Abstract

Let 𝑅 be a prime ring of char 𝑅2, 𝑑 a nonzero derivation of 𝑅 and 𝜌 a nonzero right ideal of 𝑅 such that [[𝑑(𝑥),𝑥]𝑛,[𝑦,𝑑(𝑦)]𝑚]𝑡=0 for all 𝑥,𝑦𝜌, where 𝑛0, 𝑚0, 𝑡1 are fixed integers. If [𝜌,𝜌]𝜌0, then 𝑑(𝜌)𝜌=0.

1. Introduction

Throughout this paper, unless specifically stated, 𝑅 always denotes a prime ring with center 𝑍(𝑅) and extended centroid 𝐶, 𝑄 the Martindale quotients ring. Let 𝑛 be a positive integer. For given 𝑎,𝑏𝑅, let [𝑎,𝑏]0=𝑎 and let [𝑎,𝑏]1 be the usual commutator 𝑎𝑏𝑏𝑎, and inductively for 𝑛>1, [𝑎,𝑏]𝑛=[[𝑎,𝑏]𝑛1,𝑏]. By 𝑑 we mean a nonzero derivation in 𝑅.

A well-known result proven by Posner [1] states that if [[𝑑(𝑥),𝑥],𝑦]=0 for all 𝑥,𝑦𝑅, then 𝑅 is commutative. In [2], Lanski generalized this result of Posner to the Lie ideal. Lanski proved that if 𝑈 is a noncommutative Lie ideal of 𝑅 such that [[𝑑(𝑥),𝑥],𝑦]=0 for all 𝑥𝑈,𝑦𝑅, then either 𝑅 is commutative or char 𝑅=2 and 𝑅 satisfies 𝑆4, the standard identity in four variables. Bell and Martindale III [3] studied this identity for a semiprime ring 𝑅. They proved that if 𝑅 is a semiprime ring and [[𝑑(𝑥),𝑥],𝑦]=0 for all 𝑥 in a non-zero left ideal of 𝑅 and 𝑦𝑅, then 𝑅 contains a non-zero central ideal. Clearly, this result says that if 𝑅 is a prime ring, then 𝑅 must be commutative.

Several authors have studied this kind of Engel type identities with derivation in different ways. In [4], Herstein proved that if char 𝑅2 and [𝑑(𝑥),𝑑(𝑦)]=0 for all 𝑥,𝑦𝑅, then 𝑅 is commutative. In [5], Filippis showed that if 𝑅 is of characteristic different from 2 and 𝜌 a non-zero right ideal of 𝑅 such that [𝜌,𝜌]𝜌0 and [[𝑑(𝑥),𝑥],[𝑑(𝑦),𝑦]]=0 for all 𝑥,𝑦𝜌, then 𝑑(𝜌)𝜌=0.

In continuation of these previous results, it is natural to consider the situation when [[𝑑(𝑥),𝑥]𝑛,[𝑦,𝑑(𝑦)]𝑚]𝑡=0 for all 𝑥,𝑦𝜌, 𝑛,𝑚0,𝑡1 are fixed integers. We have studied this identity in the present paper.

It is well known that any derivation of a prime ring 𝑅 can be uniquely extended to a derivation of 𝑄, and so any derivation of 𝑅 can be defined on the whole of 𝑄. Moreover 𝑄 is a prime ring as well as 𝑅 and the extended centroid 𝐶 of 𝑅 coincides with the center of 𝑄. We refer to [6, 7] for more details.

Denote by 𝑄𝐶𝐶{𝑋,𝑌} the free product of the 𝐶-algebra 𝑄 and 𝐶{𝑋,𝑌}, the free 𝐶-algebra in noncommuting indeterminates 𝑋,𝑌.

2. The Case: 𝑅 Prime Ring

We need the following lemma.

