DEPENDENCE OF HOMOGENEOUS COMPONENTS OF POLYNOMIALS

WITH SMALL DEGREE OF POISSON BRACKET

DARIA HOLIK, MAREK KARAŚ

arXiv:2101.09481v1 [math.AC] 23 Jan 2021

Abstract. Let F, G ∈ C[x1 , . . . , xn ] be two polynomials in n variables x1 , . . . , xn over the

complex numbers field C. In this paper, we prove that if the degree of the Poisson bracket
[F, G] is small enough then there are strict constraints for homogeneous components of these
polynomials (see Section 4). We also prove that there is a relationship between the homogeneous
components of the polynomial F of degrees deg F − 1 and deg F − 2 (see Section 6) as well some
results about divisibility of the homogeneous component of degree deg F − 1 (see Section 5).
Moreover we propose, possibly an appropriate, reformulation of the conjecture of Yu regarding
the estimation of the Poisson bracket degree of two polynomials (see Conjecture 1.1).

1. introduction
The research of this paper was inspired by the result of two papers. One of them is the result
of the paper Polynomial mappings with Jacobian determinant of bounded degree due to F. Leon
Pritchard [12] and the second one is the result from the paper Degree estimate for commutators
due to Vesselin Drensky and Jie-Tai Yu [1] (see also [11] and [8]).
In his paper, Pritchard considered the following problem. For a given polynomial mapping
(F, G) : C2 → C2 with small degree of the Jacobian determinant, i.e. with small degree of the
polynomial ∂F ∂G ∂F ∂G
∂x ∂y − ∂y ∂x , where F, G ∈ C[x, y] are there some relations between homogeneous
components of F and G. The main result of the paper gives such a relation.
On the other hand, Drensky and Yu, in their article consider the lower bound for the degree
of the Poisson bracket (in the commutative case) or of the degree of the commutator (in the
non-commutative case) of two polynomials F and G in n variables x1 , . . . , xn . Their considered
the conjecture which was formulated by Yu [18]. This conjecture says that, with some additional
assumptions, the following inequality holds:
deg[F, G] > min{deg F, deg G}.
Unfortunately the conjecture was showed to be false and, in the paper [1], there was presented
the counterexample, to the conjecture of Yu, due to Makar-Limanov (see [1, Example 1.1]). For-
tunately this example has some bad property. For instance this property can not be possessed
by two polynomials that are components of the same polynomial authomorphism of Cn . To be
precise, let F = F1 + · · · + Fd and G = G1 + · · · + GN , where Fi and Gj are homogeneous polyno-
mials of degree i and j, respectively. In the example of Makar-Limanov the linear components of
both polynomials F and G are linearly dependend. In the case when F and G are components
of the same polynomial authomorphism of Cn the linear components of F and G must be lineary
independent. Thus, it is natural to state the following, new version of the conjecture of Yu:
Conjecture 1.1. Let F and G be algebraically independent polynomials in C[x1 , . . . , xn ]. Let F
and G generate their own centralizers in C[x1 , . . . , xn ], respectively. Suppose, also, that deg F ∤
2020 Mathematics Subject Classification. Primary 13F20, 14R10; Secondary 16W20.
Key words and phrases. Poisson bracket, degree of Poisson bracket, polynomial ring, Jacobian determinant.
Corresponding author: Daria Holik, e-mail : holikd@agh.edu.pl.
1
2 Daria Holik, Marek Karaś

deg G, deg G ∤ deg F, F (0) = G(0) = 0 and that linear parts of F and G are linearly independent.
Then deg[F, G] > min{deg F, deg G} or maybe some other (weaker) inequality holds, for example,
deg[F, G] > c · min{deg F, deg G} for some constant c which is not dependent of F and G.
Let us notice that the assumption that the homogeneous components of maximal degree of F
and G are algebraically dependent, which appears in the original conjecture of Yu, is not nec-
essary, because if the homogeneous components of maximal degree of F and G are algebraically
independent then deg[F, G] = deg F + deg G > min{deg F, deg G}.
The problem of estimation from below of the degree of the Poisson bracket [F, G] is, of course,
connected with the results about estimation of the lower bound of the degree of nonconstant
elements of subalgebras generated by two polynomials. The first such an estimation was given
by Shestakov and Umirbaev [14]. Their used this estimation and the theorem saying that each
tame automorphism of C3 admits an elementary reduction or a reduction of some other types,
to show that the famous Nagata example
(1) σ : C3 ∋ (x, y, z) 7→ (x + 2y(y 2 + zx) − z(y 2 + zx)2 , y − z(y 2 + zx), z) ∈ C3
is wild. For information about four types of reductions defined by Shestakov and Umirbaev see
for example [13]. In the paper [9] Shigeru Kuroda showed that there is no tame automorphism
that admits a reduction of type IV.
Later in 2009 Leonid Makar-Limanov and Jie-Tai Yu [11] make an improvement of the esti-
mation due to Shestakov and Umirbaev. The improved estimation is the following
(2) deg P (f, g) ≥ D(f, g)wdeg f,deg g P,
where wdeg f,deg g P for polynomial P = P (x, y) in two variables x and y denotes the weighted
degree of P such that wdeg f,deg g x = deg f and wdeg f,deg g y = deg g, and
gcd(deg f, deg g) − (deg(f g) − deg[f, g])
(3) D(f, g) = 1 − .
deg f deg g
In the paper [11] the authors used, instead of deg[f, g], the equivalent, in this context, notion of
the degree of the following differential form df ∧ dg. On the other hand in [8] Kuroda generalized
Shestakov-Umirbaev inequality to the case of more than two polynomials. In his article the degree
of the Poisson bracket was replaced, which was very natural, by the degree of the differential
form which is obtained as the wedge product of the differentials of these polynomials.
The numerous possible results about existence or nonexistence of tame automorphisms P =
(P1 , P2 , P3 ) : C3 → C3 with a given multidegree of P (by multidegree of P we mean mdeg P =
(deg P1 , deg P2 , deg P3 )) would be possible to be achieved if the Conjecture 1.1 appears to be
true. For some results about the multidegrees see for example [2, 4, 5, 6, 7, 16].
One of a possible way of attacking the above conjecture or maybe some weaker conjecture (with
stronger assumptions) is to prove that small degree of deg[F, G] implies the relations between
F = F1 + · · · + Fd and G = G1 + · · · + GN (here, by Fi and Gj we denote the homogeneous
component of F of degree i and the homogeneous component of G of degree j, respectively)
and then show that such a relation can not be satisfied. On the other hand such a relation can
be viewed as a generalization of the  result of Pritchard.
 Indeed, in the case F, G ∈ C[x, y] it
∂F ∂G ∂F ∂G
is known that deg[F, G] = 2 + deg ∂x ∂y − ∂y ∂x . Since in the paper, we study the pair of
polynomials F, G with small degree of [F, G] we will assume that deg[F, G] < deg F + deg G.
This implies that [Fd , GN ] = 0, and so by Lemma 2 in [17] there exist a, b ∈ C, k1 , k2 ∈ N and
a homogeneous polynomial h such that Fd = ahk1 and GN = bhk2 . Of course, withuot loss of
generality, we can assume that h is not a power of any other polynomial of lower degree, and
that a = 1, because C is algebraically closed. Thus, in the rest of the paper, we will assume that
 Dependence of Homogeneous Components 3

Fd = hk1 and GN = bhk2 with b ∈ C⋆ = C \ {0} denoted by aN . Of course, in this setup, we have
N
k1 = degd h and k2 = deg h.
The paper is organized as follows. In Section 2, we recall the notion of Poisson bracket of
two polynomials, formulate some properties that can be used in the calculations and present
the main tool of the paper the so-called H-reduction method. In Section 3, we give some
auxilliary lemmas that can be used in the next section in computation. Section 4 is devoted
for establishing the general formulas for the homogeneous components of the polynomial G in
terms of the homogeneous components of the polynomial F. Then, in the following section we
use the above mentioned general formulas to prove that if deg[F, G] is small enough then the
homogeneous components of the polynomial F of degree deg F − 1 is divisible by h in the case
of deg h ≤ 2 and by sqrf h in the case deg h > 2 (for definition of sqrf h see Section 5), where
h is the polynomial such that the homogeneous components of maximal degree of F and G are
proportional to some powers of h. Under the assumption that h is a square-free polynomial we
obtain that if deg[F, G] is smaller and smaller then Fs (the homogeneous component of F of
degree s, where s = deg F − 1) is divisible by higher and higher power of h. The next section
give some other relations between homogeneous components of the polynomial F in some special
cases of deg F. More precisely, we show that there is a strong relation between Fs−1 and Fs , but
with some assumptions about the degree of F and G, and of course, about deg[F, G].
Let us notice that the assumption that h is square-free geometrically means that the compo-
nents at infinity of F = 0 (and G = 0) all have the same multiplicity. The notion of square-free
polynomials seems to be naturally arrising in the context of Jacobian Conjecture or related topics
(see for example [3]).
At the end of the introduction let us notice that all given in the present article results are
valid not only for C but for any algebraically closed field of characteristic zero.

