Let $G$ be a semisimple real lie group, $K$ be a maximal compact subgroup of $G$, $\Gamma$ be a torsion-free cocompact discrete subgroup. The Betti number the locally symmetric space $X_{\Gamma}:=\Gamma \backslash G/K$ can be expressed in term of unitary representations of $G$ by Matsushima's formula.

The formula looks quite complicated in general, the question is: when does $X_{\Gamma}$ have all vanishing odd degree Betti numbers? I am interested in some examples and some necessary conditions.

Example: Consider the toroidal compactification of Siegel threefold with $3$ level structure, its Betti numbers are $1, 0, 61, 0, 61, 0, 1$, see "The Siegel modular variety of degree two and level three" by J. William Hoffman and Steven H. Weintraub (Trans AMS 2001).


1 Answer 1


Here is an answer of sorts.

For simplicity, I restrict to cocompact torsion-free lattices. Then $b_{odd}(X_\Gamma)=0$ implies that $X$ is even-dimensional. The situation is clear if $dim(X)\le 2$, so the first interesting case is when $dim(X)=4$. There are (essentially) three 4-dimensional symmetric spaces of noncompact type (up to rescaling the metric on de Rham factors): The real-hyperbolic space, ${\mathbb H}^4$ (when $G=PO(4,1)$), the complex-hyperbolic space ${\mathbb H}^2_{\mathbb C}$ (when $G=PU(2,1)$) and the product of two hyperbolic planes ${\mathbb H}^2\times {\mathbb H}^2$ (when $G=PO(2,1)\times PO(2,1)$), up to commensuration of Lie groups.

  1. $X={\mathbb H}^4$. I do not know if there are examples with $b_1=0$. However, it is known that for every arithmetic uniform lattice $\Gamma< PO(4,1)$ there exists a congruence-subgroup $\Gamma_1<\Gamma$ such that $b_1(\Gamma_1)= b_1(X_{\Gamma_1})>0$ (Millson). I do not think anybody knows what happens with non-arithmetic manifolds.

  2. $X={\mathbb H}^2_{\mathbb C}$. Arithmetic lattices come in two types: All lattices of type I again have congruence-subgroups $\Gamma_1$ with $b_1(\Gamma_1)>0$ (Kazhdan). However, for all congruence-subgroups $\Gamma_1$ in type II arithmetic lattice, $b_1(\Gamma_1)=b_3(\Gamma_1)=0$. So all odd Betti numbers vanish. The question of positivity of 1st Betti number of finite-index non-congruence subgroups is wide-open. The non-arithmetic case is mostly open. (There is one nonvanishing result, due to Yeung.)

  3. $X={\mathbb H}^2\times {\mathbb H}^2$. The case of reducible lattices is (essentially) clear, so I will consider irreducible lattices. For all such lattices $\Gamma$, $b_1(\Gamma)=b_3(\Gamma)=0$. (This follows from Margulis' super-rigidity, but might have been known earlier.) So all odd Betti numbers vanish.

I will not attempt to analyze the case when $dim(X)\ge 6$. One gets vanishing results for $b_1$, but the situation with higher odd Betti numbers are very much unclear, unless you know something about odd Betti numbers of the dual compact metric space: Nonvanishing of some of those implies nonvanishing of $b_{odd}(\Gamma)$. There are vanishing theorems going back to Matsushima, but they do not deal with all odd Betti numbers.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.