1. Normal subgroups of GLn F for a field F . Generators for En A.

Let F be a field, n ≥ 1 an integer, GLn F the group of all invertible n × n matrices

with entries in F. It is well-known, that an n × n matrix α is invertible if and only if its
determinant det(α) is invertible, i.e., det(α) ∈ GL1 F = F \ {0}.
Our main goal here is to answer the following question: Which subgroups of GLn F
are normal?
First of all, the kernel SLn F of det: GLn F → GL1 F is normal (as any kernel is).
More generally, any subgroup H of GLn F containing SLn F is normal because it is the
inverse image of the normal subgroup det(H) of commutative GL1 F. So we have a family
of normal subgroups H parameterized by all subgroups of the multiplicative group GL1 F.
Recall that every finite subgroup of GL1 F is cyclic. So in the case of a finite F with q
elements, the number of normal subgroups H in this family equals the number of divisors
of q − 1.
Secondly, we will see soon (Corollary 1.2 below) that the center of the group GLn F
consists of all scalar matrices λ1n , where λ ∈ GL1 F. Since every central subgroup is normal,
we get another family of of normal subgroups H parameterized by all subgroups of the
multiplicative group GL1 F.
Our main result in this section (due to Dickson [Dic1] for an arbitrary field, while for
some fields this had been known previously [Dic2]) is that there are no other normal H,
i.e.,that every normal subgroup H of GLn F either contains SLn F or is contained in the
center Gn (F, 0) with the following two exceptions: n = 2, q = 2 or 3, H = [SL2 F, SL2 F ]
(the commutator subgroup). See Theorem 1.8 below.
For any associative ring A with 1, we denote by GLn A the group of all invertible n × n
matrices with entries in A. In other words, GLn A is the multiplicative group of the ring
Mn A of all n × n matrices over A.
A matrix α is elementary if α = ai,j (where a ∈ A, i 6= j) differs from the identity
matrix 1n by one entry a at a off-diagonal position (i, j). The identity matrix 1n is the
trivial elementary matrix (with a = 0). Multiplication by an elementary matrix on right
(left) is a column (row) addition operation. For any subset B of A, we denote by En B the
subgroup of GLn A generated by all elementary matrices bi,j with b ∈ B, i 6= j.
Lemma 1.1. For any ring A with 1 and any n ≥ 2, a matrix α ∈ Mn A commutes
with all elementary matrices 1i,j if and only if α is a scalar matrix c1n . Such a matrix α
commutes with all elementary matrices ai,j if and only if α is a scalar matrix c1n with c
in the center C of the ring A.
Proof. The “if” parts are obvious. Let α = (αi,j ) commute with all elementary
matrices 1i,j . Considering the positions (1,1) and (1,2) in 11,2 α = α11,2 , we obtain that
α2,1 = 0 and α1,1 = α2,2 . Similarly, all other off-diagonal entries of α are zeros and all
diagonal entries are the same.
Let now a scalar matrix α = c1n commute with all elementary matrices a1,2 . Con-
sidering the position (1,2) in a1,2 α = αa1,2 , we conclude that ac = ca for a ∈ A, i.e., c
belongs to the center of the ring A. QED.
Corollary 1.2. The centralizer of En A in GLn A coincides with the center Gn (A, 0)
of GLn A, and it coincides with (GL1 C)1n where C is the center of the ring A.

1
 For any two elements g, h of a group G, we set [g, h] = ghg −1 h−1 . The commutator
subgroup [G, G] of G is defined as the subgroup of G generated by all commutators. It is
a normal subgroup. A group G is called perfect if G = [G, G].
Lemma 1.3. For any ring A with 1 and any integer n ≥ 3, the group En A is perfect.
Proof. It follows from the identity [ai,j , 1j,k ] = ai,k where a ∈ A, i 6= j 6= k 6= i. QED.
Now we return to the case when A is a field.
Theorem 1.4. For any field F and any n, we have En F = SLn F. Moreover every
matrix in SLn F is a product of n2 elementary matrices.
Proof. It is clear that the determinant of an elementary matrix is 1. Conversely, given
any α = (αi,j ) ∈ SLn F, we can prove that it is a product of n2 elementary matrices by
induction on n as follows (the case n = 1 being trivial with the number of elementary
matrices equal to 0). If a off-diagonal entry in the last row or column of α is nonzero,
then by one addition operation we can arrange the (n, n)-entry to be 1. After this, we
can reduce the last column of α to the last column of 1n by at most n − 1 row addition
operations. Next by at most n − 1 row addition operations we can reduce the last row of
α to the last row of 1n .
Thus,
 by at most 2n − 1 addition operations we can reduce α to a matrix of the
β 0
form with β ∈ SLn−1 F. By the induction hypothesis, β is a product of (n − 1)2
0 1
elementary matrices, hence α is a product of 2n − 1 + (n − 1)2 = n2 elementary matrices.
If all off-diagonal entries in the last row and column of α are zero, then we write
 
