38.07.04 · dynamics / smooth-ergodic-theory

The Livšic Cohomological Rigidity Theorem

shipped3 tiersLean: none

Anchor (Master): Katok-Hasselblatt 1995 *Introduction to the Modern Theory of Dynamical Systems* (Cambridge) Ch. 19 (the Livšic theorem, periodic-orbit obstruction, regularity); Livšic 1971 *Cohomology of dynamical systems* (Math. USSR Izvestiya 6); Livšic 1972 *Homology properties of Y-systems* (Math. Notes 10); de la Llave-Marco-Moriyón 1986 *Canonical perturbation theory of Anosov systems and regularity results for the Livšic cohomology equation* (Annals of Mathematics 123)

Intuition Beginner

Imagine a chaotic rule that shuffles points around a space, and at each point you read off a number — think of it as the cost charged for landing there. As a point travels under the rule it racks up a running tally: cost at the start, plus cost after one step, plus cost after two steps, and so on. You would like to know whether all this cost is, in a sense, illusory — whether it can be cancelled by a clever change of bookkeeping that simply re-labels the cost at each location.

The change of bookkeeping is this: pick a new value for every point, and replace the original cost at a point by the difference between the new value at the next point and the new value at the current point. If the original cost was secretly of this form all along, then the running tally telescopes — almost everything cancels — and the cost was never doing any real work. A cost that can be re-bookkept away like this is called a coboundary, and the new labelling that does it is the transfer function.

How can you tell, just by looking at the cost, whether such a cancellation is possible? The chaotic rule has special closed loops: points that return exactly to where they started after some number of steps. Around any such loop the bookkeeping change must cancel completely, because you end where you began. So if the cost can be cancelled, its total around every closed loop has to be zero. The Livšic theorem is the striking statement that for the most chaotic systems this obvious necessary check is also sufficient: zero total around every loop is the only thing you need, and the cancelling labelling then exists and is essentially unique.

Visual Beginner

Picture the orbit of a point as a chain of beads, each bead carrying a small number — the cost there. The transfer function paints a value on every bead; the re-bookkept cost on a link is the painted value ahead minus the painted value behind. Along a long open chain these differences telescope, so only the two endpoints survive. Now look at a closed loop of beads that comes back to its start: going all the way around, every painted value is both added and subtracted once, so the painted values cancel exactly — and the only way the original costs could have summed to that is if they summed to zero around the loop.

name	what it is	everyday image
cost	the number read at each point	toll charged for landing there
transfer function	the painted value on each point	a re-labelling of the books
coboundary	a cost that is a difference of painted values	a toll that cancels out
closed loop	a point returning to its start	a round trip
loop total	sum of costs around a loop	net toll for a round trip

Worked example Beginner

Take the doubling rule on four labelled spots arranged so the rule sends spot $1$ to spot $2$ , spot $2$ to spot $3$ , spot $3$ to spot $4$ , and spot $4$ back to spot $1$ — a single closed loop of length four. Suppose the cost read at the four spots is $5, 1, 4, 2$ .

Step 1. Check the loop total. Going once around, the costs add to $5 + 1 + 4 + 2 = 12$ . Since the only loop here has total $12$ , which is not $0$ , the necessary check already fails: this cost cannot be cancelled by any re-labelling. To see the theorem in action, replace the last cost so the loop total is zero — change the four costs to $5, 1, 4, - 10$ , which sum to $5 + 1 + 4 - 10 = 0$ .

Step 2. Build the cancelling labels. Pick the painted value at spot $1$ to be $0$ . The rule says the cost on the link out of a spot equals the painted value ahead minus the painted value behind. So painted value at spot $2$ equals painted value at spot $1$ plus the cost at spot $1$ : $0 + 5 = 5$ . Then spot $3$ : $5 + 1 = 6$ . Then spot $4$ : $6 + 4 = 10$ .

Step 3. Verify the loop closes. The link from spot $4$ back to spot $1$ should carry cost equal to painted value at spot $1$ minus painted value at spot $4$ , namely $0 - 10 = - 10$ — exactly the cost we put there. The books balance all the way around.

What this tells us: the running cost $5, 1, 4, - 10$ was secretly just a difference of the painted values $0, 5, 6, 10$ , so it can be cancelled, and we could reconstruct those values one spot at a time by walking along the orbit and adding up the costs. The whole construction worked precisely because the loop total was zero — the single condition the Livšic theorem says is all that is ever needed.

Check your understanding Beginner

Exercise (easy, multiple choice).

A cost function on the points of a rule is a coboundary when it can be written as:

A. the cost at the next point minus the cost at the current point. B. a painted value at the next point minus the painted value at the current point. C. the same constant at every point. D. zero at every point.

Hint

A coboundary is a re-bookkeeping: it is built from a single painted value (the transfer function), comparing its value ahead with its value behind.

Answer

B. A coboundary is a cost of the form (painted value at the next point) minus (painted value at the current point) — a difference of one transfer function evaluated one step apart. Option A compares the cost at two points rather than a painted value; C and D are special costs (a constant cost is a coboundary only when the constant is zero, and zero is the simplest coboundary), but neither is the definition. The point of the definition is that such a cost telescopes along orbits and cancels around loops.

