the strengthened formulation of marked length spectrum rigidity requiring the realizing isometry to be smoothly isotopic to the identity is false. The standard rigidity problem with a homotopy-to-the-identity conclusion is not disproved by this construction.

Accordingly, the label \[ \texttt{ATTEMPT\_STATUS: solved} \] would be misleading if it is intended to refer to the standard Marked Length Spectrum Rigidity Conjecture. A more accurate description is:

Counterexample to the isotopy-to-the-identity strengthening of marked length spectrum rigidity in dimensions greater than ten; the standard homotopy formulation remains unresolved by this argument.

Sources

• S1 AIM workshop problem list, Open problems and questions about geodesics. https://aimath.org/pastworkshops/geodesicsproblems.pdf. Formulation source.
• E1 F. T. Farrell and L. E. Jones, A Topological Analogue of Mostow's Rigidity Theorem, Journal of the American Mathematical Society 2 $1989$, 257--370; Corollary 10.16 and equation $10.28$, pp. 350--351. https://math.uchicago.edu/~shmuel/tom-readings/FJ%20Hyperbolic.pdf. Published topology input; the argument uses closed real hyperbolic manifolds of dimension greater than ten.
• E2 Karen Butt, Quantitative marked length spectrum rigidity, Geometry & Topology 29 $2025$, Lemma 2.43 and Remark 2.44. https://arxiv.org/html/2203.12128. Explicit distinction between the homotopy quotient and the isotopy quotient of the diffeomorphism group.