2. Poisson bracket and H-reduction method

In this section we develop the main tool that will be used in the next sections of the paper.
We start with the following lemma in which we use the notation
C[x1 , . . . , xn ]d = {f ∈ C[x1 , . . . , xn ] : f is homogeneous of degree d} ∪ {0}.
Lemma 2.1. Let H be a nonconstant homogeneous polynomial that is not a power of any other
polynomial of lower degree (in particular, if H is a square-free polynomial) and let P be any
homogeneous polynomial such that
[H, P ] = 0.
Then there exist a ∈ C and k ∈ N such that
P = aH k .
Moreover, if P ∈ C[x1 , . . . , xn ]d and deg H ∤ d, then P = 0.
For the convenience of the reader, before we give the proof of the above lemma, let us recall
that for any f, g ∈ C[x1 , . . . , xn ], by [f, g] we denote the Poisson bracket of f and g, which is the
following sum:
X  ∂f ∂g ∂f ∂g

− [xi , xj ] ,
∂xi ∂xj ∂xj ∂xi
1≤i<j≤n
where [xi , xj ] can be viewed as formal objects satisfying the condition (see for example [13])
[xi , xj ] = −[xj , xi ] for all i, j.
We also define
deg [xi , xj ] = 2 for all i 6= j,
4 Daria Holik, Marek Karaś

deg 0 = −∞ and
  
∂f ∂g ∂f ∂g
deg [f, g] = max deg − [xi , xj ] .
1≤i<j≤n ∂xi ∂xj ∂xj ∂xi

Since 2 − ∞ = −∞, we have

 
∂f ∂g ∂f ∂g
deg[f, g] = 2 + max deg − .
1≤i<j≤n ∂xi ∂xj ∂xj ∂xi

From the above equality we have

(4) deg [f, g] ≤ deg f + deg g

and deg[f, g] ≥ 2 iff f and g are algebraically independent.

Proof of Lemma 2.1. Since [H, P ] = 0, it follows that H and P are algebraically dependent
and so by Lemma 2 in [17] there exist a, b ∈ C, k1 , k2 ∈ N and a homogeneous polynomial h such
that
P = ahk1 and H = bhk2 .
Since H is not a power of any other polynomial of lower degree, we conclude that k2 = 1 and so
we can take h = H. Thus, in particular, if a 6= 0 (i.e. P 6= 0), then deg P is divisible by deg H.

The mentioned H-reduction method is the statement given in the following corollary.

Corollary 2.2. Let H be a nonconstant homogeneous polynomial that is not a power of any
other polynomial of lower degree and let P be any polynomial such that

[H, P ] = 0.

Then P ∈ C[H].

Proof. Let d = deg P and let P = P0 + · · · + Pd be the homogeneous decomposition of P.

Since [H, P ] = 0, it follows that

(5) [H, Pi ] = 0 for i = 0, . . . , d.

In particular, [H, Pd ] = 0. Since Pd 6= 0 (by definition of d), it follows that d = k deg H for some
k ∈ N. By (5) and Lemma 2.1 there exist a0 , . . . , ak ∈ C, ak 6= 0 such that

Pl deg H = al H l for l = 0, . . . , k

and Pi = 0 for i ∈
/ {0, deg H, 2 deg H, . . . , k deg H}. 
We will also use the following fact that is easy to check.

Lemma 2.3. Let P ∈ C[x1 , . . . , xn ]. For any Q, R ∈ C[x1 , . . . , xn ] and α, β ∈ C we have

[P, QR] = Q [P, R] + R [P, Q] ,

[P, αQ + βR] = α [P, Q] + β [P, R] ,
[P, Q] = − [Q, P ] .

In other words, the mappings Q 7→ [P, Q] and Q 7→ [Q, P ] are C-derivations.

 Dependence of Homogeneous Components 5

3. Preparatory lemmas
We assume that are given the following two polynomials of degree, respectively, d and N with
d≤N :
(6) F = F1 + · · · + Fd and G = G1 + · · · + GN ,
where Fi and Gj are homogeneous polynomials of degree i and j, respectively, with
d N
(7) Fd = h t and GN = aN h t ,
where h is a homogeneous polynomial of degree t and aN is a nonzero constant. Of course, we
must have t|d and t|N.
Let us notice that:
s
(8) [Fi , F1α1 · · · Fsαs hα0 ] = α0 F1α1 · · · Fsαs hα0 −1 [Fi , h] + αj F1α1 · · · Fsαs Fj−1 hα0 [Fi , Fj ],
X

j=1

where s = d − 1.
Lemma 3.1. Assume that polynomials
cα(u) F1α1 · · · Fsαs hα0
X
(9) Gu = u = i, . . . , i + s − 1,
α=(α0 ,...,αs )∈Z×Ns

(u)
with cα ∈ C, satisfy the following condition:
(i+s−l) (i+s−k)
(C1) (αk + 1)cα+ek = (αl + 1)cα+el , for all 1 ≤ k, l ≤ s and α ∈ Z × Ns ,

where e0 = (1, 0, . . . , 0), e1 = (0, 1, 0, . . . , 0), . . . , es = (0, . . . , 0, 1). Then, the sum [Fs , Gi ] +
[Fs−1 , Gi+1 ] + · · · + [F1 , Gi+s−1 ] can be written as

α0 (i+s−1) α1
X
αs α0 −1
(10) 
α1 cα−e1 F1 · · · Fs h
α∈Z×N+ ×Ns−1
α0 (i+s−2) α2
X
+ α2 cα−e2 F2 · · · Fsαs hα0 −1
α∈Z×{0}×N+ ×Ns−2
..
. 
α0 (i)
X
αs α0 −1
+ αs cα−es Fs h , h .
α∈Z×{(0,...,0)}×N+

Proof. By (8) we see that [Fs , Gi ]+[Fs−1 , Gi+1 ]+· · ·+[F1 , Gi+s−1 ] is the sum of the following
summands:
n o
(i+s−l) (i+s−k)
F1α1 · · · Fsαs hα0 (αk + 1)cα+ek [Fl , Fk ] + (αl + 1)cα+el [Fk , Fl ]
X
(11)
α∈Z×Ns

for all 1 ≤ k < l ≤ s and

s
α0 cα(i+s−k) F1α1 · · · Fsαs hα0 −1 [Fk , h].
X X
(12)
k=1 α∈Z×Ns

Let us notice that (11) equals to zero by (C1).

6 Daria Holik, Marek Karaś

(u)
We can assume that for all u, c(α0 ,...,αs ) = 0, if αk < 0 for some k = 1, . . . , s. Then, by
renumbering of the summs in (12), the summand (12) can be rewritten as follows
n
(i+s−1) (i+s−2)
α0 hα0 −1 cα−e1 F1α1 −1 F2α2 · · · Fsαs [F1 , h] + cα−e2 F1α1 F2α2 −1 F3α3 · · · Fsαs [F2 , h]
P
(13)
α∈Z×Ns
o
(i) αs−1 αs −1
+ · · · + cα−es F1α1 · · · Fs−1 Fs [Fs , h] .
Assume that αk 6= 0 for some k ∈ {1, . . . , s} and set e0 = (1, 0, . . . , 0), e1 = (0, 1, 0, . . . , 0), . . . ,
es = (0, . . . , 1). Then, one can see that
h i
1 (i+s−k) α1 αs
(14) c
αk α−ek F1 · · · Fs , h
α1 (i+s−k) α1 −1 α2 (i+s−k)
= αk cα−ek F1 F2 · · · Fsαs [F1 , h] + ααk2 cα−ek F1α1 F2α2 −1 F3α3 · · · Fsαs [F2 , h]
(i+s−k) αs−1 αs −1
+ · · · + ααks cα−ek F1α1 · · · Fs−1 Fs [Fs , h]
(i+s−1) (i+s−2)
= cα−e1 F1α1 −1 F2α2 · · · Fsαs [F1 , h] + cα−e2 F1α1 F2α2 −1 F3α3 · · · Fsαs [F2 , h]
(i) α
+ · · · + cα−es F1α1 · · · Fs−1
s−1
Fsαs −1 [Fs , h].
Indeed, by assumption (C1), for l ∈ {1, . . . , s}, we have
(i+s−k) (i+s−l)
(15) [(αl − 1) + 1]c(α−ek −el )+el = [(αk − 1) + 1]c(α−ek −el )+ek .
/ Z × Ns and we can assume that
If we have αl = 0 in the above equality, then (α − ek − el ) + ek ∈
(i+s−l)
c(α−ek −el )+ek = 0.
Thus, the summand (12) can be written as in (10). Since the summands (11) are equal to
zero, the result follows. 
Lemma 3.2. Assume that h is a homogeneous polynomial that is no power of any other poly-
nomial of degree < deg h, polynomials Gi , . . . , Gi+s−1 , defined as in Lemma 3.1, satisfy the
condition (C1). If polynomial Gi−1 is such that [hr , Gi−1 ] + [Fs , Gi ] + · · · + [F1 , Gi+s−1 ] = 0, for
some r ∈ N+ , then
F1α1 · · · Fsαs hα0
X
Gi−1 = c(i−1)
α
α=(α0 ,...,αs )∈Z×Ns
(i−1)
with cα ∈ C such that for all α = (α0 , . . . , αs ) ∈ Z × Ns we have:
(i−1) α0 (i+s−l)
(P1) if αl > 0 with l ∈ {1, . . . , s}, then cα−re0 = rαl cα−el .