Formal definition Intermediate+

Let $M$ be a compact Riemannian manifold with distance $d$ , $f : M \to M$ a topologically transitive Anosov diffeomorphism (or, more generally, a homeomorphism with the closing property on a locally maximal topologically transitive hyperbolic set $Λ$ , as in 38.03.01 and 38.03.04). For $0 < α \leq 1$ write $C^{α} (M)$ for the space of $α$ -Hölder functions $φ : M \to R$ with seminorm $[φ]_{α} = sup_{x \neq = y} ∣ φ (x) - φ (y) ∣/ d (x, y)^{α}$ .

Definition (cohomological equation; coboundary). Given $φ : M \to R$ , the cohomological equation for $φ$ over $f$ is $$ \varphi = u \circ f - u, $$ where the unknown transfer function $u : M \to R$ is sought in a prescribed regularity class. A $φ$ admitting a continuous (resp. Hölder) solution $u$ is a continuous (resp. Hölder) coboundary. Two functions $φ, ψ$ are cohomologous if $φ - ψ$ is a coboundary; cohomology is an equivalence relation, and the coboundary class is the zero class.

Definition (Birkhoff cocycle; periodic sum). The Birkhoff sums of $φ$ are $S_{n} φ (x) = \sum_{k = 0}^{n - 1} φ (f^{k} x)$ , with $S_{0} φ = 0$ ; they satisfy the cocycle identity $S_{m + n} φ (x) = S_{m} φ (x) + S_{n} φ (f^{m} x)$ . For a periodic point $p$ of period $n$ (so $f^{n} p = p$ ) the periodic sum or periodic-orbit obstruction is $$ S_n\varphi(p) = \sum_{k=0}^{n-1}\varphi(f^k p). $$

Definition (periodic obstruction vanishes). The function $φ$ satisfies the periodic-orbit condition if $S_{n} φ (p) = 0$ for every $p$ with $f^{n} p = p$ (all $n \geq 1$ ). This is the Livšic hypothesis.

Sign / convention. The cohomological equation is written $φ = u \circ f - u$ (forward-difference convention), matching Katok-Hasselblatt ^{[Katok-Hasselblatt 1995 §19.2]}. With this sign, telescoping gives $S_{n} φ (x) = u (f^{n} x) - u (x)$ , so a coboundary has $S_{n} φ (p) = u (p) - u (p) = 0$ on every periodic orbit; the reconstruction $u (f^{n} x_{0}) = u (x_{0}) + S_{n} φ (x_{0})$ along an orbit follows the forward iterates, and convergence of the construction rests on the contraction in $E^{s}$ and the closing lemma of 38.03.04.

Counterexamples to common slips

The periodic condition is about the sum, not the values. $φ$ need not vanish at periodic points; only the total $S_{n} φ (p)$ around each closed orbit must vanish. A $φ$ nonzero everywhere can still be a coboundary.
Continuity of $φ$ alone is not enough. Livšic requires $φ$ Hölder. A merely continuous $φ$ with vanishing periodic sums can fail to be even a continuous coboundary; the Hölder modulus is exactly what makes the dense-orbit construction converge.
Transitivity is essential. On a non-transitive system the construction along a dense orbit is unavailable, and vanishing periodic sums on each transitive component only yields a transfer function on each component, with the components' constants unrelated.
Uniqueness is up to a constant, not absolute. If $u$ solves $φ = u \circ f - u$ then so does $u + c$ for any constant $c$ ; transitivity forces this to be the only ambiguity, since a continuous $f$ -invariant function on a transitive system is constant.

Key theorem with proof Intermediate+

Theorem (Livšic, 1971). Let $f$ be a topologically transitive Anosov diffeomorphism of a compact manifold $M$ and let $φ \in C^{α} (M)$ , $0 < α \leq 1$ , satisfy the periodic-orbit condition $S_{n} φ (p) = 0$ whenever $f^{n} p = p$ . Then there is a function $u \in C^{α} (M)$ , unique up to an additive constant, solving the cohomological equation $φ = u \circ f - u$ . Conversely, the periodic-orbit condition is necessary for $φ$ to be a continuous coboundary. (Livšic ^{[Livšic 1971]}; Katok-Hasselblatt ^{[Katok-Hasselblatt 1995 §19.2]}.)

Proof (reconstruction along a dense orbit, closed up by the closing lemma). Necessity is the telescoping identity: if $φ = u \circ f - u$ then $S_{n} φ (p) = u (f^{n} p) - u (p) = 0$ for every periodic $p$ .

Construction of $u$ on a dense orbit. By transitivity fix a point $x_{0}$ whose forward orbit ${f^{n} x_{0} : n \geq 0}$ is dense in $M$ . Define $u$ on this orbit by the only choice compatible with the equation and the normalisation $u (x_{0}) = 0$ : $$ u(f^n x_0) := S_n\varphi(x_0) = \sum_{k=0}^{n-1}\varphi(f^k x_0). $$ Then $u (f^{n + 1} x_{0}) - u (f^{n} x_{0}) = φ (f^{n} x_{0})$ , so the cohomological equation holds along the orbit. It remains to show $u$ is $α$ -Hölder on the dense orbit, for then it extends uniquely to a $C^{α}$ function on $M = \overline{{f^{n} x_{0}}}$ .