Lemma 2.1. Let 𝜌 be a non-zero right ideal of 𝑅 and 𝑑 a derivation of 𝑅. Then the following conditions are equivalent: (i) d is an inner derivation induced by some 𝑏𝑄 such that 𝑏𝜌=0; (ii) 𝑑(𝜌)𝜌=0 (for its proof refer to [8, Lemma]).

We mention an important result which will be used quite frequently as follows.

Theorem 2.2 (see Kharchenko [9]). Let 𝑅 be a prime ring, 𝑑 a derivation on 𝑅 and 𝐼 a non-zero ideal of 𝑅. If 𝐼 satisfies the differential identity 𝑓(𝑟1,𝑟2,,𝑟𝑛,𝑑(𝑟1),𝑑(𝑟2),,𝑑(𝑟𝑛))=0forany𝑟1,𝑟2,,𝑟𝑛𝐼, then either (i) 𝐼 satisfies the generalized polynomial identity 𝑓𝑟1,𝑟2,,𝑟𝑛,𝑥1,𝑥2,,𝑥𝑛=0,(2.1) or (ii) 𝑑 is 𝑄-inner, that is, for some 𝑞𝑄,𝑑(𝑥)=[𝑞,𝑥] and 𝐼 satisfies the generalized polynomial identity 𝑓𝑟1,𝑟2,,𝑟𝑛,𝑞,𝑟1,𝑞,𝑟2,,𝑞,𝑟𝑛=0.(2.2)

Theorem 2.3. Let 𝑅 be a prime ring of char 𝑅2 and 𝑑 a derivation of 𝑅 such that [[𝑑(𝑥),𝑥]𝑛,[[𝑦,𝑑(𝑦)]𝑚]𝑡=0 for all 𝑥,𝑦𝑅, where 𝑛0,𝑚0,𝑡1 are fixed integers. Then 𝑅 is commutative or 𝑑=0.

