Вы находитесь на странице: 1из 4

Jacobsons Commutativity Theorem

(Last modied on May 2, 2005.)


Problem 36 (page 30, Chapter 1) in Musilis Rings and Modules [5] is a special case of a very
beautiful theorem about commutativity of certain rings due to Jacobson [4]. This write-up is just
a compilation of the solutions to some special cases along with a proof of the general case due to
Herstein [3]. Section 1 proves two special cases, Section 2 proves the theorem for division rings,
and Section 3 proves it for any ring.
1 n = 2 and n = 3 cases
Lemma 1.1 If x
2
= x for all x R, then R is commutative.
Proof: (x + x)
2
= x + x = 4x = 2x = 2x = 0, i.e., x = x. Now x, y R we have
(x + y)
2
= x + y, which implies that xy + yx = 0, i.e., xy = yx = yx, so R is commutative. 2
Proposition 1.2 If x
3
= x for all x R, then R is commutative.
Proof: (x + x)
3
= x + x = 8x = 2x = 6x = 0. Also (x
2
x)
3
= x
2
x =
x
6
3x
5
+ 3x
4
x
3
= x
2
3x + 3x
2
x = x
2
x = 3x
2
= 3x.
Now (3x)
2
= 9x
2
= 6x
2
+ 3x
2
= 3x
2
= 3x. And since 3R is a subring of R, we can use lemma
1.1 to show that 3R is commutative. So (3x)(3y) = (3y)(3x) = 9xy = 9yx = 6xy + 3xy =
6yx + 3yx = 3xy = 3yx, for all x, y R.
Finally, we will use (x + y)
3
= x + y = x + x
2
y + xyx + xy
2
+ yx
2
+ yxy + y
2
x + y =
x + y = x
2
y + xyx + xy
2
+ yx
2
+ yxy + y
2
x = 0. Also (x y)
3
= x y = x x
2
y xyx +
xy
2
yx
2
+ yxy + y
2
x y = x y = x
2
y xyx + xy
2
yx
2
+ yxy + y
2
x = 0. Combining
these two we get 2(x
2
y + xyx + yx
2
) = 0 and 2(y
2
x + yxy + xy
2
) = 0. Now 2(x
2
y + xyx +
yx
2
) = 0 = 2x(x
2
y + xyx + yx
2
) = 2(xy + x
2
yx + xyx
2
) = 0. Also 2(x
2
y + xyx + yx
2
) =
0 = 2(x
2
y + xyx + yx
2
)x = 2(x
2
yx + xyx
2
+ yx) = 0. Now subtracting these two we have,
2(xy +x
2
yx +xyx
2
) 2(x
2
yx +xyx
2
+yx) = 2xy 2yx = 0, i.e., 2xy = 2yx. This nally gives us
xy = 3xy 2xy = 3yx 2yx = yx for all x, y R, so R is commutative. 2
2 Division Rings
In this section we will generalize the statement for division rings. Let D be the division ring that
we are working with. Consider the polynomial ring D[X]. We call g(X) a left divisor of f(X) if
f(X) = g(X)q(X), for some q(X) D[X]. The left division algorithm also remains well-dened
in D[X]. A polynomial f(X) is called a central polynomial if f(X) Z[X], where Z is the center
of D. Notice that f(X) is a central polynomial i zf(X) = f(X)z for all z D. For central
polynomials the notions of left and right divisors coincide.
Lemma 2.1 Let h(X) D[X], h(X) / D and let g(X) = 0 be such that h(X) is a left divisor of
zg(X)z
1
for all 0 = z D. Then there exists a central polynomial f(X) / D which is a divisor
of g(X).
1
Proof: First of all, observe that the set of all polynomials which satisfy the same property
as g(X) form a right ideal I in D[X]. Let f(X) be the monic polynomial of the lowest degree
in this ideal. Then I is precisely the set of all polynomials that are right multiples of f(X).
Now f
z
(X) = f(t) zf(X)z
1
is also monic and lies in I, and moreover, deg(f
z
) < deg(f). So
f(X) = zf(X)z
1
for all z D. Therefore, f(X) must be a central polynomial, and since h(X) is
a left divisor of f(X), f(X) / D. 2
Lemma 2.2 If X a, a D, is a left divisor of f(X)g(X) but not a left divisor of f(X), then
for some 0 = z D, X a is a left divisor of zg(X)z
1
.
Proof: Let f(X) = (Xa)q(X)+r, where 0 = r D. Now f(X)g(X)r
1
= (Xa)q(X)g(X)r
1
+
rg(X)r
1
= rg(X)r
1
= f(X)g(X)r
1
+ (X a)q(X)g(X)r
1
. So X a is a left divisor of
rg(X)r
1
. 2
Proposition 2.3 f(X) be a central polynomial irreducible over Z. If a, b D are both roots of
f(X) then there exists 0 = x D such that b = xax
1
.
Proof: Let f(X) = q(X)(X a). If for every 0 = z D, zq(X)z
1
were left divisible by
(X b) then by Lemma 2.1, q(X) would be divisible by a central polynomial, which would imply
that f(X) is not irreducible over Z. So for some 0 = z D, (X b) is not a left divisor of
zq(X)z
1
. But f(X) = zf(X)z
1
= zq(X)z
1
(X zaz
1
), and using Lemma 2.2, there exists
0 = w D we have X b is a left divisor of w(X zaz
1
)w
1
= X (wz)a(wz)
1
. But that
means X b = X (wz)a(wz)
1
, i.e., b = (wz)a(wz)
1
. 2
Now with these tools we are ready to prove the theorem for division rings. We can think of this
as a generalization of the well-known fact that every nite division ring is a eld.
Theorem 2.4 Let D be a division ring such that for every x D there exists 1 < n(x) N such
that x
n(x)
= x. Then D is commutative, or in other words, D is a eld.
Proof: Take 0 = a D, we have a
n(a)
= a and (2a)
n(2a)
= 2a. Let m = (n(a) 1)(n(2a) 1) +1.
Then a
m
= a and (2a)
m
= 2a = (2
m
2)a = 0 = 2
m
2 = (2
m
2)aa
1
= 0 =
(2
m
2)z = 0 for all z D. So D has characteristic equal to some prime p.
Let F
p