The Hölder estimate via closing. It suffices to bound $∣ u (f^{m} x_{0}) - u (f^{n} x_{0}) ∣$ by $C d (f^{m} x_{0}, f^{n} x_{0})^{α}$ for all $m, n$ . Set $y = f^{n} x_{0}$ and $N = m - n$ (take $N \geq 0$ ); then $u (f^{m} x_{0}) - u (f^{n} x_{0}) = S_{N} φ (y)$ , and $d (f^{N} y, y) = d (f^{m} x_{0}, f^{n} x_{0}) =: δ$ is small when the two points are close. By the Anosov closing lemma 38.03.04 there is a genuine periodic point $p$ of period $N$ with $$ d(f^k y,, f^k p) \le C,\theta^{\min(k,,N-k)},\delta, \qquad 0 \le k \le N, $$ where $0 < θ < 1$ is the hyperbolicity rate (the shadowing orbit hugs $y$ exponentially closely in the middle of the segment, the exponential closing estimate of 38.03.04). Because $S_{N} φ (p) = 0$ by hypothesis, $$ |S_N\varphi(y)| = |S_N\varphi(y) - S_N\varphi(p)| \le \sum_{k=0}^{N-1}\big|\varphi(f^k y) - \varphi(f^k p)\big| \le [\varphi]\alpha \sum{k=0}^{N-1} d(f^k y, f^k p)^\alpha. $$ Inserting the closing bound and summing the two-sided geometric series, $$ \sum_{k=0}^{N-1} d(f^k y, f^k p)^\alpha \le [\varphi]\alpha^{-1}\cdot[\varphi]\alpha, C^\alpha \delta^\alpha \sum_{k=0}^{N-1}\theta^{\alpha\min(k,N-k)} \le C',\delta^\alpha, $$ with $C^{'} = 2 C^{α} / (1 - θ^{α})$ independent of $N$ . Hence $∣ S_{N} φ (y) ∣ \leq [φ]_{α} C^{'} δ^{α} = [φ]_{α} C^{'} d (f^{m} x_{0}, f^{n} x_{0})^{α}$ . So $u$ is $α$ -Hölder on the dense orbit with seminorm $\leq [φ]_{α} C^{'}$ .

Extension and the equation. A Hölder function on a dense set extends uniquely to a Hölder function $u$ on the closure $M$ , with the same exponent and seminorm. The identity $φ = u \circ f - u$ holds on the dense orbit and both sides are continuous, hence it holds on all of $M$ .

Uniqueness. If $u_{1}, u_{2}$ both solve the equation then $w = u_{1} - u_{2}$ satisfies $w \circ f = w$ , an $f$ -invariant continuous function; on a transitive system such a function is constant, so $u_{1} - u_{2}$ is a constant. $□$

Bridge. The Livšic theorem builds toward the entire rigidity theory of hyperbolic and higher-rank systems, and the foundational reason it holds is that the periodic-orbit obstruction is a complete invariant: this is exactly the closing lemma of 38.03.04 read on the level of Birkhoff sums, where a near-return is replaced by a true periodic orbit whose vanishing sum controls the error. The central insight is that the same exponential shadowing that corrected round-off into a true orbit now corrects the cocycle $S_{N} φ$ into the periodic value $0$ , so the reconstruction along a dense orbit converges and the transfer function is Hölder. This is dual to the Hopf argument of 38.07.03: there one proved an invariant function constant by sliding along stable and unstable leaves, and the measurable Livšic theorem appears again in exactly that language, sliding the transfer function along the foliations. Putting these together, the periodic data of a hyperbolic system determine a Hölder function up to coboundary, and the bridge is that the closing lemma is the single mechanism converting periodic information into global Hölder regularity.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Prove the necessity direction in full: if $φ = u \circ f - u$ for some function $u$ (no regularity assumed), then $S_{n} φ (x) = u (f^{n} x) - u (x)$ for all $x$ and $n$ , and in particular $S_{n} φ (p) = 0$ on every periodic orbit.

Hint

Sum the relation $φ (f^{k} x) = u (f^{k + 1} x) - u (f^{k} x)$ over $k = 0, \dots, n - 1$ and telescope.

Answer

Apply the equation at $f^{k} x$ : $φ (f^{k} x) = u (f^{k + 1} x) - u (f^{k} x)$ . Sum over $k = 0$ to $n - 1$ : $$ S_n\varphi(x) = \sum_{k=0}^{n-1}\big(u(f^{k+1}x) - u(f^k x)\big) = u(f^n x) - u(x), $$ the sum telescoping. If $f^{n} p = p$ then $u (f^{n} p) = u (p)$ , so $S_{n} φ (p) = u (p) - u (p) = 0$ . This holds for any $u$ whatsoever — no continuity is needed for necessity, which is why the obstruction is a genuine invariant of the cohomology class. Rubric: full credit for the telescoping identity and the periodic specialisation.

Exercise 4 (medium, symbolic).

Carry out the Hölder estimate in the model linear case: let $f$ contract a stable coordinate by $θ$ and expand the complement by $θ^{- 1}$ , $0 < θ < 1$ , and suppose the closing lemma gives, for a segment with $d (f^{N} y, y) = δ$ , a periodic $p$ with $d (f^{k} y, f^{k} p) \leq C θ^{m i n (k, N - k)} δ$ . For $φ \in C^{α}$ with vanishing periodic sums, bound $∣ S_{N} φ (y) ∣$ and identify the constant.

Hint

Compare $S_{N} φ (y)$ with $S_{N} φ (p) = 0$ termwise, use the Hölder modulus, and sum the two-sided geometric series $\sum_{k} θ^{α m i n (k, N - k)}$ .