Proof. Let 𝑅 be noncommutative. If 𝑑 is not 𝑄-inner, then by Kharchenko's Theorem [9] 𝑔[](𝑥,𝑦,𝑢,𝑣)=𝑢,𝑥𝑛,[]𝑦,𝑣𝑚𝑡=0,(2.3) for all 𝑥,𝑦,𝑢,𝑣𝑅. This is a polynomial identity and hence there exists a field 𝐹 such that 𝑅𝑀𝑘(𝐹) with 𝑘>1, and 𝑅 and 𝑀𝑘(𝐹) satisfy the same polynomial identity [10,Lemma 1]. But by choosing 𝑢=𝑒12,𝑥=𝑒11,𝑣=𝑒11 and 𝑦=𝑒21, we get []0=𝑢,𝑥𝑛,[]𝑦,𝑣𝑚𝑡=(1)𝑡𝑛𝑒11+()𝑡𝑒22,(2.4) which is a contradiction.
Now, let 𝑑 be 𝑄-inner derivation, say 𝑑=𝑎𝑑(𝑎) for some 𝑎𝑄, that is, 𝑑(𝑥)=[𝑎,𝑥] for all 𝑥𝑅, then we have []𝑎,𝑥𝑛+1,[[𝑦,𝑎,𝑦]]𝑚𝑡=0,(2.5) for all 𝑥,𝑦𝑅. Since 𝑑0, 𝑎𝐶 and hence 𝑅 satisfies a nontrivial generalized polynomial identity (GPI). By [11], it follows that 𝑅𝐶 is a primitive ring with 𝐻=𝑆𝑜𝑐(𝑅𝐶)0, and 𝑒𝐻𝑒 is finite dimensional over 𝐶 for any minimal idempotent 𝑒𝑅𝐶. Moreover we may assume that 𝐻 is noncommutative; otherwise, 𝑅 must be commutative which is a contradiction.
Notice that 𝐻 satisfies [[𝑎,𝑥]𝑛+1,[𝑦,[𝑎,𝑦]]𝑚]𝑡=0 (see [10, Proof of Theorem 1]). For any idempotent 𝑒𝐻 and 𝑥𝐻, we have []0=𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡.(2.6) Right multiplying by 𝑒, we get []0=𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡𝑒=[]𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡1[]𝑎,𝑒𝑛+1[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚[[𝑒𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚[]𝑎,𝑒𝑛+1𝑒=[]𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡1[]𝑎,𝑒𝑛+1𝑚𝑗=0(1)𝑗𝑚𝑗[]𝑎,𝑒𝑥(1𝑒)𝑗[]𝑒𝑥(1𝑒)𝑎,𝑒𝑥(1𝑒)𝑚𝑗𝑒𝑚𝑗=0(1)𝑗𝑚𝑗[]𝑎,𝑒𝑥(1𝑒)𝑗[]𝑒𝑥(1𝑒)𝑎,𝑒𝑥(1𝑒)𝑚𝑗[]𝑎,𝑒𝑛+1𝑒=[]𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡10𝑚𝑗=0(1)𝑗𝑚𝑗(𝑒𝑥(1𝑒)𝑎)𝑗𝑒𝑥(1𝑒)(𝑎𝑒𝑥(1𝑒))𝑚𝑗[]𝑎𝑒=𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡1𝑚𝑗=0𝑚𝑗(𝑒𝑥(1𝑒)𝑎)𝑚+1𝑒=2𝑚[]𝑎,𝑒𝑛+1,[[𝑒𝑥(1𝑒),𝑎,𝑒𝑥(1𝑒)]]𝑚𝑡1(𝑒𝑥(1𝑒)𝑎)𝑚+1𝑒=()𝑡2𝑚𝑡(𝑒𝑥(1𝑒)𝑎)(𝑚+1)𝑡𝑒.(2.7)
This implies that 0=()𝑡2𝑚𝑡((1𝑒)𝑎𝑒𝑥)(𝑚+1)𝑡+1. Since char 𝑅2, ((1𝑒)𝑎𝑒𝑥)(𝑚+1)𝑡+1=0. By Levitzki's lemma [12, Lemma 1.1], (1𝑒)𝑎𝑒𝑥=0 for all 𝑥𝐻. Since 𝐻 is prime ring, (1𝑒)𝑎𝑒=0, that is, 𝑒𝑎𝑒=𝑎𝑒 for any idempotent 𝑒𝐻. Now replacing 𝑒 with 1𝑒, we get that 𝑒𝑎(1𝑒)=0, that is, 𝑒𝑎𝑒=𝑒𝑎. Therefore for any idempotent 𝑒𝐻, we have [𝑎,𝑒]=0. So 𝑎 commutes with all idempotents in 𝐻. Since 𝐻 is a simple ring, either 𝐻 is generated by its idempotents or 𝐻 does not contain any nontrivial idempotents. The first case gives 𝑎𝐶 contradicting 𝑑0. In the last case, 𝐻 is a finite dimensional division algebra over 𝐶. This implies that 𝐻=𝑅𝐶=𝑄 and 𝑎𝐻. By [10,Lemma 2], there exists a field 𝐹 such that 𝐻𝑀𝑘(𝐹) and 𝑀𝑘(𝐹) satisfies [[𝑎,𝑥]𝑛+1,[𝑦,[𝑎,𝑦]]𝑚]𝑡. Then by the same argument as earlier, 𝑎 commutes with all idempotents in 𝑀𝑘(𝐹), again giving the contradiction 𝑎𝐶, that is, 𝑑=0. This completes the proof of the theorem.

Theorem 2.4. Let 𝑅 be a prime ring of char 𝑅2, 𝑑 a non-zero derivation of 𝑅 and 𝜌 a non-zero right ideal of 𝑅 such that [[𝑑(𝑥),𝑥]𝑛,[𝑦,𝑑(𝑦)]𝑚]𝑡=0 for all 𝑥,𝑦𝜌, where 𝑛0,𝑚0,𝑡1 are fixed integers. If [𝜌,𝜌]𝜌0, then 𝑑(𝜌)𝜌=0.