  1n−2 0 0
β 0 
α= 0 γ 0 
0 1
0 0 1/γ

with β ∈ SLn−1 (F ), γ = 1/αn,n and we again done by induction on n using the following
identity known as the Whitehead lemma which holds for an arbitrary ring A and an
arbitrary γ ∈ GL1 A:
      
γ 0 1 0 1 γ−1 1 0 1 1/γ − 1
(1.5) = .
0 1/γ −1/γ 1 0 1 1 1 0 1

QED.
The group E2 F is not perfect for the fields F = Z/2Z and F = Z/3Z. Therefore the
group E2 A is not perfect for any ring A with 1 which has an ideal of index 2 or 3 (e.g.,
A = Z, the integers). Using (1.5), we can prove that E2 A is perfect for some rings.
Lemma 1.6. Let A be a ring with 1 such that the additive group A is gemerated by
the set {γaγ − a : γ ∈ GL1 A, a ∈ A}.
 Then the group E2 A is perfect.
γ 0
Proof. By (1.5), g = ∈ E2 A hence
0 1/γ
(γaγ − a)1,2 = [g, a1,2 ] ∈ [E2 A, E2 A]
and
(γaγ − a)2,1 = [g −1 , a2,1 ] ∈ [E2 A, E2 A].

2
 Now use the identity
(1.7) (a − b)1,2 = a1,2 (b1,2 )−1 . QED.
Theorem 1.8 (L. Dickson). Let F be a field, and let n ≥ 2 be an integer. In the case
when n = 2 assume that F has at least 4 elements. Then every non-central subgroup H
of GLn F which is normalized by elementary matrices, contains SLn F.
Corollary 1.9. Under the conditions of Theorem 1.8, the group SLn F modulo its
center is a simple group.
To prove Theorem 1.8, we do some computations for an arbitrary ring A with 1.
Proposition 1.10. Let H be a subgroup of GLn A which is normalized by all elemen-
tary matrices, and n ≥ 3. Assume that H contains a non-central matrix with a off-diagonal
entry equal to 0. Then H contains En B for a nonzero ideal B of A.
Proof. Set B = {b ∈ A : b1,2 ∈ H}. The identity (1.7) shows that B is an additive
subgroup of A. The identities
(1.11) [ai,j , bj,k ] = (ab)i,k (i 6= j 6= k 6= i)
show that B is an ideal of A.
We have to prove that B 6= 0. Let α = (αi,j ) be a non-central matrix in H.
Case 1: α is an elementary matrix, i.e. α = ai,j with a 6= 0 and i 6= j. The identities
(1.11) show that a ∈ B, so B is anonzero ideal. 
c1n−1 0
Case 2: α has the form α = with c ∈ C, the center of A. Pick a nonzero
u c
entry uj in the row u and k ≤ n − 1 distinct from j. Then [α, 1j,k ] = (uj )n,k ∈ H, so we
are reduced to Case  1. 
β 0
Case 3: α = . Because of Case 2, we can assume that β 6= 1n−1 . If β and
u 1  
n,i 1n−1 0
1n−1 differ in row i, then [1 , α] = ∈ H with w 6= 0 being the difference of
w 1
the k-th rows of 1n−1 and β −1 . So we are reduced to Case 2.
Case 4: α1,n = 0. If α commutes with all an,j where a ∈ A, j ≤ n − 1, then we have
n,j
Case 2. Suppose thatα does not  commute with an elementary matrix a . Then
1n−1 0
1n 6= [an,j , α] = ∈ H, so we are reduced to Case 2 with c = 1.
u 1
General case, i.e., αi,j = 0 for some i 6= j. Then (γαγ −1 )1,n = 0 for a permutation
matrix γ. The group γHγ −1 is normalized by γEn Aγ −1 = En A and contains a non-
central matrix γαγ −1 . By Case 4, En B ⊂ γHγ −1 for a nonzero ideal B of A. Therefore
En B = γ −1 En Bγ ⊂ H. QED.
Corollary 1.12. Let H be a subgroup of GLn A which is normalized by all elementary
matrices, and n ≥ 3. Assume that H contains a matrix α such that Ab = A or bA = A for
an off-diagonal entry b of α. Then En B ⊂ H for a nonzero ideal B of A.
Proof. Without loss of generality, we can assume that b = α1,n . If Ab = A, then we
write α2,n = ab with a ∈ A. The matrix β = (βi,j ) = a1,2 α(−a)1,2 ∈ H is not scalar and
β2,n = 0, so we are done by Proposition 1.10.
If bA = A, we write α1,2 = ba with a ∈ A. The matrix β = (βi,j ) = (−a)1,2 αa1,2 ∈ H
is not scalar and β2,n = 0, so we are done by Proposition 1.10. QED.