Answer

Since $S_{N} φ (p) = 0$ , $$ |S_N\varphi(y)| = \Big|\sum_{k=0}^{N-1}\big(\varphi(f^k y) - \varphi(f^k p)\big)\Big| \le [\varphi]\alpha\sum{k=0}^{N-1} d(f^k y, f^k p)^\alpha \le [\varphi]\alpha C^\alpha\delta^\alpha\sum{k=0}^{N-1}\theta^{\alpha\min(k,N-k)}. $$ The exponent $min (k, N - k)$ is symmetric about $k = N /2$ and the sum is at most twice $\sum_{j \geq 0} θ^{α j} = 1/ (1 - θ^{α})$ , so $$ |S_N\varphi(y)| \le [\varphi]\alpha,C^\alpha,\frac{2}{1-\theta^\alpha},\delta^\alpha = K,\delta^\alpha, \qquad K = \frac{2 C^\alpha [\varphi]\alpha}{1-\theta^\alpha}. $$ The bound is independent of $N$ , which is what lets $u$ be Hölder uniformly along the whole dense orbit. Rubric: full credit for the termwise comparison, the two-sided geometric sum, and $K = 2 C^{α} [φ]_{α} / (1 - θ^{α})$ .

Exercise 5 (medium, short-answer).

Two Hölder functions $φ, ψ$ over a transitive Anosov $f$ have equal periodic sums on every periodic orbit: $S_{n} φ (p) = S_{n} ψ (p)$ for all $f^{n} p = p$ . Show $φ$ and $ψ$ are Hölder-cohomologous.

Hint

Apply Livšic to the difference $φ - ψ$ .

Answer

The difference $χ = φ - ψ$ is $α$ -Hölder (with $α$ the smaller exponent) and has $S_{n} χ (p) = S_{n} φ (p) - S_{n} ψ (p) = 0$ on every periodic orbit. By the Livšic theorem there is a Hölder $u$ with $χ = u \circ f - u$ , i.e. $φ - ψ = u \circ f - u$ , which is exactly the statement that $φ$ and $ψ$ are Hölder-cohomologous. The transfer function $u$ is unique up to a constant. This is the precise sense in which periodic data are a complete invariant of Hölder cohomology: two Hölder functions are cohomologous if and only if their periodic sums agree orbit by orbit. Rubric: full credit for reducing to Livšic on the difference and the iff reading.

Exercise 6 (medium, short-answer).

Show that if $φ$ is a Hölder coboundary then $\int φ d μ = 0$ for every $f$ -invariant Borel probability measure $μ$ , and explain why the periodic-orbit condition is the sharper invariant.

Hint

Integrate $φ = u \circ f - u$ against an invariant $μ$ ; for the comparison, recall periodic orbits carry invariant measures but invariant measures are a larger class.

Answer

If $φ = u \circ f - u$ with $u$ continuous (hence bounded, $M$ compact) and $μ$ is $f$ -invariant, then $\int φ d μ = \int u \circ f d μ - \int u d μ = \int u d μ - \int u d μ = 0$ by invariance. Each periodic orbit ${p, \dots, f^{n - 1} p}$ carries the invariant measure $μ_{p} = \frac{1}{n} \sum_{k < n} δ_{f^{k} p}$ , and $\int φ d μ_{p} = \frac{1}{n} S_{n} φ (p)$ , so the integral-vanishing condition over all invariant measures contains the periodic-orbit condition. The two are in fact equivalent for transitive hyperbolic systems, because periodic-orbit measures are weak- $*$ dense in the invariant measures (a consequence of specification/closing), so vanishing on periodic measures forces vanishing on all invariant measures. The periodic-orbit condition is sharper to state — it is checkable orbit by orbit and needs no measure theory — yet equivalent in strength. Rubric: full credit for the invariance computation, the periodic-orbit measure, and the density remark.

Exercise 7 (hard, short-answer).

State and prove the measurable Livšic theorem: if $f$ is a transitive Anosov diffeomorphism preserving a smooth measure $m$ , $φ$ is Hölder, and $u$ is a measurable function with $φ = u \circ f - u$ $m$ -a.e., then $u$ coincides $m$ -a.e. with a Hölder function (in particular the periodic-orbit condition holds and Livšic applies).

Hint

Run the Hopf-type argument of 38.07.03: show $u$ is a.e. constant along stable and along unstable leaves, then use absolute continuity / local product structure to get a.e. continuity, hence Hölder.

Answer

Statement. A measurable a.e.-solution of the cohomological equation for a Hölder coboundary is a.e. equal to a Hölder solution (Livšic ^{[Livšic 1972]}). Proof. Along a stable leaf, for $y \in W^{s} (x)$ the difference $u (f^{n} y) - u (f^{n} x) = S_{n} φ (x) - S_{n} φ (y) + (u (y) - u (x))$ rearranges the equation; but $S_{n} φ (y) - S_{n} φ (x) \to$ a finite limit because $d (f^{k} y, f^{k} x) \to 0$ exponentially and $φ$ is Hölder (the series $\sum_{k} (φ (f^{k} y) - φ (f^{k} x))$ converges absolutely). Defining $\overset{u}{^} (x, y) := u (x) - u (y) - \sum_{k \geq 0} (φ (f^{k} x) - φ (f^{k} y))$ one checks it is $f$ -invariant in a suitable sense and constant along stable leaves a.e.; symmetrically $u$ is controlled along unstable leaves using backward sums. Thus the measurable $u$ differs from an explicit Hölder expression by something constant along both foliations a.e. By the absolute continuity of the stable/unstable foliations and the local product structure 38.07.03, a function constant a.e. along stable and along unstable plaques agrees a.e. with a continuous one (the same Fubini step as the Hopf argument). The resulting continuous representative satisfies the equation everywhere and has vanishing periodic sums, so by the (topological) Livšic theorem it is Hölder. Hence $u$ is a.e. Hölder. Rubric: full credit for the convergent stable/unstable sum, the constancy-along-leaves reduction, and the absolute-continuity/Fubini promotion to a Hölder representative.