We begin the proof by proving the following lemma.

Lemma 2.5. If 𝑑(𝜌)𝜌0 and [[𝑑(𝑥),𝑥]𝑛,[𝑦,𝑑(𝑦)]𝑚]𝑡=0 for all 𝑥,𝑦𝜌,𝑚,𝑛0,𝑡1 are fixed integers, then 𝑅 satisfies nontrivial generalized polynomial identity (GPI).

Proof. Suppose on the contrary that 𝑅 does not satisfy any nontrivial GPI. We may assume that 𝑅 is noncommutative; otherwise, 𝑅 satisfies trivially a nontrivial GPI. We consider two cases.Case 1. Suppose that 𝑑 is 𝑄-inner derivation induced by an element 𝑎𝑄. Then for any 𝑥𝜌,[]𝑎,𝑥𝑋𝑛+1,[[𝑥𝑌,𝑎,𝑥𝑌]]𝑚𝑡(2.8) is a GPI for 𝑅, so it is the zero element in 𝑄𝐶𝐶{𝑋,𝑌}. Expanding this, we get []𝑎,𝑥𝑋𝑚𝑛+1𝑗=0(1)𝑗𝑚𝑗[]𝑎,𝑥𝑌𝑗[]𝑥𝑌𝑎,𝑥𝑌𝑚𝑗𝑚𝑗=0(1)𝑗𝑚𝑗[]𝑎,𝑥𝑌𝑗[]𝑥𝑌𝑎,𝑥𝑌𝑚𝑗[]𝑎,𝑥𝑋𝑛+1𝐴(𝑋,𝑌)=0,(2.9) where 𝐴(𝑋,𝑌)=[[𝑎,𝑥𝑋]𝑛+1,[𝑥𝑌,[𝑎,𝑥𝑌]]𝑚]𝑡1. If 𝑎𝑥 and 𝑥 are linearly 𝐶-independent for some 𝑥𝜌, then (𝑎𝑥𝑋)𝑚𝑛+1𝑗=0(1)𝑗𝑚𝑗[]𝑎,𝑥𝑌𝑗[]𝑥𝑌𝑎,𝑥𝑌𝑚𝑗𝑚𝑗=0(1)𝑗𝑚𝑗(𝑎𝑥𝑌)𝑗[]𝑥𝑌𝑎,𝑥𝑌𝑚𝑗[]𝑎,𝑥𝑋𝑛+1𝐴(𝑋,𝑌)=0.(2.10) Again, since 𝑎𝑥 and 𝑥 are linearly 𝐶-independent, above relation implies that []𝑥𝑌𝑎,𝑥𝑌𝑚[]𝑎,𝑥𝑋𝑛+1𝐴(𝑋,𝑌)=0,(2.11) and so 𝑥𝑌(𝑎𝑥𝑌)𝑚(𝑎𝑥𝑋)𝑛+1𝐴(𝑋,𝑌)=0.(2.12) Repeating the same process yields 𝑥𝑌(𝑎𝑥𝑌)𝑚(𝑎𝑥𝑋)𝑛+1𝑡=0(2.13) in 𝑄𝐶𝐶{𝑋,𝑌}. This implies that 𝑎𝑥=0, a contradiction. Thus for any 𝑥𝜌, 𝑎𝑥 and 𝑥 are 𝐶-dependent. Then (𝑎𝛼)𝜌=0 for some 𝛼𝐶. Replacing 𝑎 with 𝑎𝛼, we may assume that 𝑎𝜌=0. Then by Lemma 2.1, 𝑑(𝜌)𝜌=0, contradiction.
Case 2. Suppose that 𝑑 is not 𝑄-inner derivation. If for all 𝑥𝜌, 𝑑(𝑥)𝑥𝐶, then [𝑑(𝑥),𝑥]=0 which implies that 𝑅 is commutative (see [13]). Therefore there exists 𝑥𝜌 such that 𝑑(𝑥)𝑥𝐶, that is, 𝑥 and 𝑑(𝑥) are linearly 𝐶-independent.
By our assumption, we have that 𝑅 satisfies
[𝑑](𝑥𝑋),𝑥𝑋𝑛,[]𝑥𝑌,𝑑(𝑥𝑌)𝑚𝑡=0.(2.14) By Kharchenko's Theorem [9], 𝑑(𝑥)𝑋+𝑥𝑟1,𝑥𝑋𝑛,𝑥𝑌,𝑑(𝑥)𝑌+𝑥𝑟2𝑚𝑡=0,(2.15) for all 𝑋,𝑌,𝑟1,𝑟2𝑅. In particular for 𝑟1=𝑟2=0, [𝑑](𝑥)𝑋,𝑥𝑋𝑛,[]𝑥𝑌,𝑑(𝑥)𝑌𝑚𝑡=0,(2.16) which is a nontrivial GPI for 𝑅, because 𝑥 and 𝑑(𝑥) are linearly 𝐶-independent, a contradiction.