3
 Proof of Theorem 1.8. For any field A = F and for any b ∈ A we have either b = 0 or
Ab = bA = A. So B = A for any nonzero ideal B of A. Therefore Theorem 1.8 with n ≥ 3
is covered by Proposition 1.10 and Corollary 1.12.
Assume now that n = 2. Since the condition of Lemma 1.6 holds when F = A has at
least 4 elements and since every non-central subgroup H in Theorem 1.8 contains a matrix
which is not diagonal, we obtain the desired conclusion from the following result involving
more general rings A :
Proposition 1.13. Under the conditions of Lemma 1.6, let H be a subgroup of GL2 A
which is normalized by E2 A. Suppose that H contains a matrix with an off-diagonal entry
in GL1 A. Then E2 A ⊂ H.
Proof. We consider the set H1,2 of (1,2)-entries of all matrices in H. Since
     
0 1 1 1 1 0 1 1
δ= = ∈ E2 A
−1 0 0 1 −1 1 0 1

and since    
a b −1 d −c
δ δ = ,
c d −b a
 
a b
it is clear that H1,2 = −H2,1 . Let α = ∈ H with b ∈ GL1 A. Then
c d
 
−1 2,1 −1 2,1 0 b
β = (b a) α(−b a) = ∈H
c0 d0

hence c0 ∈ GL1 A. Let  H1,2 = A. In fact we claim more: for any z ∈ A there
 us show that
1 z
is z 0 ∈ A such that ∈ H. Indeed,
z 1 + z0z
0

 
1 z
= [β, (−z)2,1 ] ∈ H
z0 1 + z0z

with z 0 = −c0 zb−1 .

Now we want to prove that the intersection of H with A2,1 contains all (γaγ − a)2,1
with γ ∈ GL1 A, a ∈ A.
We have    
1 γ 0 1 1
α= ,α = ∈H
c 1 + cγ c0 1 + c0
for some c, c0 ∈ A. So
 
2,1 2,1 0 γ
β = (1/γ) α(−1/γ) = ∈H
−1/γ ∗

and  
0 2,1 0 2,1 0 1
β =1 α (−1) = ∈ H.
−1 ∗

4
 Therefore
  −1
1 −a 1 −a
(γaγ − a) 2,1
= = [β, a1,2 ][β 0 , a1,2 ] ∈ H.
γaγ 1 − aγaγ a 1 − a2

Thus, A1,2 ⊂ H, hence A2,1 ⊂ H and E2 A ⊂ H. QED.

Remark 1. When F consists of 2 elements, the group SL2 F = GL2 F has order 6
and is isomorphic to the symmetric group S3 . (To see this, consider the action of GL2 F
on nonzero vectors on the palne F 2 . The action is transitive.) Its commutator subgroup
is the only nontrivial proper normal subgroup. It has order 3 and consists of the matrices
     
1 0 1 1 0 1
, , .
0 1 1 0 1 1

Remark 2. When F consists of 3 elements, the group PSL2 F = SL2 F/{±12 } has
order 12 and is isomorphic to the alternating group A4 . (To see this, consider the action of
PSL2 F on the lines passing the origin in the palne F 2 . The action is transitive.) The group
SL2 F has four normal subgroups: {12 }, {±12 }, [SL2 F, SL2 F ]. SL2 F. These subgroups are
also normal in GL2 F, so GL2 F, has five normal subgroups. The
 commutator subgroup
1 1
[SL2 F, SL2 F ] is a cyclic group of order 6 generated by . We have [SL2 F, GL2 F ]
−1 0
= [GL2 F, GL2 F ] = SL2 F.
Remark 3. Our Corollary 1.9 is Theorem 6.14 of [J]. Our elementary matrix ai,j is
is denoted by Tij (a) in [J] and Bi,j,a in [Dic1].