Exercise 8 (hard, short-answer).

Explain the regularity (Livšic-de la Llave-Marco-Moriyón) theorem: for a $C^{\infty}$ transitive Anosov $f$ and a $C^{\infty}$ coboundary $φ$ with Hölder transfer function $u$ , why is $u$ in fact $C^{\infty}$ ? Identify the two foliation-regularity inputs and the gluing lemma.

Hint

The transfer function inherits smoothness along the stable foliation from forward sums and along the unstable* foliation from backward sums; Journé's lemma glues joint regularity.

Answer

By Livšic the Hölder $u$ solves $φ = u \circ f - u$ . Smoothness along stable leaves. On $W^{s} (x)$ one has the absolutely convergent representation $u (y) - u (x) = - \sum_{k \geq 0} (φ (f^{k} y) - φ (f^{k} x))$ , a series of $C^{\infty}$ functions converging in $C^{\infty}$ along the leaf because the derivatives of $φ \circ f^{k}$ along $W^{s}$ decay exponentially (stable contraction beats the $f^{k}$ growth for $φ$ smooth). Hence $u ∣_{W^{s}}$ is $C^{\infty}$ with uniform leafwise bounds. Smoothness along unstable leaves. Backward summation $u (y) - u (x) = \sum_{k \geq 1} (φ (f^{- k} y) - φ (f^{- k} x))$ on $W^{u} (x)$ gives $u ∣_{W^{u}}$ is $C^{\infty}$ similarly. Gluing. A function that is uniformly $C^{\infty}$ along two transverse Hölder foliations whose tangent bundles span $T M$ is globally $C^{\infty}$ — this is the Journé regularity lemma (a hyperbolic analogue of elliptic regularity / Hartogs-type separate-smoothness $\Rightarrow$ joint-smoothness). Therefore $u \in C^{\infty} (M)$ . The same scheme upgrades $C^{r} \Rightarrow C^{r - ε}$ and real-analytic $\Rightarrow$ real-analytic (de la Llave-Marco-Moriyón ^{[de la Llave-Marco-Moriyón 1986]}). The point is that the regularity of $u$ is never better transverse to the foliations than along them, but joint leafwise smoothness is global smoothness by Journé. Rubric: full credit for the two leafwise summation arguments and naming Journé's lemma as the gluing step.

Advanced results Master

Theorem (Livšic regularity; de la Llave-Marco-Moriyón). Let $f$ be a $C^{\infty}$ (resp. $C^{r}$ with $r > 1$ , resp. real-analytic) transitive Anosov diffeomorphism and $φ$ of the same regularity with vanishing periodic sums. Then the Hölder transfer function $u$ of the Livšic theorem is $C^{\infty}$ (resp. $C^{r - ε}$ for every $ε > 0$ , resp. real-analytic). (Livšic ^{[Livšic 1971]}; de la Llave-Marco-Moriyón ^{[de la Llave-Marco-Moriyón 1986]}.)

The mechanism is that $u$ inherits smoothness along the stable foliation from the absolutely convergent forward series $- \sum_{k \geq 0} (φ \circ f^{k} - φ \circ f^{k} (\cdot_{x}))$ , whose leafwise derivatives decay because stable contraction dominates the chain-rule growth of $φ \circ f^{k}$ , and smoothness along the unstable foliation from the corresponding backward series. The two leafwise regularities are then glued into joint global regularity by the Journé lemma, the hyperbolic analogue of the principle that separate smoothness along transverse foliations spanning $T M$ implies joint smoothness. The loss of $ε$ in the finitely-differentiable case is the standard cost of the Hölder transversality of the foliations 38.03.01.

Theorem (measurable Livšic; Livšic 1972). For a transitive Anosov $f$ preserving a smooth measure $m$ and Hölder $φ$ , any measurable $u$ with $φ = u \circ f - u$ $m$ -a.e. agrees $m$ -a.e. with a Hölder function; consequently $φ$ is a measurable coboundary if and only if it is a Hölder coboundary if and only if its periodic sums vanish. (Livšic ^{[Livšic 1972]}; Katok-Hasselblatt ^{[Katok-Hasselblatt 1995 §19.2]}.)

The proof runs the Hopf machinery of 38.07.03: the measurable solution is shown constant-along-leaves modulo the explicit convergent foliation sums, and absolute continuity plus the local product structure promote leafwise a.e.-constancy to a continuous representative, which the topological Livšic theorem then makes Hölder. The collapse of the measurable, continuous, and Hölder coboundary classes onto one another is a rigidity phenomenon special to hyperbolicity; for parabolic or elliptic systems the regularity classes genuinely differ.