Proof. Since [hr , Gi−1 ] = rhr−1 [h, Gi−1 ] = [h, rhr−1 Gi−1 ], it follows by Lemma 3.1 and
Corollary 2.2 that

1  X
α0 (i+s−1) α1 αs α0 −1
(16) Gi−1 = α1 cα−e1 F1 · · · Fs h
rhr−1
α∈Z×N+ ×Ns−1
α0 (i+s−2) α2
X
αs α0 −1
+ α2 cα−e2 F2 · · · Fs h
α∈Z×{0}×N+ ×Ns−2
..
. 
α0 (i)
X
αs α0 −1 
+ αs cα−es Fs h + P (h),
α∈Z×{(0,...,0)}×N+
 Dependence of Homogeneous Components 7

for some polynomial P in one variable, over k. Thus, if l = 1 (and α1 > 0), then the equation in
(P1) is a direct consequence of (16).
Assume that l > 1 and αl > 0. If α1 = . . . = αl−1 = 0, then the equality in (P1) is, as above,
a direct consequence of (16). Thus, we can assume that there is k < l such that k ≥ 1 and
αk > 0. We can also assume that we have already proved that
(i−1) α0 (i+s−k)
(17) cα−re0 = rαk cα−ek .

Now, using (C1) we obtain that

(i+s−l) (i+s−k)
(18) [(αk − 1) + 1] c(α−el −ek )+ek = [(αl − 1) + 1] c(α−el −ek )+el
and so
α0 (i+s−l) α0 (i+s−k)
(19) rαl cα−el = rαk cα−ek .

This completes the proof. 

Lemma 3.3. Assume that r ∈ N+ , h is a homogeneous polynomial that is no power of any other
polynomial of degree < deg h, polynomials Gi , . . . , Gi+s−1 , defined as in Lemma 3.1, satisfy the
condition (C1) and with polynomial Gi+s , defined similarly as in Lemma 3.1, satisfy the following
condition:
for all 1 ≤ l, k ≤ s and α = (α0 , . . . , αs ) ∈ Z × Ns
(C2) (i+l−1) α0 (i+l+s−k)
if αk > 0, then cα−re0 = rα c
k α−ek

If polynomial Gi−1 is such that [hr , Gi−1 ] + [Fs , Gi ] + . . . + [F1 , Gi+s−1 ] = 0, then
(a) Gi−1 , Gi , . . . , Gi+s−2 satisfy the condition (C1),
(b) Gi−1 , Gi , . . . , Gi+s−1 satisfy the condition (C2),
all with the same r and i replaced by i − 1.
Proof. We start with the proof of (b). Take any 1 ≤ l, k ≤ s and α = (α0 , . . . , αs ) ∈ Z × Ns
with αk > 0.
Assume that l ≥ 2. Then, by (C2) we have
((i−1)+l−1) (i+(l−1)−1) α0 (i+(l−1)+s−k) α0 ((i−1)+l+s−k)
(20) cα−re0 = cα−re0 = rαk cα−ek = rαk cα−ek .
For l = 1 we can use property (P1) with l replaced by k.
Now, we prove (a). Take any 1 ≤ k, l ≤ s and α = (α0 , . . . , αs ) ∈ Z × Ns . Let us notice that
((i−1)+s−k) (i+(s−k)−1) (i+(s−k)+s−l)
(21) (αl + 1)cα+el = (αl + 1)cα+el = (αl + 1) r(ααl 0+1) cα+re0 .
Indeed, if k = s, then it is a consequence of (P1):
(i+(s−k)−1) (i−1) (i−1)
(22) (αl + 1)cα+el = (αl + 1)cα+el = (αl + 1)c(α+el +re0 )−re0
(i+s−l) (i+(s−k)+s−l)
= (αl + 1) r(ααl 0+1) cα+re0 = (αl + 1) r(ααl 0+1) cα+re0
and if k < s it follows from (C2):

(i+(s−k)−1) (i+(s−k)−1) (i+(s−k)+s−l)

(23) (αl + 1)cα+el = (αl + 1)c(α+el +re0 )−re0 = (αl + 1) r(ααl 0+1) cα+re0 .
Similarly, we have
((i−1)+s−l) (i+(s−l)−1) (i+(s−l)+s−k)
(24) (αk + 1)cα+ek = (αk + 1)cα+ek = (αk + 1) r(ααk0+1) cα+re0 .
8 Daria Holik, Marek Karaś

By (21) and (24), we have

((i−1)+s−k) ((i−1)+s−l)
(25) (αl + 1)cα+el = (αk + 1)cα+ek .

The following lemma is obvious.
Lemma 3.4. Assume that h is a homogeneous polynomial such that deg h divides N. Let GN =
N
αN h deg h , GN +1 = . . . = GN +s = 0. Then
(a) polynomials GN , GN +1 , . . . , GN +s−1 satisfy the condition (C1),
(b) for any r ∈ N+ , polynomials GN , GN +1 , . . . , GN +s satisfy the condition (C2),
all with i replaced by N.

4. Formulas for Gj
The main result of this section is the following formula for polynomials Gj .
Theorem 4.1. Let F = F1 + · · · + Fs + Fd and G = G1 + · · · + GN be elements of C[x1 , . . . , xn ],
where Fi and Gj are homogeneous polynomials of degree i and j, respectively. Assume that
(26) Fd = hd and GN = aN hN
for some homogeneous polynomial h of degree 1 and aN ∈ C \ {0}.
If deg[F, G] < d + i for i ∈ {1, . . . , N }, then there are ai , . . . , aN −1 ∈ C such that for j =
i, i + 1, . . . , N, we have
α1
X
(27) Gj = c(j) αs α0
α F1 · · · Fs h ,
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =j
α0 +d(α1 +···+αs )≤N

where for α = (α0 , . . . , αs ) ∈ Z×Ns with α0 +α1 +2α2 +· · ·+sαs = j and α0 +d(α1 +· · ·+αs ) ≤ N,
we have
Qα1 +···+αs
(j) (α0 + dk)
(28) cα = Qα1 k=1
Qαs
aα0 +d(α1 +···+αs ) .
i1 =1 di1 · · · is =1 dis
(j)
Proof. In what follows, we assume that c(α0 ,...,αs ) = 0 if α0 + α1 + 2α2 + · · · + sαs 6= j or
α0 + d(α1 + · · · + αs ) > N.
Let us notice that the above formula holds for GN . Indeed, if (α1 , . . . , αs ) ∈ Ns \ {(0, . . . , 0)}
and α0 + α1 + 2α2 + · · · + sαs = N, then α0 + d(α1 + · · · + αs ) > N. Hence, in the sum (27), for
GN , there is only one summand aN hN . Let us also notice that the above formula holds for the
polynomials GN +1 = . . . = GN +s = 0.
By Lemma 3.4, we know that the polynomials GN , . . . , GN +s−1 satisfy the condition (C1),
and that the polynomials GN , . . . , GN +s satisfy the condition (C2). Assume that i ≤ N − 1.
Then, deg[F, G] < d + N − 1 and so [hd , GN −1 ] + [Fs , GN ] + · · · + [F1 , GN +s−1 ] = 0. Since
GN , . . . , GN +s−1 satisfy the condition (C1), we see from Lemma 3.2 that the polynomial GN −1
is of the form
−1) α1
X
GN −1 = c(N
α F1 · · · Fsαs hα0
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =N −1
and satisfy (P1). But, since GN = aN hN and GN +1 = . . . = GN +s−1 = 0, we see, from (P1),
s (N −1)
that if (α1 , . . . , αs ) ∈ N \ {(0, . . . , 0), (0, . . . , 0, 1)}, then c(α0 ,...,αs ) = 0. For αs = 1, by (P1), we
have
(N −1) N (N )
c(N −d,0,...,0,1) = d·1 c(N,0,...,0) = d·1N
aN = (N −d)+d·1
d·1 a(N −d)+d(0+···+0+1) .
 Dependence of Homogeneous Components 9