For the rest of section we consider finite generation of the group En A. First of all, if A
is any associative ring with 1 and X any subset of A, the relations (1.7) (1.11) show that
En X = En X 0 where X 0 is the subring of A generated by X and En (A, X) = En (A, B)
where B is an ideal of A generated by X, provided that n ≥ 3.
Therefore, we get
Proposition 1.14. For any associative ring A and any n ≥ 3, the group En A is
finitely generated if and only if the ring A is finitely generated.
More precisely, if for a ring A and any n ≥ 2, the group En A is generated by m
matrices then the ring A is generated by 2mn2 element which are the matrix entries of the
generators and their inverses.
Conversely, if an associative ring A is generated by d elements and n ≥ 3 then it is
clear that the group En A is generated by d(n2 − n) elementary matrices.
We will improve the latter bound for rings A with 1 proving, in particular, that
the group En A is generated by 2 elements for sufficiently large n (depending on d). By
Steinberg [Ste], for any finite field F and any n ≥ 2, the group En F = SLn F is generated
by 2 elements. It is also known to be true when F is any factor ring of the integers (we
will show this below).
We will use the following refinement of Case 3 of Proposition 1.10.
Proposition 1.15. Let A be an associative ring with 1, A0 a subring containing 1,
n ≥ 3, H a subgroup of GLn A. Assume that H is normalized by En A and that H contains

5
a matrix α = (αi,j ) which coincides with the identity matrix 1n in a row or a column.
Then En B 0 ⊂ H where B 0 is the subring of A generated by all a1 αi,j a2 with i 6= j and
a1 , a2 ∈ A0 together with all a1 (αi,i − 1)a2 with a1 , a2 ∈ A0 .  
β 0
Proof. Without loss of generality we can assume that either α = or α =
  u 0
β v
with β ∈ GLn−1 A.
0 0
 
β 0
These two cases are similar so let α = ∈ H with β = (βi,j ) ∈ GLn−1 A and
u 0
an (n − 1)-row u.
Set B = {a ∈ A : a1.2 ∈ H}. By (1.7) (1.11), B is a subring of A, BA0 + A0 B ⊂ B,
and En B ∈ H. So it suffices to show that αi,j ∈ B for all i 6= j and that αi,i − 1 ∈ B for
all i ≤ n − 1.
We have (αi,j )n,i = (βi,j )n,i = [[1n,i , α], 1j,i ] ∈ H when i, j ≤ n − 1 and i 6= j, hence
αi,j ∈ B. When i ≤ n − 1, we have (αi,i − 1)n,j = [[1n,1 , α], 1i,j ] where j 6= i, j ≤ n − 1,
hence αi,i − 1 ∈ B.
Thus, B contains all entries of the matrix β − 1n−1 . Similarly, B contains all entries
of the matrix β −1 − 1n−1 .
It remains to show that ui ∈ B for all i ≤ n − 1. Without loss of generality, we can
0 0
 
1n−1 + v u 0
assume that i = 1. We have γ = [(−1)1,2 , α] = (−1)1,2 0 ∈ H where
 0 u 1 u 1
v
is the first column of α and u0 is the second row of β −1 .
u1
Since v 0 u0 = 0, wehave
1n−1 + v 0 u0 0 1n−1 v 0
    
1n−1 0
[ , ].
0 1 0 1 u0 1
Since 
1n−1 0
(−1)n,2 ∈ H
u0 1
and 
1n−1 v 0

(−1)1,n ∈ H,
0 1
1n−1 v 0
   
n,2 1,n 1n−1 0
we conclude that δ = [(−1) , (−1) ][ , ] ∈ H.
0 1 u0 1
We used that H is normalized by 11,n and 1n,2 .
Since
[(−1)n,2 , (−1)1,n ] = [(−1)1,n , (−1)n,2 ]−1 = (−1)1,2 ,
1n−1 + v 0 u0 0 1n−1 v 0
     
1,2 1n−1 0
we obtain that δ = (−1) [ , ] ∈ H,
0 1 0 1 u0 1
 
−1 1n−1 0
hence µ = δ α = ∈ H.
u1 u0 1
Recall that u0 = (u0j ) is the second row of β −1 , so u0 w = 1 for the second row w = (wi )
of β. Also recall that we have shown that En (x) ∈ H for every entry x of β − 1n−1 .