= P D be the prime eld of D. Suppose that for some x, b D, xbx
1
= b
k
. Dene
D
x,b
=
_
_
_
n(b)2

i=0
n(x)2

j=0
p
ij
b
i
x
j
, p
ij
P
_
_
_
Clearly, D
x,b
is a nite set and is closed under multiplication (by using the rule xb = b
k
x). And
because the inverse of any element z D is a power of z itself (namely, z
n(z)2
), D
x,b
is in fact a
nite division ring, and hence it is a eld. So xb = bx and b
k
= b.
Now suppose b D but b / Z. The polynomials in D[X] satised by b form a right ideal. As we
know, any right ideal in D[X] is formed by the right multiples of a monic central polynomial, so let
f(X) Z[X] be the minimal polynomial of b over D[X]. f(X) = X
m
+
m1
X
m1
+. . .+
0
, where

i
Z. Since each
i
satises the polynomial X
n(
i
)1
1 = 0, we get that P(
0
,
1
, . . . ,
m1
) is a
nite extension eld of P and has p
t
elements, for some t. So every element z P(
0
,
1
, . . . ,
m1
)
2
satises z
p
t
= z. So b
p
t
= b, because otherwise there would be p
t
+1 solutions to X
p
t
X = 0 over
in P(
0
,
1
,
m1
, b). But since f(b) = b
m
+
m1
b
m1
+. . . +
0
= 0 and D has characteristic p, we
have f(b
p
t
) = f(b)
p
t
= 0. Therefore, b and b
p
t
are both roots of the same minimal polynomial over
Z. Using Proposition 2.3, we must have b
p
t
= xbx
1
. But as we have proved above, this implies
that xb = bx and b
p
t
= b, a contradiction. Hence, Z = D which means that D is commutative. 2
3 Jacobsons Theorem
Now we want to extend Theorem 2.4 to prove the statement over any ring.
Denition 3.1 Jacobson radical J(R) is the intersection of all maximal left ideals of R.
J(R) =

max. left ideals I

It is easy to check that J(R) is a two-sided ideal.