Theorem (periodic data and smooth conjugacy). Let $f, g$ be topologically conjugate transitive Anosov diffeomorphisms, $h \circ f = g \circ h$ , such that the derivative periodic data match: for every $f$ -periodic $p$ of period $n$ , the return cocycles $D f_{p}^{n}$ and $D g_{h (p)}^{n}$ are conjugate (e.g. equal eigenvalues). Then in low dimension the conjugacy $h$ is smooth. The one-dimensional and codimension-one cases of this periodic-data rigidity reduce to a Livšic equation for $lo g$ of the derivative cocycle: the obstruction to upgrading the Hölder conjugacy to a smooth one is precisely a cohomological equation whose periodic obstruction is the mismatch of periodic eigenvalue data (de la Llave; Llave-Marco-Moriyón ^{[de la Llave-Marco-Moriyón 1986]}).

For Anosov flows the analogous statement is marked-length-spectrum rigidity: the lengths of closed geodesics are the periodic data of the geodesic flow, and matching them across two negatively curved metrics forces a time-preserving conjugacy, again through a Livšic equation for the ratio of the geodesic-flow generators (Otal, Croke in the surface case). The periodic-orbit obstruction is thus the universal carrier of rigidity: whatever invariant is read off periodic orbits is, by Livšic, a complete invariant of the relevant cohomology.

Theorem (Livšic for vector-valued and group-valued cocycles). The conclusion extends: an $R^{d}$ -valued (or compact-Lie-group-valued) Hölder cocycle over a transitive Anosov system whose periodic data vanish (the identity) on every periodic orbit is a Hölder coboundary. The abelian case is the same proof componentwise; the non-abelian case (Livšic for $G$ -valued cocycles, Parry-Pollicott, Schmidt) needs the periodic data to be conjugate within $G$ and the closing estimate to control a noncommutative product, but the architecture — periodic obstruction is complete, reconstruction along a dense orbit, closing-lemma convergence — is identical.

Synthesis. These results are one principle — the periodic-orbit obstruction is the complete invariant of hyperbolic cohomology — applied with increasing structure, and the foundational reason they cohere is that the Anosov closing lemma of 38.03.04 converts a near-return into an exact periodic orbit whose vanishing (or matching) data force a Hölder, then smooth, transfer function. This is exactly the shadowing mechanism of 38.03.04 read on Birkhoff sums rather than on orbits, and it is dual to the Hopf argument of 38.07.03: the measurable Livšic theorem is the Hopf argument applied to the transfer function, sliding it along the stable and unstable foliations until absolute continuity pins it down. The central insight is that periodic data are a complete invariant, so cohomology, smooth conjugacy, and marked-length-spectrum rigidity are three readings of the same closed-loop bookkeeping; putting these together, the regularity theorem shows the transfer function is as smooth as the data because leafwise smoothness glues to global smoothness by Journé, and the bridge is that the closing lemma, the Hopf foliation argument, and the Journé gluing are the three faces of uniform hyperbolicity that make the periodic obstruction govern everything from Hölder cohomology to the rigidity of negatively curved metrics.

Full proof set Master

Proposition (necessity and the telescoping identity). For any $φ$ and any $u$ with $φ = u \circ f - u$ , one has $S_{n} φ (x) = u (f^{n} x) - u (x)$ ; hence $S_{n} φ (p) = 0$ on every periodic orbit.

Proof. Summing $φ (f^{k} x) = u (f^{k + 1} x) - u (f^{k} x)$ over $0 \leq k < n$ telescopes to $S_{n} φ (x) = u (f^{n} x) - u (x)$ . For $f^{n} p = p$ the right side is $u (p) - u (p) = 0$ . $□$

Proposition (the dense-orbit transfer function is Hölder). Let $f$ be a transitive Anosov diffeomorphism, $φ \in C^{α}$ with vanishing periodic sums, $x_{0}$ a point with dense forward orbit, and $u (f^{n} x_{0}) := S_{n} φ (x_{0})$ . Then $u$ is $α$ -Hölder on ${f^{n} x_{0}}$ with $[u]_{α} \leq K [φ]_{α}$ for a constant $K$ depending only on the closing constants.

Proof. For $m > n$ put $y = f^{n} x_{0}$ , $N = m - n$ , $δ = d (f^{m} x_{0}, f^{n} x_{0}) = d (f^{N} y, y)$ , so $u (f^{m} x_{0}) - u (f^{n} x_{0}) = S_{N} φ (y)$ . The Anosov closing lemma 38.03.04 supplies a periodic $p$ , $f^{N} p = p$ , with $d (f^{k} y, f^{k} p) \leq C θ^{m i n (k, N - k)} δ$ . Since $S_{N} φ (p) = 0$ , $$ |S_N\varphi(y)| \le [\varphi]\alpha\sum{k=0}^{N-1} d(f^k y, f^k p)^\alpha \le [\varphi]\alpha C^\alpha\delta^\alpha\sum{k=0}^{N-1}\theta^{\alpha\min(k,N-k)} \le [\varphi]_\alpha,\frac{2C^\alpha}{1-\theta^\alpha},\delta^\alpha. $$ The constant $K = 2 C^{α} / (1 - θ^{α})$ is independent of $m, n$ , giving the uniform Hölder bound. $□$

Proposition (unique Hölder extension and the global equation). The function $u$ of the previous proposition extends uniquely to $u \in C^{α} (M)$ , and the extension satisfies $φ = u \circ f - u$ everywhere.

