Rev. Mat. Iberoam.

(online first) c European Mathematical Society


doi 10.4171/rmi/1165

On differential polynomial rings over nil algebras

Mikhail Chebotar and Wen-Fong Ke

Abstract. Let R be a nil algebra over a field of characteristic 0, and let δ

be a derivation of R. Then the differential polynomial ring R[X, δ] cannot
be mapped onto a unital simple ring homomorphically.

1. Introduction
Let δ : R → R be a derivation of a ring R. By R[X; δ] we denote the differential
polynomial ring, that is the set of left polynomials a0 + a1 X + · · · + an−1 X n−1 +
an X n , where ai ∈ R, i = 0, . . . , n. Addition of such polynomials is defined in
the usual way. Multiplication is defined by the associative rule, except that the
right polynomials should be converted into left polynomials, which can be done by
repeatedly applying the rule Xa = aX + aδ for all a ∈ R. Here, aδ denotes the
image of a by δ.
Recently, there has been a significant interest to the radical properties of differ-
ential polynomial rings [3], [4], [8], [9], [13], [14]. It is quite interesting to compare
polynomial rings over nil rings with the differential polynomial rings over nil rings.
A famous result due to Smoktunowicz [12] asserts that there exists a nil ring R
such that the polynomial ring R[x] is not nil. Obviously, this result easily extends
to differential polynomial rings.
Recall that the Koethe conjecture, one of the most challenging open questions
in ring theory, has an equivalent statement: whether a polynomial ring R[X] is
Jacobson radical if the ring R is nil. Along the long stream of attempts to solve
this problem, Puczylowski and Smoktunowicz [11] laid a cornerstone by showing
that a polynomial ring R[X] over a nil ring R is Brown–McCoy radical, i.e., it
cannot be mapped homomorphically onto a unital simple ring.
Now, it is natural to verify similar statements for differential polynomial rings
over nil rings. First of all, in their important paper [14], Smoktunowicz and Ziem-
bowski showed that there exists a locally nilpotent ring R and a derivation δ
of R such that R[X; δ] is not Jacobson radical, thus solving an open problem by

Mathematics Subject Classification (2010): 16N40.

Keywords: Differential polynomial ring, nil ring.
2 M. Chebotar and W.-F. Ke

Shestakov. Recently, Smoktunowicz ([9], Proposition 3.1) proved that if R is lo-

cally nilpotent, then R[X; δ] is Brown–McCoy radical. Hence, R[X; δ] cannot be
mapped onto a unital simple ring homomorphically. Still, one has to ponder on
the following question.
Question A. Let δ : R → R be a derivation of a nil ring R. Can R[X; δ] be
mapped homomorphically onto a unital simple ring?
In this short paper we will show that the answer is negative if the target simple
ring is of characteristic 0. Namely, we will prove the following theorem.
Theorem 1. Let R be a nil ring and let δ be a derivation of R. If ϕ is a homo-
morphism from the differential polynomial ring R[X, δ] to a unital simple ring S,
where S is of characteristic 0, then ϕ cannot be surjective.
As an obvious corollary, we state:
Corollary 2. Let R be a nil algebra over a field of characteristic 0, and let δ be
a derivation of R. Then the differential polynomial ring R[X, δ] cannot be mapped
homomorphically onto a unital simple ring.
We remark that, to answer Question A in full generality, it is sufficient to con-
sider nil algebras over Zp . From this perspective, it is interesting to compare this
question with the one about polynomial rings in several variables over nil rings
posted by Puczylowski and Smoktunowicz. In [11], Question 1a, it was asked
whether polynomial rings in several variables over nil rings are Brown–McCoy rad-
ical. For this problem, it was sufficient to consider nil algebras instead of nil rings.
For polynomial rings in two variables over nil algebras of prime characteristic, the
answer was obtained in [5]; for general situation, the answer was given only very
recently in [6] by means of convex geometry.

2. Results
One of the main tools in our proof is Kharchenko’s theory of differential identities.
For a general introduction to Kharchenko’s technique as well as the properties of
symmetric ring of quotients and extended centroid, one is referred to the book [2].
We will include a short and nice presentation from [7] on differential polynomial
identities before we put down the Kharchenko’s theorem, namely Theorem 4, that
we need for our proof.
Let us start with the following useful result on nil prime rings.
Lemma 3. Let R be a nil prime ring with symmetric ring of quotients Q and let
q ∈ Q be an element such that [q, r] = qr − rq ∈ R for all r ∈ R. Then the ring S
generated by R and q cannot be a simple ring with 1.

