Carry Groups

Tyler Cook1 , Blair Madore2 , Erica Miller3 , Tim Pollio4 , Zachariah Riel5 ,
Anneliese Spaeth67

Edited by Zachariah Riel

March 1, 2007

1
SUNY Potsdam
2
SUNY Potsdam
3
St. Edwards University
4
Princeton University
5
Kent State University
6
Xavier University
7
We thank Dr. Joel Foisy, SUNY Potsdam, Clarkson University, NSF, and NSA for their support through the
Clarkson-Potsdam REU program.
Introduction

Consider the subset of Z Z2 Z4 Z8 consisting of those elements where all but …nitely many
coordinates are zero. We will de…ne an additive operation on this set. Instead of adding these elements in the
usual coordinate-wise fashion, we will introduce the notion of "carrying" from the …nite cyclic groups into
Z. In the following example, we will add coordinate-wise, but we will carry 1, twice, into the Z coordinate.
Here, we carry a 1 into the Z coordinate whenever the sum of the elements of a coordinate exceeds its
modulus. Let a +c b denote the least nonnegative residue of a + b, modulo c; let a c b denote the least
nonnegative residue of ab, modulo c. For x 2 R, let the ‡oor of x be denoted by bxc. In this example, we
"add" (5; 1; 2; 3; 0; 0; :::) and (1; 1; 1; 7; 6; 0; 0; :::) as follows.

Z Z2 Z4 Z8 Z16 :::
( 5; 1; 2; 3; 0; 0; ::: )
( 1; 1; 1; 7; 6; 0; ::: )
= ( 5 + 1 + 2; 1 +2 1; 2 +4 1; 3 +8 7; 0 +16 6; 0; ::: )
= ( 8; 0; 3; 2; 6; 0; ::: )

The subset of elements, where all but …nitely many coordinates are zero, of Z Z2 Z4 Z8
forms an Abelian group under this additive operation, which will be denoted by . We denote this group by
1 1 1 1
Z Z2 Z4 Z8 ; more precisely, we may denote it by Z Z2 Z4 Z8 , since each time we exceed
the modulus of a coordinate when adding two elements, 1 is carried into the Z coordinate. We say this,
a1 a2 a3 a4
because we also consider in…nitely (and …nitely) generated groups of the form Z Zb1 Zb2 Zb3 (or
a1 a2 an
Z Zb1 Zb2 Zbn ). Groups of this type, we will call carry groups.
Dr. Jonathan L. King …rst de…ned the carry group Z Z2 Z4 Z8 and claimed that it could be used
to solve a problem in ergodic theory concerning the roots of a measure preserving transformation1 . This
problem was solved by Dr. Blair Madore2 , though a complete understanding of the structure of this group
and its …nitely generated subgroups was not needed for the solution. During the 2001 and 2006 Potsdam-
Clarkson NSF-REU summer programs, Madore and his students have further investigated the structure of
carry groups. We will now de…ne carry groups more formally.
All of our groups are written additively and all …nite cyclic groups Zn are represented by f0; 1; :::; n 1g.
Let (nk ) be a (…nite or in…nite) sequence of positive integers greater than 1. Also, let H be an Abelian group
and let (hk ) be a (…nite or in…nite) sequence of elements of H. We de…ne a group with an operation on
the Cartesian product H Zn1 Zn2 Zn3 Znm . The operation is de…ned by
(a; a1 ; a2 ; a3 ; :::; am ) (b; b1 ; b2 ; b3 ; :::; bm ) =
m
!
X ai + bi
a+b+ hi ; a1 +n1 b1 ; a2 +n2 b2 ; a3 +n3 b3 ; :::; am +nm bm .
i=1
ni

Simply, this is coordinate-wise addition, where hi is carried into the H coordinate if ai +bi ni . Conveniently,
we denote this group by
h1 h2 h3 hm
H Zn1 Zn2 Zn3 Znm
1 J. L. King, The generic transformation has roots of all orders. Dedicated to the memory of Anzelm Iwanik. Colloq. Math.