(N −1)
Setting aN −1 = c(N −1,0,...,0) , we obtain that GN −1 satisfy formulas (27) and (28). By Lemma 3.3,
we also know that the polynomials GN −1 , . . . , GN +s−2 satify the condition (C1), and that the
polynomials GN −1 , . . . , GN +s−1 satify the condition (C2). Thus, we can use the induction.
Now, assume that for some l ∈ {i + 1, i + 2, . . . , N }, polynomials Gl , . . . , Gl+s are given by
the formulas (27) and (28). Assume also that Gl , . . . , Gl+s−1 satisfy the condition (C1) and
Gl , . . . , Gl+s satisfy the condition (C2). Since deg[F, G] < d + i ≤ d + (l − 1), it follows that
[hd , Gl−1 ] + [Fs , Gl ] + · · · + [F1 , Gl+s−1 ] = 0. Hence, by Lemma 3.2, the polynomial Gl−1 is of
the form

cα(l−1) F1α1 · · · Fsαs hα0

X
(29) Gl−1 =
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =l−1

and satisfy (P1). Moreover, by Lemma 3.3, the polynomials Gl−1 , . . . , Gl+s−2 satisfy the con-
dition (C1) and Gl−1 , . . . , Gl+s−1 satisfy the condition (C2). Since Gl−1 satisfy (P1), we have
that

α0 +d (l+s−1) α1
X
(30) Gl−1 = al−1 hl−1 + dα1 c(α0 +d,α1 −1,α2 ,...,αs ) F1 · · · Fsαs hα0
α=(α0 ,...,αs )∈Z×N+ ×Ns−1
α0 +α1 +2α2 +···+sαs =l−1
α0 +d (l+s−2) α2
X
+ dα2 c(α0 +d,0,α2 −1,α3 ,...,αs ) F2 · · · Fsαs hα0
α=(α0 ,...,αs )∈Z×{0}×N+ ×Ns−2
α0 +α1 +2α2 +···+sαs =l−1
..
.
α0 +d (l+s−s)
X
αs α0
+ dαs c(α0 +d,0,...,0,αs −1) Fs h .
α=(α0 ,...,αs )∈Z×{(0,...,0)}×N+
α0 +α1 +2α2 +···+sαs =l−1

(l+s−1)
Notice that c(α0 +d,α1 −1,α2 ,...,αs ) can be different from zero only if (α0 +d)+d(α1 −1+α2 +· · ·+αs ) ≤
N (because there is no components with degree > N in the polynomial G), that is only if
(l+s−2) (l)
α0 + d(α1 + α2 + · · · + αs ) ≤ N. The same holds for c(α0 +d,0,α2 −1,α3 ...,αs ) , . . . , c(α0 +d,0,...,0,αs −1) .
Hence, in (29), we can add the restriction α0 + d(α1 + · · · + αs ) ≤ N.
Assume that α1 > 0, α0 + α1 + 2α2 + · · · + sαs = l − 1 and α0 + d(α1 + · · · + αs ) ≤ N. Then,
by (30) and the above assumptions, we have

(l−1) α0 +d (l+s−1)
(31) c(α0 ,...,αs ) = dα1 c(α0 +d,α1 −1,α2 ,...,αs )
Qα1 −1+···+αs
((α + d) + dk)
= α0 +d  k=1
dα1 Qα1 −1  0Q
a(α0 +d)+d((α1 −1)+α2 +···+αs )
αs
i1 =1 di1 · · · is =1 dis
Qα1 +···+αs
(α0 + dk)
= Qα1 k=1
Qαs
aα0 +d(α1 +α2 +···+αs ).
i1 =1 di1 · · · is =1 dis
10 Daria Holik, Marek Karaś

Now, assume that α1 = 0, α2 > 0, α0 + 2α2 + · · · + sαs = l − 1 and α0 + d(α2 + · · · + αs ) ≤ N.

Similarly, as before, we have
(l+s−2) α0 +d (l+1)
(32) c(α0 ,...,αs ) = dα2 c(α0 +d,0,α2 −1,α3 ,...,αs )
Qα2 −1+α3 +···+αs
α0 +d k=1 ((α0 + d) + dk)
= dα2
Qα2 −1  Qαs
a(α0 +d)+d((α2 −1)+α3 +···+αs )
i2 =1 di2 · · · is =1 dis
Qα2 +α3 +···+αs
(α0 + dk)
= Qαk=1
2

Qαs
aα0 +d(α2 +α3 +···+αs )
i2 =1 di2 · · · is =1 dis

The similar calculation shows that formula (28) holds also in the cases α1 = . . . = αq = 0,
and αq+1 > 0, with α0 + α1 + 2α2 + · · · + sαs = l − 1 and α0 + d(α1 + · · · + αs ) ≤ N, for all
q = 3, . . . , s − 1.
Since, by Lemma 3.3 we can use induction, the result follows. 
Using similar arguments as above one can prove the following result in which we consider the
case deg h ≥ 1; interested one is deg h > 1. Of course, the following theorem is a generalization
of the above one.
Theorem 4.2. Let F = F1 + · · · + Fs + Fd and G = G1 + · · · + GN be elements of C[x1 , . . . , xn ],
where Fi and Gj are homogeneous polynomials of degree i and j, respectively. Assume that
h ∈ C[x1 , . . . , xn ] is a homogeneous polynomial that is no power of any other polynomial of
degree < deg h, such that deg h| gcd(d, N ) and
(33) Fd = hd/ deg h and GN = aN hN/ deg h
for some aN ∈ C \ {0}. In what follows, we put r = degd h and t = deg h.
If deg[F, G] < d + i for i ∈ {1, . . . , N }, then there are ai , . . . , aN −1 ∈ C such that for j =
i, i + 1, . . . , N, we have
α1
X
(34) Gj = c(j) αs α0
α F1 · · · Fs h ,
α=(α0 ,...,αs )∈Z×Ns
tα0 +α1 +2α2 +···+sαs =j
tα0 +d(α1 +···+αs )≤N

where for α = (α0 , . . . , αs ) ∈ Z×Ns with tα0 +α1 +2α2 +· · ·+sαs = j and tα0 +d(α1 +· · ·+αs ) ≤
N, we have
Qα1 +···+αs
(j) (α0 + rk)
(35) cα = Qα1 k=1
Qαs
atα0 +d(α1 +···+αs ) .
i1 =1 ri1 · · · is =1 ris

Moreover, we have aj = 0 when deg h ∤ j.

5. Divisibility of Fs
In this section, we show that, if deg[F, G] is small enough then Fs is divisible by h (see
Theorem 5.1) or even by suitable power of h (see Theorem 5.5). This will be established in the
case deg h = 1 and in the case deg h > 1 with the additional assumption that h is a square-free
polynomial (see Theorem 5.8). In the case deg h > 1 but without assumption that h is square-free,
we obtain that Fs is divisible by sqrf h, where sqrf h is defined as the product of all irreducible
factors of the polynomial h but taken without multiplicity, for example sqrf x31 (x2 − 1)2 =

x1 (x2 − 1). A polynomial sqrf h for a given h ∈ C[x1 , . . . , xn ] is defined only up to a constant
factor u ∈ C \ {0}, but for our purpose it is not a disadvantage. If the polynomial h is itself
a square-free polynomial, that is with irreducible factors with multiplicity one, then we put
 Dependence of Homogeneous Components 11

sqrf h = h. For a general definition of square-free elements of any commutative rings we refer the
reader to [3].

5.1. The case deg h = 1. In this subsection we consider the case deg h = 1. First of all, let us
notice that if deg h ≤ 2 and h is no power of any other polynomial with degree < deg h, then
the polynomial h is square-free.
Theorem 5.1. Let F and G be as in Theorem 4.1. If N 6≡ 0 ( mod d) and deg[F, G] < d+i with
i < ds N = d−1d N, then Fs is divisible by h. In other words, there is a homogeneous polynomial
F̃s−1 of degree s − 1 such that Fs = F̃s−1 h.
In the proof of the above theorem we use the following
Lemma 5.2. For all integers 1 ≤ j < N, the minimal value
(36) min{ α0 | (α0 , . . . , αs ) ∈ Z × Ns , α0 + α1 + 2α2 + · · · + sαs = j, α0 + d(α1 + · · · + αs ) ≤ N }
is equal to dj − sN and is attained at only one point (dj − sN, 0, . . . , 0, N − j).
Proof. Let Zj,N,d := { (α0 , . . . , αs ) ∈ Z×Ns | α0 +α1 +2α2 +· · ·+sαs = j, α0 +d(α1 +· · ·+αs ) ≤
N }. Notice that for all (α0 , . . . , αs ) ∈ Zj,N,d , we have sα1 + (s − 1)α2 + · · · + 2αs−1 + αs ≤ N − j
(in particular, αs ≤ N − j). Thus, sαs ≤ s(N − j) − s (sα1 + (s − 1)α2 + · · · + 2αs−1 ) and
α1 + 2α2 + · · · + sαs ≤ s(N − j) − [s2 − 1]α1 − [s(s − 1) − 2]α2 − · · · − [s · 2 − (s − 1)]αs−1 .
Since s2 − 1, s(s − 1) − 2, . . . , s · 2 − (s − 1) are strictly positive numbers, it follows that for
(α1 , . . . , αs ) ∈ Zj,N,d if (α1 , . . . , αs−1 ) 6= (0, . . . , 0) ∈ Ns−1 , then α0 > dj − sN. Of course, if
(α0 , 0, . . . , 0, αs ) ∈ Zj,N,d with αs < N − j, then also α0 = j − sαs > j − s(N − j) = dj − sN. 
Now we can prove Theorem 5.1.
Proof of Theorem 5.1. Since deg[F, G] < d + i, we see by Theorem 4.1 that

cα(i) F1α1 · · · Fsαs hα0 .