Proof. Suppose that the ring S generated by R and q is a simple ring with 1.
We claim that q is invertible in S. As 1 ∈ S, there are a0, a1 , . . . , ak ∈ R such
On differential polynomial rings over nil algebras 3

that 1 = a0 + a1 q + · · · + ak q k . Write a0 = 1 − a1q − · · · − ak q k , which is nilpotent,

and so (1 − a1 q − · · · − ak q k ) = 0 for some  ≥ 1. Expanding, we see that

1 = b1 q + · · · + bt q t = (b1 + · · · + bt q t−1 )q

for some t ∈ N and b1 , . . . , bt ∈ R. Since b1 + · · · + bt q t−1 ∈ S, q has a left inverse

in S. Similarly, changing the side on which the element q appears, we see that q
is has a right inverse in S. Hence q is invertible in S.
Now, for some positive integer N and ui,0 , vj,0 ∈ R (i, j ∈ {0, 1, . . . , N }), we
can write
N 
N
q −1 = ui,0 q i = q j vj,0 .
i=0 j=0

Let k ≥ 1, and assume that


N 
N
q −k = ui,k q i = q j vj,k ,
i=0 j=0

where ui,k , vj,k ∈ R (i, j ∈ {0, 1, . . . , N }). Then


N  
N 
q −(k+1) = ui,k q i · q −1 = u0,k · q −1 + ui,k q i · q −1
i=0 i=1

N  
N 
N
= u0,k ui,0 q i + ui,k q i−1 = u0,k u0,0 + (u0,k ui,0 + ui,k+1 ) q i
i=0 i=1 i=1

N
= ui,k+1 q i ,
i=0

and similarly,


N
q −(k+1) = q j vj,k+1 ,
j=0

where ui,k+1 , vj,k+1 ∈ R (i, j ∈ {0, 1, . . . , N }). Thus, by induction, we have for
any positive integer t, there are ui,t , vj,t ∈ R i, j ∈ {0, 1, . . . , N }, such that


N 
N
q −t = ui,t q i = q j vj,t .
i=0 j=0

Take a nonzero ideal I of R with the property that q i I ⊆ R and Iq j ⊆ R for

i, j ∈ {1, . . . N }. Such an ideal exists by Proposition 2.2.3 in [2]. Let J = I 2 . Then
we have

(2.1) q −s J q −t ⊆ R for all positive integers s and t.

4 M. Chebotar and W.-F. Ke

We note that the idealgenerated by J and q is S due to the simplicity of S. Hence,

the unity 1 is a sum (i,j)∈Λ q i wij q j , where Λ is a finite subset of N0 × N0 , and
wij ∈ J for all (i, j) ∈ Λ. Let K and L be positive integers such that for every
(i, j) ∈ Λ, K > i and L > j. With (2.1), for each (i, j) ∈ Λ, there is an rij ∈ R
with
q i wij q j = q K (q i−K wij q j−L )q L = q K rij q L .
Thus,
    
1= q i wij q j = q K rij q L = q K rij q L = q K r q L ,
(i,j)∈Λ (i,j)∈Λ (i,j)∈Λ


where r = (i,j)∈Λ rij ∈ R. But then, r = q −K−L , which is impossible since r is
nilpotent. Therefore, S cannot be a simple ring with 1. 2