We are now ready to prove our main theorem.

Proof of Theorem 2.4. Suppose that 𝑑(𝜌)𝜌0, then we derive a contradiction. By Lemma 2.5, 𝑅 is a prime GPI ring, so is also 𝑄 by [14]. Since 𝑄 is centrally closed over 𝐶, it follows from [11] that 𝑄 is a primitive ring with 𝐻=𝑆𝑜𝑐(𝑄)0.
By our assumption and by [7], we may assume that [𝑑](𝑥),𝑥𝑛,[]𝑦,𝑑(𝑦)𝑚𝑡=0(2.17) is satisfied by 𝜌𝑄 and hence by 𝜌𝐻. Let 𝑒=𝑒2𝜌𝐻 and 𝑦𝐻. Then replacing 𝑥 with 𝑒 and 𝑦 with 𝑒𝑦(1𝑒) in (2.17), then right multiplying it by 𝑒, we obtain that [𝑑]0=(𝑒),𝑒𝑛,[]𝑒𝑦(1𝑒),𝑑(𝑒𝑦(1𝑒))𝑚𝑡𝑒=[𝑑](𝑒),𝑒𝑛,[]𝑒𝑦(1𝑒),𝑑(𝑒𝑦(1𝑒))𝑚𝑡1[]𝑑(𝑒),𝑒𝑛𝑚𝑗=0(1)𝑗𝑚𝑗𝑑(𝑒𝑦(1𝑒))𝑗𝑒𝑦(1𝑒)𝑑(𝑒𝑦(1𝑒))𝑚𝑗𝑒𝑚𝑗=0(1)𝑗𝑚𝑗𝑑(𝑒𝑦(1𝑒))𝑗𝑒𝑦(1𝑒)𝑑(𝑒𝑦(1𝑒))𝑚𝑗[]𝑑(𝑒),𝑒𝑛𝑒.(2.18)
Now we have the fact that for any idempotent 𝑒, 𝑑(𝑦(1𝑒))𝑒=𝑦(1𝑒)𝑑(𝑒), 𝑒𝑑(𝑒)𝑒=0 and so []0=𝑑(𝑒),𝑒𝑛,[]𝑒𝑦(1𝑒),𝑑(𝑒𝑦(1𝑒))𝑚𝑡10𝑚𝑗=0(1)𝑗𝑚𝑗𝑒(𝑦(1𝑒)𝑑(𝑒))𝑗𝑦(1𝑒)𝑑(𝑒𝑦(1𝑒))𝑚𝑗.𝑑(𝑒)𝑒(2.19) Now since for any idempotent 𝑒 and for any 𝑦𝑅, (1𝑒)𝑑(𝑒𝑦)=(1𝑒)𝑑(𝑒)𝑦, above relation gives []0=𝑑(𝑒),𝑒𝑛,[]𝑒𝑦(1𝑒),𝑑(𝑒𝑦(1𝑒))𝑚𝑡1𝑒𝑚𝑗=0𝑚𝑗(𝑦(1𝑒)𝑑(𝑒))𝑗𝑦(1𝑒)(𝑑(𝑒)𝑦(1𝑒))𝑚𝑗=[]𝑑(𝑒)𝑒𝑑(𝑒),𝑒𝑛,[]𝑒𝑦(1𝑒),𝑑(𝑒𝑦(1𝑒))𝑚𝑡1𝑒𝑚𝑗=0𝑚𝑗(𝑦(1𝑒)𝑑(𝑒))𝑚+1𝑒=[]𝑑(𝑒),𝑒𝑛,[]𝑒𝑦(1𝑒),𝑑(𝑒𝑦(1𝑒))𝑚𝑡12𝑚𝑒(𝑦(1𝑒)𝑑(𝑒))𝑚+1𝑒=2𝑚𝑒(𝑦(1𝑒)𝑑(𝑒))𝑚+1𝑡𝑒.