1.1. Preliminaries on Euclidean spaces
If and are Euclidean spaces (finite dimensional inner product spaces
over then so is with the inner product
(1.1) |
|
|
|
If is an orthonormal basis of and an orthonormal basis of then is an orthonormal basis of In this way is endowed with a Euclidean structure.
The -th exterior power is the quotient of
by the subspace spanned by tensors in which, for some , The
orthogonal complement is the space of alternating -tensors, spanned by
(1.2) |
|
|
|
and this tensor projects modulo to the image modulo of We may therefore identify with . As such it inherits from a Euclidean
structure. If is an orthonormal basis of then for is a basis of which is orthogonal
but not normalized: the norm of is To correct it, we
modify the inner product that inherits from
by a factor of and set
(1.3) |
|
|
|
In this inner product the form an orthonormal basis. Moreover, the
volume of the -dimensional parallelopiped spanned by in is nothing but the norm To see it, choose an orthonormal basis of
the first vectors of which span
(assuming the are linearly independent) and observe that
(1.4) |
|
|
|
where
1.2. Length and norm in and the rank 1 case of the
theorem
To begin the proof, note that
(1.5) |
|
|
|
If is a vector in we denote by the length of which is the minimal number such that
(1.6) |
|
|
|
(To avoid abuse of language, will always be called the norm of , and not its length.) In such a case, the and the are linearly independent vectors in and
respectively. Thus the length of a vector is at most
Let us write, for any lattice
(1.7) |
|
|
|
The maximum is over all sublattices of all ranks (it is easily seen that the
maximum is attained). A unimodular lattice is semistable if and only if
Proposition 1.1.
Let and be any two lattices, normalized so that and Then
(1.8) |
|
|
|
Proof.
Writing , a direct
computation gives
(1.9) |
|
|
|
Consider the Gramians and These are symmetric
positive definite by matrices with and because we normalized and so that the volume of
the parallelopiped formed by the or the is at
least 1.
We are indebted to Assaf Goldberger for pointing out the following fact.
Lemma 1.2.
Let and be two symmetric positive definite by
matrices. Then the spectrum of is real and positive.
The matrix although not necessarily symmetric, is conjugate
to
(1.10) |
|
|
|
which is symmetric positive definite since is.
We now continue the proof of the proposition. By our assumption, the product
of the eigenvalues of is at least 1, and the lemma implies
that they are real and positive. By the inequality between the geometric and
arithmetic means, However,
(1.11) |
|
|
|
so as needed.
∎
The proposition proves the case of the theorem, but to
prove the cases or we shall need its full strength.
1.3. The rank 2 case of the theorem
Let now be a sublattice of rank in Let be a nonzero vector of shortest norm in and another
vector in such that . Subtracting a
multiple of from we may assume that the angle between and is between and It
follows that
(1.12) |
|
|
|
If either of or is indecomposable (of length ) we
are done, since its norm is then at least and the other’s norm is
at least 1, but We may therefore assume that
(1.13) |
|
|
|
are both decomposable tensors. We now use the identity
(1.14) |
|
|
|
|
|
|
|
|
|
|
which follows at once from the fact that
(1.15) |
|
|
|
If and are proportional, then and are
independent, and
while the second and the third terms vanish. Otherwise, the first term is
larger than the third, and the second is by our assumption. This
concludes the rank 2 case.
1.4. The rank 3 case of the theorem
The rank 3 case is handled similarly, but the details are more
complicated. Let Let be a shortest nonzero vector in
and a second shortest vector independent from and assume
as before that the angle between and is between and Complete to a basis of by a vector Using orthonormal
coordinates in the real subspace spanned by such that the -axis is in the -direction and the -plane is the plane
spanned by and we may assume, subtracting from
a suitable integral linear combination of and that
(1.16) |
|
|
|
with and This is because the
rectangle
(1.17) |
|
|
|
where is the angle between and is a
fundamental domain for We now have
(1.18) |
|
|
|
and
(1.19) |
|
|
|
(by the rank 2 case), so
(1.20) |
|
|
|
If , or then by the proposition
and the fact that we must have and we are done.
There remains the case and in which the proposition does not
help us. The key in this case is the following identity.
Lemma 1.3.
We have an 18-term relation
(1.21) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
where we have used the convention that
are the two indices complementary to .
Proof.
Expanding the grammian of three vectors in Euclidean space we get
(1.22) |
|
|
|
|
|
|
|
|
|
|
For our decomposable tensors we get the expression
(1.23) |
|
|
|
|
|
|
|
|
|
|
We proceed to replace all the inner products by expressions involving only
wedge-products (sines instead of cosines). The elements of the form can be replaced by as in the
case of rank 2. For use
(1.24) |
|
|
|
|
|
|
|
|
|
|
This is an 5-term expression and when we multiply the one for the with the one for the we get 25 terms,
and a total of 38 terms for A
direct computation (carried out by a computer) reveals that there are many
cancellations, and the 18-term relation falls out.
∎
We continue the proof of the theorem, when , and
are decomposable as above. Choose an orthonormal basis of so that and similarly an orthonormal basis of putting the
in a triangular form. Assume first that the are linearly
independent. Expanding in the orthonormal basis of and in the corresponding orthonormal
basis of of (recall the
normalization from 1.1)
(1.25) |
|
|
|
|
|
|
|
|
|
|
Assume next that both the and the are linearly dependent. If
and are proportional then and can not be
proportional, otherwise In this case we may assume that and while and We find
(1.26) |
|
|
|
|
|
|
|
|
|
|
There remains the case where the and the are linearly
dependent, but no two vectors in each triplet are proportional.
Introduce the notation
(1.27) |
|
|
|
and similarly Note that
|
|
|
where is the angle between and in the plane spanned by the We may assume
that A
direct computation shows then that
(1.28) |
|
|
|
Similarly denote by the angle between and We now observe that in the 18-term relation, nine
terms drop out and the remaining nine give
(1.32) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
(1.36) |
|
|
|
|
|
This concludes the proof of the theorem.
1.5. An argument from duality
Duality allows us to conclude more. Quite generally, if
(1.37) |
|
|
|
is an exact sequence of Euclidean lattices (this means that is
torsion-free and that the metrics on and are the
ones induced from the metric on ) then
(1.38) |
|
|
|
If is the dual lattice, From these
two observations it follows at once that if is semistable, so is
Let and be two unimodular semistable lattices of ranks and .
Then and are unimodular and semistable too. Suppose
we know the desired results for all of some rank Let be a rank (we shall say it has corank ) sublattice of for
which we want to prove that We may assume that is torsion free. We equip it with the quotient metric and call it .
Then is a rank lattice and is a rank
sublattice of By what we have seen already, but
(1.39) |
|
|
|
since and This shows that
if we know the desired result for all rank lattices, we also know it for
all corank lattices.
Corollary 1.4.
If and are semistable and they are both of rank 2, or one of them is
of rank 2 and the other is of rank 3, then is
semistable.
[Fa] Faltings, G.: Mumford-Stabilität in der algebraischen
Geometrie, Proceedings of the 1994 ICM.
[To] Totaro, B.: Tensor products in -adic Hodge theory, Duke Math.J. 83 (1996), 79-104.