Proof. A uniformly $α$ -Hölder function on a dense subset $D$ of a compact metric space extends uniquely to an $α$ -Hölder function on $\overline{D} = M$ (Cauchy continuity: $α$ -Hölder maps Cauchy sequences to Cauchy sequences, and the extension preserves the seminorm by passing to limits). On $D$ the relation $u (f x) - u (x) = φ (x)$ holds by construction; $f (D) \subseteq D$ and all three functions $u \circ f$ , $u$ , $φ$ are continuous, so the identity persists to $\overline{D} = M$ . $□$

Proposition (uniqueness up to a constant). On a transitive system two solutions of the cohomological equation differ by a constant.

Proof. If $φ = u_{i} \circ f - u_{i}$ , $i = 1, 2$ , then $w = u_{1} - u_{2}$ has $w \circ f = w$ . A continuous $f$ -invariant function on a topologically transitive system is constant: it is constant on the dense orbit ${f^{n} x_{0}}$ (invariance makes $w (f^{n} x_{0}) = w (x_{0})$ ) and hence, by continuity, on $M$ . So $u_{1} - u_{2} \equiv$ const. $□$

Proposition (periodic obstruction is the complete invariant of Hölder cohomology). Two Hölder functions over a transitive Anosov $f$ are Hölder-cohomologous if and only if they have equal periodic sums on every periodic orbit.

Proof. ( $\Rightarrow$ ) If $φ - ψ = u \circ f - u$ then $S_{n} (φ - ψ) (p) = u (f^{n} p) - u (p) = 0$ on periodic orbits, so the periodic sums agree. ( $\Leftarrow$ ) If $S_{n} φ (p) = S_{n} ψ (p)$ for all periodic $p$ , then $χ = φ - ψ$ is Hölder with vanishing periodic sums, and the Livšic theorem yields a Hölder $u$ with $χ = u \circ f - u$ , i.e. $φ$ and $ψ$ are Hölder-cohomologous. $□$

Connections Master

Shadowing, the closing lemma, and structural stability 38.03.04. This unit is the cohomological payoff of the closing lemma: where 38.03.04 turned a near-return pseudo-orbit into a genuine periodic orbit, the Livšic proof uses that periodic orbit's vanishing sum to control the Birkhoff cocycle $S_{N} φ$ and force the transfer function to be Hölder. The exponential closing estimate of 38.03.04 is the exact input that makes the dense-orbit reconstruction converge, so Livšic rigidity rests on shadowing precisely as ergodicity rested on it for density of periodic measures.
The Hopf argument and ergodicity of Anosov flows 38.07.03. The measurable Livšic theorem is the Hopf argument applied to a transfer function rather than a Birkhoff average: $u$ is shown constant-modulo-explicit-sums along the stable and unstable foliations, and absolute continuity together with the local product structure promotes that leafwise control to a continuous representative. Both theorems convert two leafwise constancies into one global conclusion through the same Fubini step licensed by absolute continuity.
Hyperbolic sets, Anosov and Axiom-A systems 38.03.01. The stable/unstable splitting and the Hölder transversality of the foliations proved there are what make both the reconstruction and the de la Llave-Marco-Moriyón regularity work; the $ε$ -loss in the $C^{r}$ regularity statement is exactly the price of the Hölder (not $C^{1}$ ) transverse regularity established in 38.03.01. Transitivity, supplied by the spectral decomposition of 38.03.01, is what makes the dense orbit and the up-to-a-constant uniqueness available.
Topological pressure and equilibrium states 38.06.04. Livšic cohomology is the equivalence relation underlying the thermodynamic formalism: two Hölder potentials cohomologous (equal periodic sums) define the same equilibrium state and the same pressure up to the additive constant, so the periodic-orbit obstruction classifies potentials exactly as the variational principle classifies their measures. The transfer operator's spectral data depend on a potential only through its Livšic class.

Historical & philosophical context Master

The cohomological equation over hyperbolic systems was solved by Alexander N. Livšic in two papers of the early 1970s. The 1971 Izvestiya article ^{[Livšic 1971]} introduced the periodic-orbit obstruction and proved that, for a U-system (Anosov diffeomorphism or flow) and a Hölder function whose sums vanish on every periodic orbit, a Hölder transfer function exists and is unique up to an additive constant; the proof reconstructed the transfer function along a dense orbit and controlled the error by the closing lemma. The companion 1972 Matematicheskie Zametki note ^{[Livšic 1972]} proved the measurable version — that a measurable solution is automatically a.e. Hölder — completing the rigidity picture by collapsing the regularity classes.

The smooth-regularity question — whether a $C^{\infty}$ coboundary has a $C^{\infty}$ transfer function — was settled by Rafael de la Llave, José Manuel Marco, and Roberto Moriyón in their 1986 Annals of Mathematics paper ^{[de la Llave-Marco-Moriyón 1986]}, which established smoothness of $u$ along the stable and unstable foliations and glued the two through a separate-smoothness regularity lemma (later abstracted by Journé). Their work connected Livšic theory to the rigidity programme: matching periodic data — eigenvalues at periodic points, or marked length spectra of negatively curved metrics — forces smooth conjugacy, the periodic-orbit obstruction serving as the universal carrier. Katok and Hasselblatt's 1995 treatise ^{[Katok-Hasselblatt 1995]} gives the canonical textbook account, and the higher-rank cocycle-rigidity extensions appear in the Katok-Nițică monograph on abelian group actions.