Let R be a prime ring with symmetric ring of quotients Q and extended cen-
troid C, which is a field. Let Der(R) and Der(Q) denote the set of all derivations
of R and Q, respectively. Since every derivation of R can be uniquely extended
to a derivation of Q (cf. [2], Proposition 2.5.1), we regard Der(R) ⊆ Der(Q). For
α ∈ C, σ ∈ Der(Q) and x ∈ Q, define xσα = xσ α. This makes σα a derivation
on Q, and turns Der(Q) into a right vector space over C. Let D = Der(R)C and
let Dint be a subspace of Der(Q) consisting of all inner derivations of Q.
Now, a differential polynomial is a generalized polynomial involving noncommu-
tative indeterminates Yi which are acted by derivations of R as unary operations.
The coefficients of a differential polynomial are allowed to lie in Q. Every differen-
Δ
tial polynomial can be transformed and written in the form χ(Yi j ), where χ(Zij )
is an ordinary generalized polynomial in Zij and Δj ’s are words of derivations
Δ
of R. We say that χ = 0 is a differential identity for R if χ(ri j ) = 0 for all ri ∈ R.
Choose a basis M0 for Dint , and augment it to a basis M of Dint + D. Fix
a total order < in the set M such that μ < δ for δ ∈ M0 and μ ∈ M \ M0 ,
and extend it to the set of all derivation words in M : a shorter word is smaller
than a longer one, while words of the same length are ordered lexicographically. A
Δ
differential identity can always be transformed into one in the form χ(Yi j ) = 0,
where (1) χ(Zij ) is a generalized polynomial in distinct indeterminates Zij , and
(2) each Δj is a regular word. Here a word Δ = δ1s1 δ2s2 . . . δm
sm
, is regular if
1. δj ∈ M \ M0 for all j = 1, . . . , m;
2. δ1 < δ2 < · · · < δm ; and
3. if R has characteristic p > 0, then sk < p for all k = 1, . . . , m.
We now state Kharchenko’s theorem as it was done in [7], p. 254.
Δ
Theorem 4. If χ(Yi j ) = 0 is a differential identity on a prime ring R, where Δj
are distinct regular derivation words in M and where χ(Zij ) is a generalized poly-
nomial in indeterminates Zij , then χ(Zij ) = 0 is a generalized polynomial identity
on R.
On differential polynomial rings over nil algebras 5

Applying the theorem, we obtain a result on linear generalized differential poly-

nomials, which will be used for Lemma 6. It also illustrates well how Kharchenko’s
theorem works.
In our case, we will have just one derivation d on R, so the linear
 i generalized
i
differential polynomial in our consideration is of the form i=0 sj=0 aij Y d bij ,
s 
aij , bij ∈ Q, where Y is an indeterminate. When j=0 aj Y d bj = 0, we say that
the leading exponent is d . If there is some q ∈ Q such that rd = [q, r] for all
r ∈ R, we say that d is X-inner.
Corollary 5. Let R be a prime ring of characteristic 0 and let d be a derivation
of R which is not X-inner. Let

i 
 
si
i
χ(Y d ) = aij Y d bij
i=0 j=0

be a generalized differential polynomial, where aij , bij ∈ Q and Y is an indetermi-

i
nate. Then there exists r ∈ R such that χ(rd ) = 0.
Proof. Since d is not X-inner, we may assume that d ∈ M \M0 and all di are regular
i  
words. Let d be the leading exponent of χ(Y d ), that is sj=0 aj Y d bj = 0.
i i
Suppose that χ(rd ) = 0 for all r ∈ R, that is χ(Y d ) is a differential identity
on R. According to Theorem 4 we get


 
si
χ(Zi ) = aij Zi bij = 0
i=0 j=0

for all Zi ∈ R, i = 0, . . . , . Taking Z0 = · · · = Z−1 = 0, we get


s
aj Z bj = 0
j=0

for all Z ∈ R, but this contradicts Corollary 6.1.3 in [2], which asserts that there
are no nonzero linear generalized polynomial identities in one variable. 2
Lemma 6. Let S be a simple ring of characteristic 0. Let R be a prime ring with
extended centroid C, and x ∈ S an element such that [x, r] = xr − rx ∈ R for all
r ∈ R. Assume that S is generated by R and x, and that the derivation d on R given
by rd = [x, r] for all r ∈ R is not X-inner. Then for elements a0 , a1 , . . . , an ∈ RC,
a0 + a1 x + · · · + an xn = 0 only if ai = 0 for all i.
Proof. We proceed by induction on n. The case n = 0 is obviously true. Assume
that the statement holds for n = k > 0, and that a0 , a1 , . . . , ak+1 ∈ RC are such
that a0 + a1 x + · · · + ak+1 xk+1 = 0. For convenience, write

s = a0 + a1 x + · · · + ak+1 xk+1 = 0.