Denition 3.2 A ring R is called semisimple if J(R) = (0).
Denition 3.3 An element x R is called a unit if there exists y R such that xy = yx = 1.
Lemma 3.4 If x J(R) then 1 + x is a unit.
Proof: If x J(R) then x I

, for every maximal left ideal I

of R. So 1+x / I

for any maximal


left ideal, and therefore it does not lie in any proper left ideal of R. Hence R(1 +x) = R and there
exists a left inverse y R such that y(1 + x) = 1. Now we know that if x J(R) then 1 +x has a
left inverse, and we want to show that a right inverse also exists. Since y = 1yx and yx J(R),
so y has a left inverse z such that zy = 1. But y(1 + x) = 1 = zy(1 + x) = z = (1 + x) = z.
So (1 + x)y = y(1 + x) = 1, which means that 1 + x is a unit. 2
We will assume the following famous theorem. Its proof can be found in [1].
Theorem 3.5 (Wedderburns Structure Theorem [1]) Every semisimple ring R is isomorphic
to direct sum of matrix rings over division rings, i.e., R

=

k
i=1
M
c
i
(D
i
) where every M
c
i
(D
i
) is
the ring of c
i
c
i
matrices over division ring D
i
.
Theorem 3.6 Let R be a ring with a unit element (i.e., 1 R) such that for every x R there
exists 1 < n(x) N, depending on x, such that x
n(x)
= x. Then R is commutative.
Proof: Firstly, it suces to look only at semisimple rings R. Because if R is not semisimple,
its Jacobson radical J(R) = (0). Consider 0 = x J(R). There exists 1 < n(x) N such that
x
n(x)
= x, i.e., x(1 x
n(x)1
) = 0. But x J(R) = x
n(x)1
J(R). So using lemma 3.4
we get that 1 x
n(x)1
is a unit and there is y R such that (1 x
n(x)1
)y = 1. Therefore
x(1 x
n(x)1
)y = 0 = x = 0, a contradiction.
Now by Wedderburns structure theorem for semisimple rings, R

k
i=1
M
c
i
(D
i
) where every
M
c
i
(D
i
) is the ring of c
i
c
i
matrices over division ring D
i
. And for every x M
c
i
(D
i
), there exists
1 < n(x) N such that x
n(x)
= x. If some c
i
> 1 then consider the following element in M
c
i
(D
i
).
3
x =
_
_
_
_
_
0 1 0
0 0 0
.
.
.
0 0 0 0
_
_
_
_
_
x
2
= 0 which gives a contradiction because there cannot be any 1 < n(x) N such that
x
n(x)
= x. So c
i
= 1 for all i, and R

k
i=1
D
i
, where each D
i
is commutative (by Theorem 2.4).
Hence R must also be commutative. 2
We can extend Theorem 3.6 for rings which do not contain a unit element 1.
Theorem 3.7 Let R be any ring such that for every x R there exists 1 < n(x) N with
x
n(x)
= x, then R is commutative.
Proof: Take x, y R and let n(xy yx) = m. So (xy yx)
m
= (xy yx) = ((xy yx)
m1
)
2
=
(xy yx)
m1
, i.e., (xy yx)
m1
is an idempotent element in R.
Now consider any idempotent e R. For any x R, we have (xe exe)
2
= 0 (by expanding
and using e
2
= e). And if (xe exe) = 0, it cannot be the case that (xe exe)
s
= (xe exe) = 0
for some 1 < s = n(xe exe). So xe exe = 0. Similarly, we can show that ex exe = 0. Now by
subtracting one of them from the other we have xe = ex which means that every idempotent e lies
in the center Z of the ring. Consider R

= eR = Re. R

is a ring with the same property as and


with e as an identity element. So using Theorem 3.6, R

is commutative. Hence for all x, y R,


we get xye = (xe)(ye) = (ye)(xe) = yxe. Therefore, (xy yx)e = 0 for every idempotent e.
Since (xy yx)
m1
is an idempotent, (xy yx)(xy yx)
m1
= (xy yx)
m
= (xy yx) = 0.
So xy = yx and R is commutative. 2
References
[1] B. Farb, R. K. Dennis, Noncommutative Algebra, Springer-Verlag, 1993.
[2] I. N. Herstein, Noncommutative rings, The Carus Mathematical Monograph No. 15, MAA,
1968.
[3] I. N. Herstein, An elementary proof of a theorem of Jacobson, Duke Mathematical Journal,
21 (1954), 45-48.
[4] N. Jacobson, Structure theory for algebraic algebras of bounded degree, Annals of Math, 46
(1945), 695-707.
[5] C. Musili, Rings and Modules, Narosa Publishing House.
4

Вам также может понравиться