X
(37) Gi =
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =i
α0 +d(α1 +···+αs )≤N

(i)
By Lemma 5.2, all summands of (37), except c(di−sN,0,...,0,N −i) FsN −i hdi−sN , involves hu with
u > di − sN. Let us also notice that di − sN < 0. Thus, multiplying both sides of (37) by hsN −di ,
(i)
we obtain that h divides c(di−sN,0,...,0,N −i) FsN −i . Since N 6≡ 0( mod d), we see that −sN 6≡ 0(
QN−i
(i) k=1 (di−sN +dk)
mod d) and so c(di−sN,0,...,0,N −i) = QN−i aN 6= 0. Thus, we see that h|FsN −i and so
is =1 dis
h|Fs because h is a square-free polynomial. 
By Theorems 4.1 and 5.1, we obtain the following
Corollary 5.3. Let F and G be as in Theorem 4.1. If N 6≡ 0 ( mod d) and deg[F, G] < d + i
with i < ds N, then for j = i, i + 1, . . . , N, we have
α1 αs−1 αs α0 +αs
X
(38) Gj = c(j)
α F1 · · · Fs−1 F̃s−1 h ,
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =j
α0 +d(α1 +···+αs )≤N

(j)
where ai , . . . , aN −1 ∈ C and cα are as in Theorem 4.1, and F̃s−1 is as in Theorem 5.1.
The following lemma we use in the proof of Theorem 5.5 below.
12 Daria Holik, Marek Karaś

Lemma 5.4. For all k ∈ N such that d > 2k and for all integers 1 ≤ j < N, the minimal value
(39) min{ α0 + kαs | (α0 , . . . , αs ) ∈ Zj,N,d }
is equal to (d − k)j − (s − k)N and is attained at only one point (dj − sN, 0, . . . , 0, N − j).
Proof. As in the proof of Lemma 5.2, we have
(s − k)αs ≤ (s − k)(N − j) − (s − k) (sα1 + (s − 1)α2 + · · · + 2αs−1 ) .
From this we see that α1 + 2α2 + · · · + (s − 1)αs−1 + (s − k)αs ≤ (s − k)(N − j) − [(s − k)s −
1]α1 − [(s − k)(s − 1) − 2]α2 − · · · − [(s − k) · 2 − (s − 1)]αs−1 and so
(40) α0 + kαs = j − [α1 + 2α2 + · · · + (s − 1)αs−1 + (s − k)αs ]
≥ j − (s − k)(N − j) + [(s − 1)s − 1]α1
+[(s − k)(s − 1) − 2]α2 + · · · + [(s − k) · 2 − (s − 1)]αs−1
Since (s−k)s−1 > (s−k)(s−1)−2 > . . . > (s−k)·2−(s−1) = s−(2k−1) > 0, it follows that for
(α0 , . . . , αs ) ∈ Zj,N,d if (α1 , . . . , αs−1 ) 6= (0, . . . , 0) ∈ Ns−1 , then α0 + kαs > (d − 1)j − (s − 1)N.
Of course, if (α0 , 0, . . . , 0, αs ) ∈ Zj,N,d with αs < N − j, then also α0 + kαs = j − (s − k)αs >
j − (s − k)(N − j) = (d − k)j − (s − k)N. 
By the above lemma, one can use, several times, the similar arguments as in the proof of
Theorem 5.1 to show the following result (see also proof of Theorem 5.8 in the next subsection).
Theorem 5.5. Let F, G and h be as in Theorem 4.1. If N 6≡ 0 ( mod d) and deg[F, G] < d + i
s−k
with i < d−k N with 2k < d, then Fs is divisible by hk+1 . In other words, there is a homogeneous
polynomial F̃s−k−1 of degree s − k − 1 such that Fs = F̃s−k−1 hk+1 .
By Theorems 4.1 and 5.5, we obtain the following
Corollary 5.6. Let F and G be as in Theorem 4.1. If N 6≡ 0 ( mod d) and deg[F, G] < d + i
s−k
with i < d−k N with 2k < d, then we have
α1 αs−1 αs
X
(41) Gj = c(j)
α F1 · · · Fs−1 F̃s−k−1 h
α0 +(k+1)αs
,
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =j
α0 +d(α1 +···+αs )≤N

(j)
where ai , . . . , aN −1 ∈ C and cα are as in Theorem 4.1 and F̃s−k−1 is as in Theorem 5.5.
5.2. The case deg h > 1. The results given in this subsection are also true for deg h = 1, but
of course are interested in the case deg h > 1. We start with the following lemma.
Lemma 5.7. Let d, t and N be as in Theorem 4.2. For all k ∈ N such that d > 2kt and for all
integers 1 ≤ j < N, the minimal value
(t)
(42) min{ α0 + kαs | (α0 , . . . , αs ) ∈ Zj,N,d }
1
is equal to t [(d − kt)j − (s − kt)N ] and is attained at only one point ( 1t (dj −sN ), 0, . . . , 0, N −j),
(t)
where Zj,N,d = {(α0 , . . . , αs ) ∈ Z × Ns : tα0 + α1 + 2α2 + · · · + sαs = j, tα0 + d(α1 + · · · + αs ) ≤ N }
Proof. As in the proof of Lemma 5.2, we have
(s − kt)αs ≤ (s − kt)(N − j) − (s − kt) (sα1 + (s − 1)α2 + · · · + 2αs−1 ) .
From this we see that α1 + 2α2 + · · · + (s − 1)αs−1 + (s − kt)αs ≤ (s − kt)(N − j) − [(s − kt)s −
1]α1 − [(s − kt)(s − 1) − 2]α2 − · · · − [(s − kt) · 2 − (s − 1)]αs−1 .
 Dependence of Homogeneous Components 13

1
Since α0 = t [j − (α1 + 2α2 + · · · + sαs )] , it follows that

1 ktαs
(43) α0 + kαs = t [j − (α1 + 2α2 + · · · + sαs )] +
t
1
= t [j − (α1 + 2α2 + · · · + (s − 1)αs−1 + (s − kt)αs )]
1
≥ t {j + (s − kt)(j − N ) + [(s − kt)s − 1]α1 + [(s − kt)(s − 1) − 2]α2 +
· · · + [(s − kt)2 − (s − 1)]αs−1 } .

Since (s − kt)s − 1 > (s − kt)(s − 1) − 2 > . . . > (s − kt) · 2 − (s − 1) = s − (2kt − 1) > 0,

(t)
it follows that for (α0 , . . . , αs ) ∈ Zj,N,d if (α1 , . . . , αs−1 ) 6= (0, . . . , 0) ∈ Ns−1 , then α0 + kαs >
1 (t)
t{(d − kt)j − (s − kt)N } . Of course, if (α0 , 0, . . . , 0, αs ) ∈ Zj,N,d with αs < N − j, then also
α0 + kαs = 1t {j − (s − kt)αs } > 1t {j − (s − k)(N − j}) = 1t {(d − kt)j − (s − kt)N } . 

Now, we can use the above lemma and Theorem 4.2 to prove the following result.

Theorem 5.8. Let F, G and h be as in Theorem 4.2. Assume also that the polynomial h is
s−k
square-free. If N 6≡ 0 ( mod d) and deg[F, G] < d + i with i < d−k N and 2kt < d, then
k+1
Fs is divisible by h . In other words, there is a homogeneous polynomial F̃s−t(k+1) of degree
s − t(k + 1) such that Fs = F̃s−t(k+1) hk+1 .