If ak+1 = 0, there is nothing to prove. So we assume ak+1 = 0.

6 M. Chebotar and W.-F. Ke

By Proposition 2.5.1 in [2], we can extend the derivation d uniquely to a deriva-

tion on RC, and by the assumption, we have [x, b] = bd for all b ∈ RC. Hence,

(2.2) xb = b x + bd for all b ∈ RC.

For r ∈ R, set t(r) = sr ak+1 − ak+1 rs = 0. After using (2.2) to rearrange the
terms of t(r) in the powers of x, we get

t(r) = b0 (r) + b1 (r)x + · · · + bk (r)xk + bk+1 (r)xk+1 = 0,

where each bm (Y ), 0 ≤ m ≤ k + 1, is a linear generalized differential polynomial

in Y . In particular,

bk+1 (r) = ak+1 r ak+1 − ak+1 r ak+1 = 0,

while k+1
b0 (r) = ak+1 rd ak+1 + b(r),
with b(Y ) a linear generalized differential polynomial in Y whose leading exponent
is smaller than dk+1 . Therefore, t(r) is in the form

t(r) = b0 (r) + b1 (r)x + · · · + bk (r)xk = 0,

and the induction hypothesis says that bm (r) = 0 for m = 0, . . . , k. Moreover, this
is true for arbitrary r ∈ R. However, as ak+1 is not zero, b0 (Y ) is a nonzero linear
generalized differential polynomial in Y , and so, according to Corollary 5, there
exists an element r ∈ R such that b0 (r ) is not zero. This is a contradiction, and
the lemma is proved. 2
Proof of Theorem 1. Suppose that the skew polynomial ring R[X, δ] is homomor-
phically mapped onto a unital simple ring S of characteristic 0. Let ϕ be such an
epimorphism, and A the image of R. Let p(X) ∈ R[X, δ] be a preimage of 1 and
let x be the image of Xp(X) under this homomorphism.
Suppose that r ∈ R and ϕ(r) = a. Then

ϕ(rX) = ϕ(rX) ϕ(p(X)) = ϕ(rX p(X)) = ϕ(r) ϕ(Xp(X)) = ax,

and for i > 1,

ϕ(r(X p(X))i ) = ϕ(rX p(X)) ϕ(X p(X))i−1 = ax · xi−1 = axi .

k k k
Consequently, if h = i=0 ri X i , then ϕ(h) = i=0 ϕ(ri X i ) = i ϕ(ri )xi . There-
fore, every element u ∈ S can be written in the form u = a0 + a1 x + · · · + ak xk for
some k ≥ 0 and each ai ∈ A for i = 0, . . . , k.
Since R is nil, we have that A is nil as well. Further, with S being of zero
characteristic, we will show that A is prime. Our approach is based on the idea of
the proof of Theorem 4 in [10] with appropriate modifications.
First, we observe that if I is a nonzero differential ideal of A, then I cannot not
be locally nilpotent. In particular, A itself is not locally nilpotent. Indeed, for a
On differential polynomial rings over nil algebras 7

nonzero differential ideal I of A, consider the differential polynomial ring I[Y ; d],
where ad = Y a − aY for all a ∈ I. Take the homomorphism ψ from I[Y ; d] into S
defined by a0 + a1 Y + · · · + ak Y k → a0 + a1 x + · · · + ak xk for all a0 , a1 , . . . , ak ∈ I.
Then ψ(I[Y ; d]) is a nonzero ideal of S, and so ψ(I[Y ; d]) = S since S is simple.
By Proposition 3.1 in [9], this cannot happen if I is locally nilpotent.
Next, we claim that A is primary in the sense of [10], p. 338, i.e., every ideal
divisor of zero of A is nilpotent. Let I be a nonzero ideal of A such that Iβ = 0
for some nonzero β ∈ A. Assume that I is not nilpotent, and set

J = {γ ∈ A | ∃M (γ) > 0 such that I M(γ) γ = 0}.

Then J is a nonzero ideal of A. For any γ ∈J, as I M(γ) γ =0, we have I 2M(γ) γ d =0,
and so γ d ∈ J as well. Therefore, J is a nonzero differential ideal of A. We will
derive a contradiction that J is locally nilpotent.
Let Γ = {γ1 , . . . , γl } be a finite subset of J and, for any k > 0, denote Γk =
{γi1 γi2 · · · γik | 1 ≤ ij ≤ l}. It is sufficient to show that Γk = 0 for some k. Let

K = {α ∈ A | ∃N (α) > 0 such that αAΓN (α) = 0}.

