Consider the quadratic Gauss sum
where is an odd prime. Writing the expression using the quadratic character modulo
,
and squaring gives
Thus the magnitude and possible direction of are already known:
The squaring argument leaves exactly one ambiguity: which sign occurs?
There are several ways to settle this ambiguity. One can use Poisson summation, theta functions, contour integration, or Estermann’s estimate. Schur’s proof is different in character. It treats the Gauss sum as the trace of the finite Fourier transform. The square of that transform reveals its possible eigenvalues, while its determinant fixes the remaining sign.
Let and let
be the vector space of complex-valued functions on the finite field
. Define the unnormalized Fourier transform
In the standard basis, is the
matrix
Its diagonal entries are , so its trace is exactly the Gauss sum:
Thus the problem of evaluating becomes a spectral problem: determine the trace of the Fourier matrix. It is convenient to normalize the transform:
The orthogonality of additive characters gives
Hence so
is unitary. In particular, it is diagonalizable and all of its eigenvalues have absolute value
. More importantly, Fourier inversion gives a simple formula for the square of
:
Interchanging the sums gives
The inner sum vanishes unless , in which case it equals
. Therefore
Let
denote the reflection operator
Then
This is the main structural fact. The space splits into its even and odd parts:
where
The dimensions are
On the even subspace, acts as
, so
. Thus the eigenvalues of
on
are
and
. On the odd subspace,
acts as
, so
. Thus the eigenvalues of
on
are
and
. Equivalently, the eigenvalues of
are among
Let be the multiplicities of the eigenvalues
of
. The even-odd decomposition gives
The trace is
At this stage, the spectrum has narrowed the Gauss sum to the expected four directions. To determine which direction occurs, we need one further piece of information: the determinant.
The determinant of is a Vandermonde determinant:
For ,
The sine factors are all positive, because
The exponential factors contribute
Every integer appears in exactly
pairs, so
Therefore the exponential factor is because
is odd. Thus
The remaining product is positive. Also,
Hence and after normalization,
On the other hand, the determinant is the product of the eigenvalues:
Since , this becomes
Comparing this with the Vandermonde evaluation yields
This parity relation is exactly the missing information.
First suppose that Then
Since we already know that
the imaginary part of
must vanish. Hence
The real part is either or
, so
The determinant condition gives
Thus
, which selects
Therefore
and so
Now suppose that Then
This time
so the real part of the trace vanishes. Hence
The imaginary part is
or
, so
The determinant condition now gives Thus
which forces
Therefore
and hence
We have proved the classical evaluation
The conceptual structure of the proof is worth emphasizing. The identity confines the spectrum of the Fourier transform to the four fourth roots of unity. This almost determines the trace, but not its sign. The Vandermonde determinant supplies the missing orientation information: it records the total phase of the Fourier matrix. Once that phase is known, the multiplicities, and therefore the trace, are forced.