6
 Therefore, for any distinct i, j ≤ n − 1 we have [µ, (wj )j,i ] = (u1 u0j wj )n,i ∈ H, hence
u1 u0j wj 3 B. Taking the sum over j, we obtain that u1 ∈ B. QED.
0
Corollary 1.16. Let A be an associative ring with 1, A a subring containing 1,
n ≥ 3, H a subgroup of GLn A. Assume that H is normalized by En A and that H contains
a matrix α = (αi,j ) such that α1,1 ) = 1, αn,1 ) = 0, and (α−1 )n,n ) = 1. Then En (B 0 ) ⊂ H
where B 0 is the subring of A generated by all a1 αi,j a2 with i 6= j and a1 , a2 ∈ A0 together
with all a1 (αi,i − 1)a2 with a1 , a2 ∈ A0 .
Proof. We apply Proposition 1.15 to the matrix β = [(−1)1,n , α] = 11,n , (1n +vu) ∈ H
where v is the first column of α and u is the last row of α−1 . Note that the last row of
β coincides with the last row of 1n and that the βi,n = αi,1 = vi for 2 ≤ i ≤ n − 1. We
obtain that (ui )1,2 ∈ H for 2 ≤ i ≤ n − 1.
Now we apply Proposition 1.15 to the matrix
n−1
Y
γ= (−ui )i,1 α.
i=2

The first column of γ is the same as the first column of 1n , and the ring generated by all
entries of γ − 1n together with all vi (2 ≤ i ≤ n − 1) is the same as the ring generated by
all entries of α − 1n . QED.
Theorem 1.17. Let A be an associative ring with 1 which is generated by 1 together
with d elements. Then
(a) the group En A is generated by 2 elements for all n ≥ 2 when d = 0, i.e.., A is a
factor ring of the integers;
(b) the group En A is generated by m + 2 elements provided that
n ≥ 3 and m(n2 − n − 1) ≥ d;
(c) the group En A is generated by m + 2 elements when
n ≥ 7 and (n − 5)2 /4 + m(n2 − n − 1) ≥ d;
in particular, En A is generated by 2 elements when n ≥ 7 and (n − 5)2 /4 ≥ d.
Proof. Let α = (αi,j ) ∈ En A be the monomial matrix defined by αi,i+1 = 1 for
i = 1, . . . , n − 1, αn,1 = (−1)n−1 , and αi,j = 0 when j 6= i + 1 and (i, j) 6= (n, 1).
(a) If d = 0, we claim that En A is generated by α and 11,2 . Indeed, let H be the
subgroup generated by α and 11,2 .
When n = 2 we have 12,1 = α(−1)1,2 α−1 ∈ H, hence H = E2 A because the ring A is
generated by 1.
Let now n ≥ 3. Then 1i,i+1 = α1−i 11,2 αi−1 ∈ H for i = 1, . . . , n − 1 and (−1n−1 )n,1 =
α1−n 11,2 αn−1 ∈ H.
Using the relations (1.7), (1,11), we obtain that H contains all elementary matrices,
hence H = En A.
(b) Our m + 2 generators are going to be the monomial matrix α, the elementary
matrix 11,2 , and m matrices of the form βk γk with 1 ≤ k ≤ m where each βk is a lower
diagonal matrix with 1 along the diagonal and (n, 1)-entry 0 and each βk is an upper
diagonal matrix with 1 along the diagonal.
We place d generators of the ring A as entries of the matrices βk , γk . Note that for
each k the subring of A generated by the entries of βk and γk is the same as the the subring