(2.20) This implies that 0=(1)𝑡2𝑚𝑡((1𝑒)𝑑(𝑒)𝑒𝑦)(𝑚+1)𝑡+1 for all 𝑦𝐻. Since char 𝑅2, we have by Levitzki's lemma [12,Lemma 1.1] that (1𝑒)𝑑(𝑒)𝑒𝑦=0 for all 𝑦𝐻. By primeness of 𝐻, (1𝑒)𝑑(𝑒)𝑒=0. By [15,Lemma 1], since 𝐻 is a regular ring, for each 𝑟𝜌𝐻, there exists an idempotent 𝑒𝜌𝐻 such that 𝑟=𝑒𝑟 and 𝑒𝑟𝐻. Hence (1𝑒)𝑑(𝑒)𝑒=0 gives (1𝑒)𝑑(𝑒)=(1𝑒)𝑑(𝑒2)=(1𝑒)𝑑(𝑒)𝑒=0 and so 𝑑(𝑒)=𝑒𝑑(𝑒)𝑒𝐻𝜌𝐻 and 𝑑(𝑟)=𝑑(𝑒𝑟)=𝑑(𝑒)𝑒𝑟+𝑒𝑑(𝑒𝑟)𝜌𝐻. Hence for each 𝑟𝜌𝐻, 𝑑(𝑟)𝜌𝐻. Thus 𝑑(𝜌𝐻)𝜌𝐻. Set 𝐽=𝜌𝐻. Then 𝐽=𝐽/(𝐽𝑙𝐻(𝐽)), a prime 𝐶-algebra with the derivation 𝑑 such that 𝑑(𝑥)=𝑑(𝑥), for all 𝑥𝐽. By assumption, we have that 𝑑𝑥,𝑥𝑛,𝑦,𝑑𝑦𝑚𝑡=0,(2.21) for all 𝑥,𝑦𝐽. By Theorem 2.3, we have either 𝑑=0 or 𝜌𝐻 is commutative. Therefore we have that either 𝑑(𝜌𝐻)𝜌𝐻=0 or [𝜌𝐻,𝜌𝐻]𝜌𝐻=0. Now 𝑑(𝜌𝐻)𝜌𝐻=0 implies that 0=𝑑(𝜌𝜌𝐻)𝜌𝐻=𝑑(𝜌)𝜌𝐻𝜌𝐻 and so 𝑑(𝜌)𝜌=0. [𝜌𝐻,𝜌𝐻]𝜌𝐻=0 implies that 0=[𝜌𝜌𝐻,𝜌𝐻]𝜌𝐻=[𝜌,𝜌𝐻]𝜌𝐻𝜌𝐻 and so [𝜌,𝜌𝐻]𝜌=0, then 0=[𝜌,𝜌𝜌𝐻]𝜌=[𝜌,𝜌]𝜌𝐻𝜌 implying that [𝜌,𝜌]𝜌=0. Thus in all the cases we have contradiction. This completes the proof of the theorem.

