Recently I am reading Wall's paper "On the Orthogonal Groups of Unimodular Quadratic Forms II". In this paper, I encountered with the map $E^1_\omega$, which now I am interested in.

Let $X$ be an unimodular integral lattice, and $H$ denotes the unimodular lattice which has a basis $x$, $y$ satisfying $x\cdot x=y\cdot y=0$, and $x\cdot y=1$. For $\omega \in X$ with $\omega\cdot\omega\in 2\mathbb{Z}$, we define the isometry $E^1_\omega$ of $X\oplus H$ as follows.

  • For $\xi\in X$, $E^1_\omega(\xi)=\xi-(\xi\cdot\omega)x$.
  • $E^1_\omega(x)=x$.
  • $E^1_\omega(y)=y+\omega-2^{-1}(\omega\cdot\omega)x$.

My question is: What is algebraic intuition behind the definition of $E^1_\omega$?

EDIT 1 : As D. Ruberman answered below, there exists intuition behind the map $E^1_\omega$ from the view point of differential topology. In fact, Wall proved that, for any closed simply-connected 4-manifold $M$ which has indefinite unimodular form $X$ as intersection form, any automorphism of $X\oplus H$ is realized by some diffeomorphisms of $M\#S^2\times S^2$.

He proved it as follows.

Step 1 : He had showed that the automorphism group of $X\oplus H$ is generated by $E^1_\omega$ and $E^2_\omega$, where $E^2_\omega$ is defined by interchanging the roles of $x$ and $y$ in $E^1_\omega$.

Step 2 : He also proved that each of $E^i_\omega$ is induced from a diffeomorphism of $M\#S^2\times S^2$. This completes the proof.

I already knew the proof of Step 2 which intuitively explains the map $E^i_\omega$ from the topological point of view. But I want to know some intuition behind Step 1, which is algebraic part of the proof.

EDIT 2 : The map $E^i_\omega$ is called the Siegel-Eichler transformation in a lot of literature.

  • $\begingroup$ Writing $X\oplus H= \langle x \rangle \oplus H \oplus \langle y \rangle$, the given automorphisms are the upper/lower triangular matrices with 1 on the diagonal. $\endgroup$
    – user83633
    Jul 22, 2016 at 12:22
  • $\begingroup$ @user83633 Thank you for your comment. (I think you made a subtle typo $X\oplus H = \langle x \rangle \oplus X \oplus \langle y \rangle$.) $\endgroup$ Jul 24, 2016 at 8:31

2 Answers 2


This is explained very clearly in Wall's paper, Diffeomorphisms of 4-manifolds, J. London Math. Soc. (1964) 131-140. The idea is that if you do surgery on a circle $C$ in a simply-connected $4$-manifold $X$, you get $X \# T_r$ where $T$ is a 2-sphere bundle over $S^2$ with fiber $S^2$, and intersection form generated by $x,y$ as above, except that maybe $y\cdot y = r$ may be non-zero. (Up to diffeomorphism, only the parity of $r$ matters.) Now if you perform an isotopy of that $C$, ending up at $C$, you get a self-diffeomorphism of $X \# T_r$. (You can get $r\neq 0$ by twisting the framing of the circle in this isotopy.) If the $2$-dimensional homology class swept out by the circle (ie a map of a torus) in this isotopy is $\omega$, then the induced map on homology is of the form $E^1_\omega$.

I don't know what was the original intuition behind the algebraic formula (due to Siegel), but Wall's geometric picture makes it a lot clearer to me.

  • $\begingroup$ @Ruberman Thank you for your answer. Yes, I have already known geometric intuition (via Kirby calculus) of the map $E^1_\omega$. But I know next to nothing about algebraic point of view, too. $\endgroup$ Jul 22, 2016 at 5:57
  • $\begingroup$ Maybe you want to revise your question to ask for the algebraic intuition behind Siegel's definition. $\endgroup$ Jul 22, 2016 at 13:36
  • $\begingroup$ @Ruberman You are right. I have edited the question to make it clear. I apologize any confusion because of my bad English language. $\endgroup$ Jul 23, 2016 at 4:32

Recently, I had found some algebraic intuition by generalizing the construction.

We can define similar isomorphism $E^1_a$ of $X\oplus H$ for any $a\in X^{*}(=\text{Hom}_{R}(X,R)$, where $R$ is the coefficient ring of lattices).

To begin with, we define the isomorphism $E^1_a$ of $X\oplus\langle x\rangle$ as follows:

  • For $\xi\in X$, $E^1_a(\xi):=\xi+a(\xi)x$,
  • $E^1_a(x)=x$.

It seems that the construction of $E^1_a$ is natural and easy to come up with. And more, the following lemma holds if we assume the coefficient ring of lattices $R$ has 2 as an unit:

The isomorphism $E^1_a$ can be uniquely extended as isomorphism of $X\oplus H$.

Proof : Let's assume that there exists such an extension and write $E^1_a(y)=\omega+sx+ty$ by using some $\omega\in X$. Because $E^1_a$ preserves the metric, we have

  • For $\xi\in X$, $0=E^1_a(\xi)\cdot E^1_a(y)=\xi\cdot\omega+ta(\xi)$,
  • $1=E^1_a(x)\cdot E^1_a(y)=t$,
  • $0=E^1_a(y)\cdot E^1_a(y)=\omega\cdot\omega +2st$.

Hence, we have $t=1$, for any $\xi\in X$, $\underline{\xi\cdot\omega+a(\xi)=0}_{(*)}$, and $s=-2^{-1}(\omega\cdot\omega)$. Note that there always exists the unique element $\omega\in X$ which satisfies $(*)$ (because $X$ has non-degenerate metric). This suggests the method of construction and completes the proof.

Finally, in the case of $R=\mathbb{Z}$ and $\omega\in X$ with $\omega\cdot\omega\in 2\mathbb{Z}$, if we take $a\in X^*$ as $(*)$, $E^1_a$ agrees with the Siegel-Eichler transformation $E^1_\omega$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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