Then K is an ideal of A. From I M(γi1 ) Aγi1 γi2 · · · γik ⊆ I M(γi1 ) γi1 γi2 · · · γik = 0,
we have I M(γi1 ) ⊆ K. Since I is assumed to be not nilpontent, K is nonzero. Now,
as shown in the proof of Theorem 4 in [10], for α ∈ K, it holds that

αd AΓ2N (α) = 0,

thus d(α) ∈ K. This shows that K is a differential ideal of A. Therefore S is

generated by K and x, and we have

(2.3) α0 + xα1 + · · · + xt αt = 1

for some α0 , . . . , αt ∈ K. Let N = max{N (α0 ), . . . , N (αt )}. Then αi ΓN +1 ⊆

αi AΓN = 0 for all i, 0 ≤ i ≤ t. Multiplying ΓN +1 from the right on the iden-
tity (2.3), we get 0 = ΓN +1 . Therefore, J is locally nilpotent. This contradiction
assures us that I is nilpotent. Hence, A is primary.
Let us continue to show that A is prime. This will be the case if we could show
that A has no nonzero nilpotent ideal. Let T be the sum of all nilpotent ideals,
and we set to show that T is a differential ideal and conclude that T = 0. To this
goal, it is sufficient to show that if an element α ∈ A is contained in a nilpotent
ideal, then αd is sitting is some nilpotent ideal as well.
As in the proof of Theorem 4 in [10], we assume that (A, +) is torsion-free,
and let α ∈ A be an element of some nilpotent ideal. Let us show that αd belongs
to some nilpotent ideal as well. If αd belongs to an ideal divisor of zero, then it
is in a nilpotent ideal and there is nothing to do. So we assume that αd is not
in any ideal divisor of zero. As some power of αA is zero, there is a smallest
integer m such that (αA)m = 0. If m = 1, then α = 0 and αd = 0, and we are
done. Assume m > 1, and let ai , bi ∈ A, 1 ≤ i ≤ m − 1. For all a ∈ A, we have
8 M. Chebotar and W.-F. Ke

αa1 αa2 . . . αam−1 αa = 0. Differentiating this equation m times and multiplying it

from the left with αb1 αb2 . . . αbm−1 , we get

(m + 1)αb1 αb2 . . . αbm−1 αd a1 αd a2 . . . αd am−1 αd a = 0 for all a ∈ A.

Since A is torsion-free,

αb1 αb2 . . . αbm−1 αd a1 αd a2 . . . αd am−1 αd a = 0 for all a ∈ A.

As A is primary, we have

αb1 αb2 . . . αbm−1 αd a1 αd a2 . . . αd am−1 αd = 0.

The elements ai , bj ∈ A, 1 ≤ i ≤ m − 1, 1 ≤ j ≤ m − 1, are arbitrary, and since αd
is not in any ideal divisor of zero, we have

αb1 αb2 . . . αbm−1 αd a1 αd a2 . . . αd am−1 = 0

for all ai , bj ∈ A, 1 ≤ i ≤ m − 1, 1 ≤ j ≤ m − 1. Continuing in this fashion, we

shall reach the conclusion that

αb1 αb2 . . . αbm−1 αd = 0

for all bj ∈ A, 1 ≤ j ≤ m − 1, and so

αb1 αb2 . . . αbm−1 = 0

for all bj ∈ A, 1 ≤ j ≤ m − 1. This contradicts the minimality of m. Thus, αd is

contained in some nilpotent ideal of A as required. Therefore, as said, A has no
nonzero nilpotent ideals, and so A is prime.
To finish the proof of the theorem, we consider two cases.
If the derivation d is X-inner, i.e., x is contained in the symmetric ring of
quotients Q of A, then we are done by Lemma 3. Hence, we assume that d is not
X-inner. Suppose that for some a0 , a1 , . . . , an ∈ A,

a0 + a1 x + · · · + an xn = 1.