84/85 part 2, pp 521-547, 2000.

2 B. F. Madore, Rank-one group actions with simple mixing Z subactions, Ph.D. Thesis, University of Toronto, 2000.

1
so that we know we’re using the operation . This also helps so that we can see which elements of H are
carried for each coordinate. Note that these groups are …nitely generated. We can similarly de…ne in…nitely
generated carry groups. In particular, these groups are de…ned on the subset of H Zn1 Zn2 Zn3 ,
where all but …nitely many coordinates of the elements are zero. The operation , in this case, is de…ned by

(a; a1 ; a2 ; a3 ; :::) (b; b1 ; b2 ; b3 ; :::) =

1
!
X ai + bi
a+b+ hi ; a1 +n1 b1 ; a2 +n2 b2 ; a3 +n3 b3 ; ::: .
i=1
ni
P1 j k
ai +bi
We’re guaranteed that i=1 hi ni is an element of H, since all but …nitely many coordinates of the
h1 h2 h3
elements are zero. We similarly denote this group by H Zn1 Zn2 Zn3 . One can check that these
are all Abelian groups; it is tedious, but straightforward. In this paper, for most of our carry groups, we will
have H = Z.
2 4
Example. Consider (2; 1; 2); (1; 3; 4) 2 Z Z4 Z5 . We compute (2; 1; 2) (1; 3; 4) as follows.
2 4
Z Z4 Z5
( 2; 1; 2 )
( 1; 3; 4 )
= ( 9; 0; 1 )

Example. In the carry group Z Z4 Z5 , consider the element a = (4; 2; 1). We calculate a, the
inverse of a, to be
a = ( 4 1 1; 4 2; 5 1) = ( 6; 2; 4) ,
since
(4; 2; 1) ( 6; 2; 4) = ( 6; 2; 4) (4; 2; 1) = (0; 0; 0) .

a1 a2 an
In general, given any element (m; m1 ; m2 ; :::; mn ) of the carry group H Zb1 Zbn , the inverse is
given by
(m; m1 ; m2 ; :::; mn ) = (m0 ; m01 ; m02 ; :::; m0n ) ,
where
X 0 if mi = 0,
m0 = m ai and m0i = for i = 1; :::; n.
bi mi if mi 6= 0,
1 i n
mi 6=0
a1 a2 an
Now, for (m; m1 ; m2 ; :::; mn ) 2 H Zb1 Zbn , let the element

(m; m1 ; m2 ; :::; mn ) (m; m1 ; m2 ; :::; mn ) (m; m1 ; m2 ; :::; mn )

| {z }
\k" tim es

be denoted by k (m; m1 ; m2 ; :::; mn ). What does the element k (m; m1 ; m2 ; :::; mn ) look like explicitly? The
answer to that question may not be immediately clear. However, a straightforward induction argument will
a1 a2 an
verify that, given k 2 N and (m; m1 ; m2 ; :::; mn ) 2 H Zb1 Zbn ,
0 1
k
X1 X n
i bj mj + mj
k (m; m1 ; m2 ; :::; mn ) = @km + aj ; k b1 m1 ; k b2 m2 ; :::; k bn mn A . ( 1)
i=0 j=1
bj

Furthermore, let’s …nd an explicit formula for k (m; m1 ; m2 ; :::; mn ). Since we’re working with an Abelian
group, we have that k (m; m1 ; m2 ; :::; mn ) = k ( (m; m1 ; m2 ; :::; mn )). For all j = 1; 2; :::; n, let

0 if mj = 0,
m0j =
bj mj if mj 6= 0.