7
of A generated by the entries of βk γk . Therefore it suffices to show that the subgroup Hk
of En A generated by the three matrices α, 11,2 , and βk γk contains En Bk where Bk is the
subring of A generated by the entries of βk γk .
By (a), H contains En A0 where A0 is the subring of A generated by 1. Now we apply
Corollary 1.16 to each matrix βk γk and conclude that En Ak ⊂ H where Bk is the subring
of A generated by all entries of βk γk − 1n .
It remains to observe that Ak is also the subring of of A generated by all entries of
βk − 1n and γk − 1n .
(c) We set s = (n − 5)/2 when n ≥ 7 is odd and s == (n − 6)/2 when n ≥ 8 is
even. We define the matrix γ0 ∈ En A by placing s(n − 5 − s) generators at the positions
(i, j) with 4 ≤ i ≤ s + 3 and s + 6 ≤ i ≤ n − 1 (if d < s(n − 5 − s), we place zeros at
some of those positions) in the identity matrix 1n . The remaining generators of the ring,
if any, we arrange at matrices βk , γk . as in (b). In the view of (b), it remains to show that
the subgroup H of En A generated by the monomial matrix α and the matrix δ = 11,2 γ0
contains En A0 .
Since the matrix γ0 commutes with the matrices 11,2 , 12,3 = α−1 11,2 α, and α−1 γ0 α,
we have 11,3 = [δ, α−1 δα] ∈ H.
If n is odd, the conjugates of 11,3 by powers of α generate En A0 , so En A0 ⊂ H.
When n is even, H contains all 1i,j with even i − j. In particular, 12,n ∈ H. So
11,n = [δ, 12,n ] ∈ H.
For any n, the conjugates of 1n,1 by powers of α generate En A0 . Thus, En A0 ⊂ H.
GED.
Remarks. Any homomorphism f : A → B of associative rings with 1 induces ho-
momorphisms fn : GLn A → GLn B for all n ≥ 1. Notice that fn (En A) ⊂En B. When
f (A) = B, fn (En A) =En B for all n but fn (GLn A) need not to be the whole group GLn B.
(Consider Z → Z/5Z).

Problems and exercises.

Here are some problems related with this section. Some of them are easy exercises
while others are very difficult or even open. Some exercises are about associative rings and
intended to prepare the reader for the following sections.
 
−1 0
1. Prove that the matrix ∈ E2 A (which is a product of 4 elementary
0 −1
matrices by (1.5)) is a product of 3 elementary matrices if and only if 2 = 0 in the ring A.
2. For any field F show that a matrix in E2 F = SL2 F is product of three elementary
matrices if and only if the matrix is either non-diagonal or trivial.
3. For every field F and every n ≥ 3, show that every matrix in En F = SLn F is a
product of n2 − 1 elementary matrices.
4. For any infinite field F and any n ≥ 2, show that not every matrix in En F is a
product of n2 − 2 elementary matrices. (Hint: use the fact that every m + 1 polynomials
in m variables over a field are algebraically dependent.)

8
 5. For any field F and any n ≥ 2, show that there is a number q such that not every
matrix in En F is a product of n2 − 2 elementary matrices provided that F has at least
q elements. (Hint: Over a field of q elements, there are (q − 1)(n2 − n) + 1 elementary
matrices while the group SLn A has order (q n − 1)(q n − q 2 ) · · · (q n − q n−1 /(q − 1)).)
6. (Open problem) For which finite fields F and integers n ≥ 3, every matrix in En F
is a product of n2 − 2 elementary matrices?
any field F, show that every matrix in Mn F of rank k ≤ n − 1 can be reduced
 7. For 
1k 0
to by n2 row and column addition operations.
0 0
8. Prove that SLn Z = En Z for all n. It is known (Carter-Keller [CK]) that for n ≥ 3
every matrix in SLn Z is a product of a bounded number (depending on n) of elementary
matrices, but this is not true for n = 2 (prove it!). In other words, the product A of
infinitely many copies of Z has the following property:
(*) SLn A = En A for n ≥ 3 but SL2 A 6= E2 A.
9. Prove (*) for A = F [x, y] where F is a field (Cohn, Suslin).
10. Prove (*) when A is the ring of all real continuous functions of one variable
(Vaserstein).
11. Prove (*) when A is the ring of all real continuous functions of three variables
(Thurston-Vaserstein).
12. (Open problem) Is SL2 A = E2 A when A = Z[x, 1/x]?
13. (Open problem) Given a non-central matrix α ∈ SLn F over a field F, is it true
that every non-central matrix in SLn F is a product of n matrices, each similar to α?
14. Prove that every non-central matrix in SLn F over a field F is a commutator.
What about scalar matrices?
Qn λ1 , µ1 , . . . , λn , µn in a field F and a nonscalar matrix α ∈ GLn F
15. Given elements
such that det(α) = i=1 λi , µi show that there are β, γ ∈ GLn F such that α = βγ, the
eigenvalues of β are λi , and the eigenvalues of γ are µi .
16. Let A be an associative ring with 1, a ∈ A, and aA = Aa = A. Prove that a ∈
GL1 A.
 