Bibliography Master

@article{Livsic1971,
  author  = {Liv{\v{s}}ic, Alexander N.},
  title   = {Cohomology of dynamical systems},
  journal = {Mathematics of the USSR-Izvestiya},
  volume  = {6},
  number  = {6},
  year    = {1972},
  pages   = {1278--1301},
  note    = {Russian original: Izv. Akad. Nauk SSSR Ser. Mat. 35 (1971) 1296--1320}
}

@article{Livsic1972,
  author  = {Liv{\v{s}}ic, Alexander N.},
  title   = {Certain properties of the homology of {Y}-systems},
  journal = {Mathematical Notes},
  volume  = {10},
  number  = {5},
  year    = {1971},
  pages   = {758--763}
}

@article{LlaveMarcoMoriyon1986,
  author  = {de la Llave, Rafael and Marco, Jos{\'e} Manuel and Moriy{\'o}n, Roberto},
  title   = {Canonical perturbation theory of {A}nosov systems and regularity results for the {L}iv{\v{s}}ic cohomology equation},
  journal = {Annals of Mathematics},
  volume  = {123},
  number  = {3},
  year    = {1986},
  pages   = {537--611}
}

@book{KatokHasselblatt1995,
  author    = {Katok, Anatole and Hasselblatt, Boris},
  title     = {Introduction to the Modern Theory of Dynamical Systems},
  publisher = {Cambridge University Press},
  series    = {Encyclopedia of Mathematics and its Applications},
  volume    = {54},
  year      = {1995}
}

@book{KatokNitica2011,
  author    = {Katok, Anatole and Ni{\c{t}}ic{\u{a}}, Viorel},
  title     = {Rigidity in Higher Rank Abelian Group Actions. Volume I: Introduction and Cocycle Problem},
  publisher = {Cambridge University Press},
  series    = {Cambridge Tracts in Mathematics},
  volume    = {185},
  year      = {2011}
}

@article{Journe1988,
  author  = {Journ{\'e}, Jean-Lin},
  title   = {A regularity lemma for functions of several variables},
  journal = {Revista Matem{\'a}tica Iberoamericana},
  volume  = {4},
  number  = {2},
  year    = {1988},
  pages   = {187--193}
}

@article{Otal1990,
  author  = {Otal, Jean-Pierre},
  title   = {Le spectre marqu{\'e} des longueurs des surfaces {\`a} courbure n{\'e}gative},
  journal = {Annals of Mathematics},
  volume  = {131},
  number  = {1},
  year    = {1990},
  pages   = {151--162}
}

Prerequisites

38.03.01
38.03.04
38.07.03

Tier anchors

beginner: Katok-Hasselblatt 1995 *Introduction to the Modern Theory of Dynamical Systems* (Cambridge) Ch. 19 (informal: when a running tally over a chaotic rule can be cancelled by a change of bookkeeping); 3Blue1Brown cat-map and Anosov visual essays for the closed-loop / periodic-orbit imagery
intermediate: Katok-Hasselblatt 1995 *Introduction to the Modern Theory of Dynamical Systems* (Cambridge University Press) Ch. 19 §19.2 (the Livšic theorem and the cohomological equation over hyperbolic systems); Coudène 2016 *Ergodic Theory and Dynamical Systems* (Springer Universitext) Ch. 11 (cohomological equations, the Livšic theorem)
master: Katok-Hasselblatt 1995 *Introduction to the Modern Theory of Dynamical Systems* (Cambridge) Ch. 19 (the Livšic theorem, periodic-orbit obstruction, regularity); Livšic 1971 *Cohomology of dynamical systems* (Math. USSR Izvestiya 6); Livšic 1972 *Homology properties of Y-systems* (Math. Notes 10); de la Llave-Marco-Moriyón 1986 *Canonical perturbation theory of Anosov systems and regularity results for the Livšic cohomology equation* (Annals of Mathematics 123)

References

Livšic 1971 — Cohomology of dynamical systems · Izvestiya Akademii Nauk SSSR, Seriya Matematicheskaya 35 (1971) 1296-1320; Mathematics of the USSR-Izvestiya 6 (1972) 1278-1301; the cohomological equation over a U-system (Anosov), the periodic-orbit obstruction as the complete invariant, the Hölder transfer function and its uniqueness up to a constant
Livšic 1972 — Certain properties of the homology of Y-systems · Matematicheskie Zametki 10 (1971) 555-564; Mathematical Notes 10 (1971) 758-763; the measurable Livšic theorem — a measurable solution of the cohomological equation for a Hölder coboundary is a.e. equal to a Hölder one
de la Llave, Marco, Moriyón 1986 — Canonical perturbation theory of Anosov systems and regularity results for the Livšic cohomology equation · Annals of Mathematics 123 (1986) 537-611; the regularity theorem — for a $C^\infty$ (or $C^r$, or real-analytic) Anosov system and coboundary the transfer function is $C^\infty$ (resp. $C^{r-\varepsilon}$, analytic), via smoothness along the stable and unstable foliations and an elliptic-regularity / Journé-lemma argument
Katok, Hasselblatt 1995 — Introduction to the Modern Theory of Dynamical Systems · Cambridge University Press 1995, Ch. 19 §19.2 (the Livšic theorem via the closing lemma, periodic data, regularity); Encyclopedia of Mathematics and its Applications 54
Katok, Nițică 2011 — Rigidity in Higher Rank Abelian Group Actions · Cambridge University Press 2011, Vol. I Ch. 2 (the Livšic theory, cocycle rigidity, the periodic-cycle functional); the higher-rank cohomological-rigidity context

Estimated time

beginner: 19m
intermediate: 52m
master: 92m