Then
(a0 − 1) + a1 x + · · · + an xn = 0,
and it follows from Lemma 6 that a0 − 1 = a1 = · · · = an = 0. This yields a0 = 1
which is impossible since a0 is nilpotent. The proof is complete. 2

After we have proved the main result, we ask the following natural question.
Question B. Let δ : R → R be a derivation of a nil ring R. Can R[X; δ] be
mapped homomorphically onto a ring with a nonzero idempotent?
On differential polynomial rings over nil algebras 9

We remark that the questions A and B are “differential analogies” of well-known

solved problems. After showing that R[x] cannot be mapped homomorphically onto
a simple ring with unity if R is nil, Puczylowski and Smoktunowicz asked if R[x]
can be mapped homomorphically onto a simple ring with a nonzero idempotent
(see Question 1b in [11]). The answer is “no”, as shown by Beidar, Fong and
Puczylowski in [1].
To conclude the paper, we note that in case of finite characteristic, a differen-
tially simple ring may not be prime, or semiprime. This prevents us from applying
Kharchenko’s theory directly as we have done above.
Acknowledgement. We would like to thank the referee for her/his useful sug-
gestions.

References
[1] Beidar, K. I., Fong, Y. and Puczylowski, E. R.: Polynomial rings over nil rings
cannot be homomorphically mapped onto rings with nonzero idempotents. J. Algebra
238 (2001), no. 1, 389–399.
[2] Beidar, K. I., Mikhalev, A. V. and Martindale, W. S.: Rings with generalized
identities. Monographs and Textbooks in Pure and Applied Mathematics 196, Marcel
Dekker, New York, 1996.
[3] Bell, J. P. Madill, B. M. and Shinko, F.: Differential polynomial rings over
rings satisfying a polynomial identity. J. Algebra 423 (2015), 28–36.
[4] Chebotar, M.: On differential polynomial rings over locally nilpotent rings. Israel
J. Math. 227 (2018), no. 1, 233–238.
[5] Chebotar, M., Ke, W.-F., Lee, P.-H. and Puczylowski, E. R.: A note on
polynomial rings over nil rings. In Modules and comodules, 169–172. Trends Math.,
Birkhä user Verlag, Basel, 2008.
[6] Chebotar, M., Ke, W.-F., Lee, P.-H. and Puczylowski, E. R.: On polynomial
rings over nil rings in several variables and the central closure of prime nil rings.
Israel J. Math. 223 (2018), no. 1, 309–322.
[7] Chuang, C.-L.: ∗-differential identities of prime rings with involution. Trans. Amer.
Math. Soc. 316 (1989), no. 1, 251–279.
[8] Greenfeld, B., Smoktunowicz, A. and Ziembowski, M.: Five solved problems
on radicals of Ore extensions. Publ. Mat. 63 (2019), no. 2, 423–444.
[9] Nielsen, P. P. and Ziembowski, M.: Derivations and bounded nilpotence index.
Internat. J. Algebra Comput. 25 (2015), no. 3, 433–438.
[10] Posner, E. C.: Differentiably simple rings. Proc. Amer. Math. Soc. 11 (1960),
337–343.
[11] Puczylowski, E. R. and Smoktunowicz, A.: On maximal ideals and the Brown–
McCoy radical of polynomial rings. Comm. Algebra 26 (1998), no. 8, 2473–2482.
[12] Smoktunowicz, A.: Polynomial rings over nil rings need not be nil. J. Algebra 233
(2000), no. 2, 427–436.
[13] Smoktunowicz, A.: How far can we go with Amitsur’s theorem in differential
polynomial rings. Israel J. Math. 219 (2017), no. 2, 555–608.
10 M. Chebotar and W.-F. Ke

[14] Smoktunowicz, A. and Ziembowski, M.: Differential polynomial rings over lo-
cally nilpotent rings need not be Jacobson radical. J. Algebra 412 (2014), 207–217.

Received July 11, 2018; revised October 25, 2018. Published online January 7, 2020.

Mikhail Chebotar: Department of Mathematical Sciences, Kent State University,

Kent, USA.
E-mail: [email protected]

Wen-Fong Ke: Department of Mathematics and Research Center for Theoretical

Sciences, National Cheng Kung University, Tainan, Taiwan.
E-mail: [email protected]