a 0
17. Let A be an associative ring with 1 and ∈ GL2 A. Prove that a.d ∈
0 d
GL1 A.
18. Let A be a finitely generated associative ring with 1. Show that for sufficiently
large n the ring Mn A admits two generators.
19. Let F be an infinite field of characteristic p 6= 0. Let F 0 be a field of characteristic
6= p. Let f : SL2 F → GLn F 0 be a group morphism (integer n > 0 is arbitrary). Prove that
f is trivial.
20. Let A be an associative domain with 1 (“domain” means that A has no zero
divisors, i.e., if xy = 0 in A then either x = 0 or y = 0). Prove that if xy = 1 then yx = 1.
21. Let A be a ring with 1, and A∞ the direct product of infinitely many copies of A.
For any n, clearly GLn (A∞ ) = (GLn A)∞ so the natural group homomorphism En (A∞ ) →

9
(En A)∞ is injective. Show that this homomorphism is surjective if and only if every matrix
in En A is a product of a bounded number of elementary matrices.
22. Let A be an associative ring with 1. The columns An with n entries over A form
a free right A-mosule with n free generators. Every finitely generated free right A-mosule
is isomorphic to An for some n ≥ 0. (In general, n is not unique.) When every right ideal
of A is finitely generated, A is called Noetherian. Prove that if A is right Noetherian then
every submodule of An is finitely generated for every n.
Prove that if for some k every right ideal of A is generated by k elements, then every
submodule of An is generated by kn elements, for any n ≥ 1.
There are rings with k = 1, known as right principal idea rings (PIRs), e.g., A = Z,
the integers, and with k = 2, e.g., the rings of algebraic integers and the rings of differential
operators with polynomial coefficients.
23. Let A be an associative domain with 1 It is called a rigt Ore donain if aA∩bA 6= {0}
for any nonzero a, b ∈ A, Clearly, every commutative domain is an Ore domain.
Prove that every right Ore domain is isomorphic to a subring of a division ring.
Prove that every right Noetherian domain (see the previous exercise) is a right Ore
domain.
24. Let A be an associative domain with 1. It is called a right Bézout domain if for
every a, b ∈ A there is c ∈ A such that aA + bA = cA. Clearly, every right principal ideal
domain is a Bézout domain and that every finitely generated right ideal in right Bézout
domain is principal. Here are two examples of commutative Bézout domains which are not
PIRs: the ring of entire functions, the ring of all algebaic integers.
Prove that every right Bézout domain is an Ore domain (see the previous exercise).
25. An associative ring with 1 is called a right Euclidean if there a function f from
A to the non-negative integers such that for for any nonzetro a, b ∈ A there is q ∈ A such
that f (a − bq) < f (b). Here are three examples of commutative Euclidean domains:
Z is Euclidean with f (a) = |a|,
any field F is Euclidean with f (a) = 1 for a 6= 0 and f (0) = 0.
F [t] is Euclidean for any field F with f (a) = deg(a) + 1 for a 6= 0 and f (0) = 0.
Prove that every right Euclidean ring is a right PIR (see Exercise 22 above).
26. Let A1 be an associative ring with 1, n ≥ 2, and A = Mn A1 . Show that every
element of A is the sum of elements of the form γaγ − a with a ∈ A and γ ∈ GL1 A =
GLnA.
27. Let A be an associative ring with 1 and n ≥ 2. Show that 1n is the sum of
2 invertible matrices in Mn A and that every matrix in Mn A is the sum of 3 invertible
matrices (i.e., matrices in GLn A.)
28. Let A be an associative ring with 1 and n ≥ 1. Show that every ideal of Mn A has
the form Mn B where B is an ideal of A.

10