Let
be a finite directed graph with multiple edges allowed, and let
denote the vertex set of
and
the edges set of
. Let
and
be the functions from
to
defined by
where es is an edge from vertex i to vertex j. For a vertex
we put

and

We say that
is an Eulerian path of
if
is an element of Sym(d) (the symmetric group acting on the set
and
for
.
It is well known that a connected graph
has an Eulerin path starting at vertex p and ending at vertex q if and only if one of the following two conditions applies:
1)
and
for each
;
2)
and
,
and
for each
.
A directed connected graph
with fixed vertices p and q is called Eulerian if either condition 1) or 2) is satisfied. We note that if
is an Eulerian graph of type (b), then the vertices p, q are uniquely determined, but in the other case we may choose any vertex
. For an Eulerian graph
denote by

2. Main Results
Let
be an Eulerian graph with d edges
and distinguished points p and q. the polynomial
associated with
is defined as follows:

Thus
is a multilinear polynomial in the set
of non-commuting indeterminates.
Let
be an integer, C a commutative ring with 1 and
a set map where the
’s are the standard matrix units over C. It is clear that T can be viewed as a substitution. we shall define a directed graph
induced from
by T. First consider the directed graph on the vertex set
with edge set
where
,
and
. Now we define
by restricting the vertex set to
. We note that the graph so obtained need by no means be connected let alone Eulerian. If it is Eulerian however, by construction
has at most
min
vertices, where
, i.e.,
for all
and
,
. Those elements of
which do lift to an Eulerian path of
will be called admissible (with respect to T). It is clear that
is admissible if and only if
is an Eulerian path of
. For the remainder of this section, we introduce Swan’s theorem and our main results.
Swan [1]. Let
be an Eulerian graph with d edges and k vertices satisfying
. Then
has the same number of odd and even permutations (with respect to the fixed order)
Theorem 1. Let
be an Eulerian graph with vertex set
and d edges. Further let
be an integer such that
.
Then
is a polynomial identity on the ring
of
matrices over a commutative ring C with 1 Corollary 2. Let
be an Eulerian graph with k vertices and d edges. Further let
be an integer and assume that
. Then
is a polynomial identity on
.
3. Proof of Theorem 1
Since
is multilinear, it suffices to show that
for any substitution T of
matrix units over C. Fix such an T and put
,
. Then
(*)
Now consider
. Clearly, and summand in (*) vanishes unless, for the given
,

for all
, i.e., if
is admissible. If so, on multiplying the matrix units, we obtain
.
It follows that
where the inner sum is taken over all admissible permutations with
and
. If no such admissible
exists, the inner sum is 0 by definition. We want to prove that this inner sum is 0 anyway. It is readily seen that for any choice of u and b, a sum and
in the inner sum arises precisely of
lifts to an Eulerian path of
from (p, (u) to (q, v). Thus, on applying Swan’s theorem to
with
and
, we find that the number of even and odd admissible permutations
with
and
coincide whence the inner sum is 0 for any choice of u and v. This completes the proof.
4. Applications
1) Let
be the Eulerian graph on one vertex with d loops. Then
and

the standard polynomial [2] in d indeterminates.
More generally, let
be the Eulerian graph on k vertices with distinguished points
and the number
of edges from vertex i to j:

Now clearly
and
k times. On putting
and labelling the indeterminates, corresponding to the edges from i to
by
, from the corollary 2 it follows that

is a polynomial identity on
[3] if
, i.e., if
.
2) For
we define a sequence
,
of staircase steps, and the staircase height
. We will construct a substitution T, such that
lifts to the unique convering directed path of
(i.e.,
). First define a function
by
;
,
,
.
Next we define by recursion the sequence of pair
,
, where
is a natural number and
is a subset of
. We put
and
. Having
in hand
. There are three cases to consider:
a)
b)
,
c)
,
.
We now put

and

Let
for all
, it is clear that
gives a substitution of
matrix units over C. Now

is the unique covering directed path of
from
to
[4-7]. Since the
entry of the
matrix
is
, we have Theorem 3. Let
be an Eulerian graph and
. If
, then
is not a polynomial identity on the ring
of
matrices over a commutative ring C with 1.
Remark. It is an obvious consequence of the above theorem that if
is not the least integer
for which
is not a polynomial identity on
.
We note that, in general min
is not the least integer
for which
is not a polynomial identity on
.
Let
be the Eulerian graph on one vertex d loops. It is easily see that

Thus
for all
and the minimality assertion of the Amitsur-Levitzki theorem follows; the main part is an immediate consequence of the corollary.
Let
be the Eulerian graph on k vertices with distinguished points
and the number
of edges from vertex i to j:

Analogously, for any
we have

In consequence
for all
.
For
we get the double Capelli polynomial; it is known, however, that in this case
is not the smallest n for which
is not a polynomial identity on
.
When
we use x, y and z instead of the symbols
,
, and
respectively to denote the indeterminates of the triple Capelli polynomial and continue to write m for the number of edges from vertex i to i + 1. Thus the triple Capelli polynomial is