Proof. Since deg[F, G] < d + i, we see by Theorem 4.2 that

α1
X
(44) Gi = c(i) αs α0
α F1 · · · Fs h .
α=(α0 ,...,αs )∈Z×Ns
tα0 +α1 +2α2 +···+sαs =i
tα0 +d(α1 +···+αs )≤N

(i) 1
By Lemma 5.7 used for k = 0, all summands of (44), except c( 1 (di−sN ),0,...,0,N −i) FsN −i h t (di−sN ) ,
t
1 1
involves hu with u > t (di − sN ). Let us also notice that t (di − sN ) < 0. Thus, multiplying
1 (i)
(sN −di)
both sides of (44) by h t , we obtain that h divides c( 1 (di−sN ),0,...,0,N −i) FsN −i . Since N 6≡ 0(
t
QN−i 1
(i) ( (di−sN )+rk)
mod d), we see that −sN 6≡ 0( mod d) and so c( 1 (di−sN ),0,...,0,N −i) = k=1 QtN−i aN =
t is =1 ris
QN−i
k=1 (di−sN +dk)
QN−i aN 6= 0. Thus, we see that h|FsN −i and so h|Fs because h is a square-free poly-
is =1 dis
nomial.
Now, we can (44) rewrite in the form
α Fs αs
cα(i) F1α1 · · · Fs−1
X
hα0 +αs ,
s−1


(45) Gi = h
α=(α0 ,...,αs )∈Z×Ns
tα0 +α1 +2α2 +···+sαs =i
tα0 +d(α1 +···+αs )≤N

Fs
where h is a homogeneous polynomial of degree s − t. Using Lemma 5.7 for k = 1, we see
(i)
N −i 1 [(d−1)i−(s−1)N ]
that all summands of (45), except c( 1 (di−sN ),0,...,0,N −i) Fhs ht , involves hu with
t
u > 1t [(d − 1)i − (s − 1)N ]. Let us also notice that 1t [(d − 1)i − (s − 1)N ] < 0. Thus, multiplying
1 (i)
N −i
both sides of (45) by h t [(d−1)i−(s−1)N ] , we obtain that h divides c( 1 (di−sN ),0,...,0,N −i) Fhs .
t
(i) Fs
Since c( 1 (di−sN ),0,...,0,N −i) 6= 0, it follows as above that h divides h .
t
14 Daria Holik, Marek Karaś

Now, we can (45) rewrite in the form

α Fs αs
cα(i) F1α1 · · · Fs−1
X
hα0 +2αs ,
s−1


(46) Gi = h 2

α=(α0 ,...,αs )∈Z×Ns

tα0 +α1 +2α2 +···+sαs =i
tα0 +d(α1 +···+αs )≤N

where Fh2s is a homogeneous polynomial of degree s−2t. Thus, we can repeat the above arguments
once more again.
Repeating this arguments several times we obtain the thesis. 
Corollary 5.9. Let F and G be as in Theorem 4.2. Moreover, let h be a square-free polynomial.
s−k
If N 6≡ 0 ( mod d) and deg[F, G] < d + i with i < d−k N and 2kt < d, then we have
α1 αs−1 αs
X
(47) Gj = c(j)
α F1 · · · Fs−1 F̃s−t(k+1) h
α0 +t(k+1)αs
,
α=(α0 ,...,αs )∈Z×Ns
tα0 +α1 +2α2 +···+sαs =j
tα0 +d(α1 +···+αs )≤N

(j)
where ai , . . . , aN −1 ∈ C and cα are as in Theorem 4.2 and F̃s−t(k+1) is as in Theorem 5.8.

6. Dependence between Fs−1 and Fs

In this section we consider the situation when d1 = gcd(d, N ) = gcd(deg F, deg G) is such that
d = 2d1 . In such a situation N must be an odd multiplicity of d1 and so must be of the form
N = d1 (2k + 1) for some nonnegative integer k. Since we always assume that N ≥ d, we see that
the number k maybe considered as positive integer. In this setup we will show that if deg[F, G]
is small enough that there is a relation between Fs−1 and Fs .
We start this with the following numerical lemma which we will use in both considered cases,
namely the case deg h = 1 and the case deg h > 1.
Lemma 6.1. For a given positive integers j, N, d, t such that j < N, d ≤ N and t| gcd(d, N ) let
us consider the following sets
s s
( )
X X
(48) Zj,N,d = (α0 , . . . , αs ) ∈ Z × Ns | α0 + lαl = j, α0 + d αl ≤ N ,
l=1 l=1
s s
( )
(t)
X X
s
(49) Zj,N,d = (α0 , . . . , αs ) ∈ Z × N | tα0 + lαl = j, tα0 + d αl ≤ N ,
l=1 l=1
s
( )
X
s
(50) ∆j,N,d = (α1 , . . . , αs ) ∈ R+ | (d − l)αl ≤ N − j .
l=1
Then, the following assertions hold:

The mapping ϕ : ∆j,N,d ∩ Ns ∋ (α1 , . . . , αs ) 7→ (j − sl=1 lαl , α1 , . . . , αs ) ∈ Zj,N,d is

P
(a)
a bijection.
(t)
(b) The mapping ψ : Zj,N,d ∩ (tZ × Ns ) ∋ (α̃0 , α1 , . . . , αs ) 7→ ( α̃t0 , α1 , . . . , αs ) ∈ Zj,N,d is
a bijection.
The set ∆j,N,d is the simplex with the following set of vertices: (0, . . . , 0), N s−j e1 ,
N −j
e , . . . , N2−j es−1 and (N − j)es , where e1 = (1, 0, . . . , 0), . . . , es = (0, . . . , 0, 1) is
(c) s−1 2
the standard basis of Rs . In other words vertices of ∆j,N,d others than (0, . . . , 0) are
of the form Nd−l −j
el for l = 1, . . . , s.
 Dependence of Homogeneous Components 15

For any (β1 , . . . , βs ) ∈ Ns andPfor the function (a linear function actually) fβ1 ,...,βs :
s
(d) ∆j,N,d ∋ (α1 , . . . , αs ) 7→ j − l=1 (l − βl )αl ∈ R we have fβ1,...,βs (0, . . . , 0) = j and
fβ1 ,...,βs ( Nd−l
−j
el ) = j − l−β
d−l (N − j).
l

If 2βs < d then the minimal value of f0,...,0,βs on the set ∆j,N,d equals to j − (s −
(e)
βs )(N − j) and is attained exactly at one point (0, . . . , 0, N − j).
If 2βs = d then the minimal value of f0,...,0,βs on the set ∆j,N,d equals to j − (s −
(f ) βs )(N − j) and is attained exactly on the line segment with the following endpoints
(0, . . . , 0, N2−j , 0) and (0, . . . , 0, N − j).
Proof. The assertions (a)-(d) are easy to check. To prove assertions (e) and (f) let us notice
that, since the set ∆j,N,d is a convex set and the function f0,...,0,βs is linear, it follows that we
only need to examine the values of the function on the set of vertices of ∆j,N,d . Since we are
interested in minimal value of f0,...,0,βs and N − j > 0, we see that what is interested for us is
the maximal value of the fraction l−β d−l for l = 1, . . . , s. Since we have β1 = . . . = βs−1 = 0, the
l

values of l−β 1 2 d−2 s−1

d−l for l = 1, . . . , s − 1 are the following: d−1 < d−2 < . . . < 2 = 2 . To conclude
l

the proof, we only need to notice that s−1 s−1

2 < s − βs in the case 2βs < d and 2 = s − βs in the
case 2βs = d. 
Before we go to the subsection devoted to the case deg h = 1 we give the one more lemma,
which is a consequence of the above numerical one.
Lemma 6.2. Let d1 ∈ N \ {0, 1} and set d = 2d1 , N = d1 (2k + 1) for some k ∈ N+ . Then, the
following statements hold
min{ α0 + d1 αs | (α0 , . . . , αs ) ∈ Zj,N,d } is equal to (d − d1 )j − (s − d1 )N = d1 j −
(a) (d1 − 1)N and is attained exactly at the points of the following set { (dj − sN +
dl, 0, . . . , 0, l, N − j − 2l) ∈ Z × Ns | 0 ≤ 2l ≤ N − j },
d d1
Let t be an integer such that t divides d1 and set r = t and r̃ = t (in particular
(t)
(b) r = 2r̃). Then, min{ α0 + r̃αs | (α0 , . . . , αs ) ∈ is equal to r̃j − (d1 − 1)r̃(2k +
Zj,N,d }
1) and is attained exactly at the points of the following set { (rj − sr̃(2k + 1) +
rl, 0, . . . , 0, l, N − j − 2l) ∈ Z × Ns | 0 ≤ 2l ≤ N − j }.
 s
Ps
Proof. Consider the set ∆j,N,d Ps = (α1 , . . . , αs ) ∈ R+ | l=1 (d − l)αl ≤ N − j and the func-
tion f0,...,d1 (α1 , . . . , αs ) = j − l=1 (l − δls d1 )αl , where δls is the Kronecker delta. Let us notice
that from Lemma 6.1 (a), we have Zj,N,d = ϕj (∆j,N,d ∩Ns ). Notice also that f0,...,d1 (α1 , . . . , αs ) =
α0 + d1 αs , where (α0 , α1 , . . . , αs ) = ϕj (α1 , . . . , αs ) for any (α1 , . . . , αs ) ∈ ∆j,N,d ∩ Ns . Now using
the above considerations and Lemma 6.1 (c) and (f), we obtain the assertion (a).
In order to prove (b) we apply Lemma 6.1 (b) and already proved Lemma 6.2 (a). In-
(t)
deed, by Lemma 6.1 (b) we have that (α0 , . . . , αs ) ∈ Zj,N,d if and only if (α̃0 , . . . , αs ) =
(tα0 , . . . , αs ) ∈ Zj,N,d ∩ (tZ × Ns ). On the other hand minimizing α0 + r̃αs is equivalent to
minimizing t(α0 + r̃αs ) = α̃0 + d1 αs . Thus searching the points at which α0 + r̃αs achieve min-
(t)
imal values on the set Zj,N,d is the same as searching the points at which α̃0 + d1 αs achieve
minimal values on the set Zj,N,d ∩ (tZ × Ns ) and so we can use Lemma 6.2 (a). It should be
mentioned that from Lemma 6.2 (a) we get information about the points at which the minimum
is achieved and what is its value. This completes the proof. 