3. The Case: 𝑅 Semiprime Ring

In this section we extend Theorem 2.3 to the semiprime case. Let 𝑅 be a semiprime ring and 𝑈 be its right Utumi quotient ring. It is well known that any derivation of a semiprime ring 𝑅 can be uniquely extended to a derivation of its right Utumi quotient ring 𝑈 and so any derivation of 𝑅 can be defined on the whole of 𝑈 [7,Lemma 2].

By the standard theory of orthogonal completions for semiprime rings, we have the following lemma.

Lemma 3.1 (see [16, Lemma 1 and Theorem 1] or [7,pages 31-32]). Let 𝑅 be a 2-torsion free semiprime ring and 𝑃 a maximal ideal of 𝐶. Then 𝑃𝑈 is a prime ideal of 𝑈 invariant under all derivations of 𝑈. Moreover, {𝑃𝑈𝑃isamaximalidealof𝐶with𝑈/𝑃𝑈2-torsionfree}=0.

Theorem 3.2. Let 𝑅 be a 2-torsion free semiprime ring and 𝑑 a non-zero derivation of 𝑅 such that [[𝑑(𝑥),𝑥]𝑛,[𝑦,𝑑(𝑦)]𝑚]𝑡=0 for all 𝑥,𝑦𝑅, 𝑛,𝑚0,𝑡1 fixed are integers. Then 𝑑 maps 𝑅 into its center.

Proof. Since any derivation 𝑑 can be uniquely extended to a derivation in 𝑈, and 𝑅 and 𝑈 satisfy the same differential identities [7, Theorem 3], we have [𝑑](𝑥),𝑥𝑛,[]𝑦,𝑑(𝑦)𝑚𝑡=0,(3.1) for all 𝑥,𝑦𝑈. Let 𝑃 be any maximal ideal of 𝐶 such that 𝑈/𝑃𝑈 is 2-torsion free. Then by Lemma 3.1, 𝑃𝑈 is a prime ideal of 𝑈 invariant under 𝑑. Set 𝑈=𝑈/𝑃𝑈. Then derivation 𝑑 canonically induces a derivation 𝑑 on 𝑈 defined by 𝑑(𝑥)=𝑑(𝑥) for all 𝑥𝑈. Therefore, 𝑑𝑥,𝑥𝑛,𝑦,𝑑𝑦𝑚𝑡=0,(3.2) for all 𝑥,𝑦𝑈. By Theorem 2.3, either 𝑑=0 or [𝑈,𝑈]=0, that is, 𝑑(𝑈)𝑃𝑈 or [𝑈,𝑈]𝑃𝑈. In any case 𝑑(𝑈)[𝑈,𝑈]𝑃𝑈 for any maximal ideal 𝑃 of 𝐶. By Lemma 3.1, {𝑃𝑈𝑃isamaximalidealof𝐶with𝑈/𝑃𝑈2-torsionfree}=0. Thus 𝑑(𝑈)[𝑈,𝑈]=0. Without loss of generality, we have 𝑑(𝑅)[𝑅,𝑅]=0. This implies that 𝑅0=𝑑2[][][][].𝑅,𝑅=𝑑(𝑅)𝑅𝑅,𝑅+𝑅𝑑(𝑅)𝑅,𝑅=𝑑(𝑅)𝑅𝑅,𝑅(3.3) Therefore [𝑅,𝑑(𝑅)]𝑅[𝑅,𝑑(𝑅)]=0. By semiprimeness of 𝑅, we have [𝑅,𝑑(𝑅)]=0, that is, 𝑑(𝑅)𝑍(𝑅). This completes the proof of the theorem.