To conclude this part of the section, we make the following observation.

Remark 6.3. One can notice that Lemma 6.1 (e) can be used, instead of Lemmas 5.2 and 5.7,
in order to obtain alternative proofs of Theorems 5.1 and 5.8.
16 Daria Holik, Marek Karaś

6.1. The case deg h = 1. In this subsection we consider the case deg h = 1. Thus, we have
F = F1 + · · · + Fs−1 + hd and G = G1 + · · · + GN −1 + aN hN with d = 2d1 for some d1 ∈ N \ {0, 1},
aN ∈ C \ {0} and N = d1 (2k + 1) for some positive integer k.
In the proof of the main result of this subsection we will use the following lemma.
Lemma 6.4. Assume that d = 2d1 for some d1 ∈ N \ {0, 1}, N = 2d1 k + d1 for some k ∈ N+ ,
and that deg[F, G] < d + N − 2k − 2. Assume also that Fd = hd and GN = aN hN for some
homogeneous polynomial h of degree one and aN ∈ C \ {0}. Then, for all l ∈ {0, 1, . . . , k + 1},
we have
 
(N −2k−2) k + 1 (N −2k−2)
(51) c(−2d1 k−3d1 +2d1 l,0,...,0,l,2k+2−2l) = (−4)l c(−2d1 k−3d1 ,0,...,0,0,2k+2) .
l
Proof. Since deg[F, G] < d + N − 2k − 2, we see that the polynomial GN −2k−2 is given by
the formulas (27) and (28) with d = 2d1 and s = 2d1 − 1. One can check, that
(N −2k−2)
(52) c(−2d1 k−3d1 +2d1 (l+1),0,...,0,l+1,2k+2−2(l+1))
Q2k+2−(l+1)
i=1 (−2d1 k − 3d1 + 2d1 (l + 1) + 2d1 i)
= Ql+1 Q2k+2−2(l+1) aN
is−1 =1 2d1 is−1 · is =1 2d1 is
and
Q2k+2−l
(N −2k−2) i=1 (−2d1 k − 3d1 + 2d1 l + 2d1 i)
(53) c(−2d1 k−3d1 +2d1 l,0,...,0,l,2k+2−2l) = Ql aN ,
· i2k+2−2l
Q
is−1 =1 2d1 is−1 s =1
2d1 is

for any l = 0, 1, . . . , k. From the above, one can see that

(N −2k−2)
(54) c(−2d1 k−3d1 +2d1 (l+1),0,...,0,l+1,2k+2−2(l+1))
2d1 (2k + 1 − 2l)2d1 (2k + 2 − 2l) (N −2k−2)
= c
2d1 (l + 1)(−2d1 k − d1 + 2d1 l) (−2d1 k−3d1 +2d1 l,0,...,0,l,2k+2−2l)
(k + 1 − l)(−4) (N −2k−2)
= c(−2d1 k−3d1 +2d1 l,0,...,0,l,2k+2−2l).
l+1
In particular, for l = 0, we obtain that
(N −2k−2) (k + 1)(−4) (N −2k−2)
(55) c(−2d1 k−d1 ,0,...,0,1,2k) = c(−2d1 k−3d1 ,0,...,0,0,2k+2)
1 
k + 1 (N −2k−2)
= (−4) c(−2d1 k−3d1 ,0,...,0,0,2k+2) .
1
Similarly, one can see that
(N −2k−2) (k + 1 − 1)(−4) (N −2k−2)
c(−2d1 k+d1 ,0,...,0,2,2k−2) = c(−2d1 k−d1 ,0,...,0,1,2k)
1+1
(k + 1)k(−4)2 (N −2k−2)
= c
1· 2  (−2d1 k−3d1 ,0,...,0,0,2k+2)
k + 1 (N −2k−2)
= (−4)2 c(−2d1 k−3d1 ,0,...,0,0,2k+2) .
2
Continuing the above calculation inductively, we obtain the result. 
Now we can prove the following theorem.
Theorem 6.5. Assume that d = 2d1 for some d1 ∈ N \ {0, 1}, N = 2d1 k + d1 for some
k ∈ N+ , and that deg[F, G] < d + N − 2k − 2. Assume also that Fd = hd and GN = aN hN for
 Dependence of Homogeneous Components 17

some homogeneous polynomial h of degree one and aN ∈ C \ {0}. Then, there is a homogeneous
polynomial F̂s−2 of degree s − 2 such that
(56) Fs−1 = 14 (F̃d21 −1 + hF̂s−2 ).
s−(d1 −1) d1 d1 −1 s−(d1 −1) d1 d1 −1
Proof. First of all notice that d−(d1 −1) = d1 +1 > d1 and so d−(d1 −1) N = d1 +1 N > d1 N =
d1 −1
d1 (2d1 k + d1 )= N − 2k − 1. By the above and the assumption that deg[F, G] < d + N − 2k − 2,
we can use Theorem 5.5 to obtain that Fs is divisible by hd1 . Thus, there exists a homogeneous
polynomial F̃d1 −1 of degree d1 − 1 such that Fs = hd1 F̃d1 −1 . Now, using the above considerations,
assumption that deg[F, G] < d + N − 2k − 2 and Theorem 4.1 with d = 2d1 , s = 2d1 − 1 and
j = N − 2k − 2 we obtian that
αs−1 αs
cα(N −2k−2) F1α1 · · · Fs−1
X
(57) GN −2k−2 = F̃d1 −1 hα0 +d1 αs .
α=(α0 ,...,αs )∈Z×Ns
α0 +α1 +2α2 +···+sαs =j
α0 +d(α1 +···+αs )≤N

Notice that, for N = 2d1 k + d1 and j = N − 2k − 2, the set from Lemma 6.2 (a) is equal to
the following one { (−2d1 k − 3d1 + 2d1 l, 0, . . . , 0, l, 2k + 2 − 2l) ∈ Z × Ns | l = 0, 1, . . . , k + 1 }.
Thus, in the polynomial GN −2k−2 all summand other than included in the following sum
k+1
(N −2k−2)
X
(58) l
c(−2d1 k−3d1 +2d1 l,0,...,0,l,2k+2−2l) Fs−1 F̃d2k+2−2l
1 −1
h−d1
l=0
involves hu with u > −d1 . By Lemma 6.4, the above sum can be written as follows
k+1  
(N −2k−2)
X k+1
(59) c(−2d1 k−3d1 ,0,...,0,2k+2) h−d1
(−4)l Fs−1
l
F̃d2k+2−2l
1 −1
l
l=0
(N −2k−2)
= c(−2d1 k−3d1 ,0,...,0,2k+2) h (F̃d21 −1
−d1
− 4Fs−1 )k+1 .
(N −2k−2)
Now, we see by multiplying both sides of (57) by hd1 that h divides c(−2d1 k−3d1 ,0,...,0,2k+2) (F̃d21 −1
Q2k+2
(N −2k−2) (−2d1 k−3d1 +2d1 i)
−4Fs−1 )k+1 . But, since c(−2d1 k−3d1 ,0,...,0,2k+2) = i=1
Q2k+2 a2d1 k+d1 6= 0, it follows that
is =1 2d1 is

h|(F̃d21 −1 − 4Fs−1 )k+1 , and since h is not a power of any other polynomial (in this case this means
that h is square-free), h|(F̃d21 −1 − 4Fs−1 ). Thus there is a homogeneous polynomial F̂s−2 of degree
s − 2 such that −hF̂s−2 = F̃d21 −1 − 4Fs−1 . Solving the last equality with respect to Fs−1 we obtain
the thesis. 
6.2. The case deg h > 1. As the title of subsection suggests, we consider here the situation of
d N
deg h > 1. Thus, we have F = F1 + · · · + Fs−1 + h t and G = G1 + · · · + GN −1 + aN h t where
t = deg h and aN ∈ C \ {0}. As before, we assume that d = 2d1 for some d1 ∈ N \ {0, 1} and
N = d1 (2k + 1) for some positive integer k. Since t must divide d and N and gcd(d, N ) = d1 ,
one can see that we must have, in this situation, that d1 is divisible by t.
In the proof of the main result of this subsection we will use the following lemma.
d
Lemma 6.6. Let d, N and deg[F, G] be as in Lemma 6.4. Moreover, suppose that Fd = h t
N
and GN = aN h t for some square-free, homogeneous polynomial h of degree t, where t ≥ 1,
aN ∈ C \ {0} and r = dt , r̃ = dt1 . Then, for all l ∈ {0, 1, . . . , k + 1}, we have
 
(N −2k−2) l k+1 (N −2k−2)
(60) c((−2k−3+2l)r̃,0,...,0,l,2k+2−2l) = (−4) c((−2k−3)r̃,0,...,0,0,2k+2) .
l
18 Daria Holik, Marek Karaś

Proof. According with the assumption that deg[F, G] < d + N − 2k − 2, the polynomial
GN −2k−2 is given by the formulas (34) and (35) with d = 2d1 , s = 2d1 − 1, so
(N −2k−2)
(61) c((−2k−3+2(l+1))r̃,0,...,0,l+1,2k+2−2(l+1))
(N −2k−2)
= c −2d1 k−3d1 +2d1 (l+1) 
t
,0,...,0,l+1,2k+2−2(l+1)
Q2k+2−(l+1) −2d1 k−3d1 +2d1 (l+1)+2d1 i
i=1 t
= Ql+1 2d1 is−1 Q2k+2−2(l+1) 2d1 is N
a
is−1 =1 t · is =1 t
Q2k+2−(l+1)
i=1 (−2d1 k − 3d1 + 2d1 (l + 1) + 2d1 i)
= Ql+1 Q2k+2−2(l+1) aN
is−1 =1 2d1 is−1 · is =1 2d1 is

and
Q2k+2−l
(N −2k−2) i=1 (−2d1 k − 3d1 + 2d1 l + 2d1 i)
(62) c((−2k−3+2l)r̃,0,...,0,l,2k+2−2l) = Ql aN ,
· i2k+2−2l
Q
is−1 =1 2d1 is−1 s =1 2d1 is

for any l = 0, 1, . . . , k. Further part of this proof is similar to the proof of Lemma 6.4. 
Theorem 6.7. Assume that d = 2d1 for some d1 ∈ N \ {0, 1}, N = 2d1 k + d1 for some k ∈ N+ ,
d N
and that deg[F, G] < d + N − 2k − 2. Assume also that Fd = h t = hr and GN = aN h t for some
square-free, homogeneous polynomial h of degree t, where t ≥ 1, aN ∈ C \ {0} and r = dt , r̃ = dt1 .
Then, there is a homogeneous polynomial F̂s−t−1 of degree s − t − 1 such that
(63) Fs−1 = 14 (F̃d21 −1 + hF̂s−t−1 ).
Proof. Using the assumption that deg[F, G] < d + N − 2k − 2, and the following observation
d1 d1
s−( −1) d1 (2t−1) d1 −1 s−( −1) d1 −1
that d
t
= d1 (2t−1)+t > d1 and so d
t
N > d1 N = N − 2k − 1 we can apply
d−( t1 −1) d−( t1 −1)
Theorem 5.8 to obtain the divisibility of Fs by hr̃ . It means that, there exists a homogeneous
polynomial F̃d1 −1 of degree d1 − 1 such that Fs = hr̃ F̃d1 −1 . Referring to the above calculations
and Theorem 4.2, the polynomial GN −2k−2 takes the form
αs−1 αs
cα(N −2k−2) F1α1 · · · Fs−1
X
(64) GN −2k−2 = F̃d1 −1 hα0 +r̃αs .
α=(α0 ,...,αs )∈Z×Ns
tα0 +α1 +2α2 +···+sαs =j
tα0 +d(α1 +···+αs )≤N

According to Lemma 6.2 (b) for j = N − 2k − 2, the set of summands with minimal possible
power of h, in the above sum, corresponds to α belonging to { ((−2k − 3 + 2l)r̃, 0, . . . , 0, l, 2k +
2 − 2l) ∈ Z × Ns | l = 0, 1, . . . , k + 1 }, so all components under than occurring the presenting sum
k+1
(N −2k−2)
X
(65) l
c((−2k−3+2l)r̃,0,...,0,l,2k+2−2l)Fs−1 F̃d2k+2−2l
1 −1
h−r̃
l=0

involves hu with u > −r̃. Applying Lemma 6.6, we can rewrite the above sum in the following
way
k+1  
(N −2k−2)
X k+1
(66) c((−2k−3)r̃,0,...,0,2k+2) h−r̃ (−4)l Fs−1
l
F̃d2k+2−2l
1 −1
l
l=0
(N −2k−2)
= c((−2k−3)r̃,0,...,0,2k+2) h−r̃ (F̃d21 −1 − 4Fs−1 )k+1 .
 Dependence of Homogeneous Components 19

The rest of the proof is analogous to the proof of the Theorem 6.5 using the assumption that
h is square-free. 

References
[1] V. Drensky, J.-T. Yu, Degree estimate for commutators. J. Algebra 322, no. 7, 2321–2334 (2009)
[2] E. Edo, T. Kanehira, M. Karaś, S. Kuroda, Separability of wild automorphisms of a polynomial ring. Trans-
form. Groups 18, no. 1, 81–96 (2013)
[3] P. Jędrzejewicz, J. Zieliński, Analogs of Jacobian conditions for subrings. J. Pure Appl. Algebra 221, 2111-
2118 (2017)
[4] M. Karaś, There is no tame automorphism of C3 with multidegree (3, 4, 5). Proc. Am. Math. Soc. 139, no.
3, 769-775 (2011)
[5] M. Karaś, Tame automorphisms of C3 with multidegree of the form (3, d2 , d3 ). J. Pure Appl. Algebra 214,
no. 12, 2144-2147 (2010)
[6] M. Karaś, Multidegrees of tame automorphisms of Cn . Diss. Math. 477, 55 p. (2011)
[7] M. Karaś, J. Zygadło, On multidegree of tame and wild automorphisms of C3 . J. Pure Appl. Algebra 215,
2843–2846 (2011)
[8] S. Kuroda, A generalization of the Shestakov-Umirbaev inequality. J. Math. Soc. Japan 60, no. 2, 495–510
(2008)
[9] S. Kuroda, Shestakov-Umirbaev reductions and Nagata’s conjecture on a polynomial automorphisms. Tohoku
Math. J. 62, 75-115 (2010)
[10] S. Kuroda, On the Karaś type theorems for the multidegrees of polynomial automorphisms. J. Algebra 423,
441–465 (2015)
[11] L. Makar-Limanov, J.-T. Yu, Degree estimate for subalgebras generated by two elements. J. Eur. Math. Soc.
(JEMS) 10, 533–541 (2008)
[12] F. L. Pritchard, Polynomial mappings with Jacobian determinant of bounded degree. Arch. Math. 48, no. 6,
495–504 (1987)
[13] I. P. Shestakov, U. U. Umirbaev, The Nagata automorphism is wild. Proc. Natl. Acad. Sci. USA 100, 12561-
12563 (2003)
[14] I. P. Shestakov, U. U. Umirbaev, Poisson brackets and two-generated subalgebras of rings of polynomials. J.
Amer. Math. Soc. 17, 181-196 (2004)
[15] I. P. Shestakov, U. U. Umirbaev, The tame and the wild automorphisms of polynomial rings in three variables.
J. Amer. Math. Soc. 17, 197-227 (2004)
[16] X. Sun, Y. Chen, Multidegrees of tame automorphisms in dimension three. Publ. Res. Inst. Math. Sci. 48,
no. 1, 129–137 (2012)
[17] U. U. Umirbaev, J.-T.Yu, The strong Nagata conjecture. Proc. Natl. Acad. Sci. USA 101, 4352-4355 (2004)
[18] J.-T. Yu, Degree estimate for subalgebras and automorphisms of free algebras. In: S.-T. Yau, et al. (Eds.),
Proceedings of the Fourth International Congress of Chinese Mathematicians, Hangzhou, China, December
17–23, (2007), Higher Education Press, Beijing (2008), pp. 359–366; International Press, Boston (2009)

Daria Holik
AGH University of Science and Technology,
Faculty of Applied Mathematics
al. A. Mickiewicza 30
30-059 Kraków
Poland
e-mail: holikd@agh.edu.pl
ORCID: https://orcid.org/0000-0002-2133-9711

Marek Karaś
AGH University of Science and Technology,
Faculty of Applied Mathematics
al. A. Mickiewicza 30
20 Daria Holik, Marek Karaś

30-059 Kraków
Poland
e-mail: mkaras@agh.edu.pl
ORCID: https://orcid.org/0000-0003-0821-521X