2
We have that

k (m; m1 ; m2 ; :::; mn )
= k ( (m; m1 ; m2 ; :::; mn ))
0 1
B X C
= kB
@ m ai ; m01 ; m02 ; :::; m0n C
A
1 i n
mi 6=0
0 1
X k
X1 X
n
B i bj m0j + m0j C
= B
@ km k ai + aj ;k b1 m01 ; k b2 m02 ; :::; k bn m0n C
A
i=0 j=1
bj
1 i n
mi 6=0
0 1
B X k
X1 X i m0j + m0j C
B bj C
= B km k ai + aj ; k b1 m1 ; k b2 m2 ; :::; k bn mn C
@ i=0 1 j n
bj A
1 i n
mi 6=0 m0j 6=0
0 1
X k
X1 X
B i bj (bj mj ) + bj mj C
= B
@ km k ai + aj ; k b1 m1 ; k b2 m2 ; :::; k bn mn C
A
i=0 1 j n
bj
1 i n
mi 6=0 mj 6=0
0 1
k
X1 X
B i bj mj + bj mj C
= B
@ km + aj 1 ; k b1 m1 ; k b2 m2 ; :::; k bn mn C
A
i=0 1 j n
bj
mj 6=0
0 1
k
X1 X
B i bj mj + ( mj ) C
= B
@ km + aj ; k b1 m1 ; k b2 m2 ; :::; k bn mn C
A ( 2)
i=0 1 j n
bj
mj 6=0

Using ( 1 ) and ( 2 ), we come upon the following proposition.

a1 a2 an
Proposition. Given any nonzero integer d and any (m; m1 ; m2 ; :::; mn ) 2 H Zb1 Zbn ,

d (m; m1 ; m2 ; :::; mn )
0 $ % 1
jdj 1 n
X X i sgn (d) bj mj + mj sgn (d)
= @dm + aj ;d b1 m1 ; d b2 m2 ; :::; d bn mn A . ( )
i=0 j=1
bj

By now, the reader should be a bit more familiar with what we call carry groups and the operation
. Now, our aim is to present some theory regarding the structure of these carry groups. For the most
part, our classi…cation of these groups is complete. For …nitely generated carry groups, we can surely use
the Fundamental Theorem of Finitely Generated Abelian Groups and the Smith Normal Form to …nd their
structure. However, our theory provides alternative and interesting methods that yield their structure. We
have come upon an understanding of …nitely and in…nitely generated carry groups, which will be discussed
in the following sections.

3
Chapter 1

Structure Theory for Finitely

Generated Uniform Carry Groups

During the summer of 2001, Dr. Blair Madore’s1 NSF-REU group formulated a structure theory for
…nitely generated carry groups of the form
a a a a
Z Zb1 Zb2 Zbn .

For the sake of simplicity, let’s refer to them as uniform carry groups. We will now state and prove their
main results.2 The following fact is trivial. However, it will be used often throughout this paper. It would
be best to keep it in mind.

Lemma 1.1. If a; b; n 2 Z, 0 a < n, and 0 b < n, then

a+b
n + (a +n b) = a + b. (1)
n

Proof. If a + b n, then 1 a+b

n < 2, meaning that
a+b
n = 1. Also, a +n b = a + b n. Therefore
n n + (a +n b) = n 1 + (a + b n) = a + b. If a + b < n, then a+b
a+b
n < 1 so that
a+b
n = 0. Hence,
a+b
n n + (a +n b) = (a +n b) = a + b.

Proposition 1.2. Z Zn = Z.

Proof. De…ne ' : Z Zn ! Z by ' (a; b) = na + b. Let (a1 ; b1 ) ; (a2 ; b2 ) 2 Z Zn . Therefore,

b1 + b 2 b1 + b 2
' ((a1 ; b1 ) (a2 ; b2 )) = ' a1 + a2 + ; b1 +n b2 = n a1 + a2 + + (b1 +n b2 )
n n
b1 + b 2 (1)
= n (a1 + a2 ) + n + (b1 +n b2 ) = n (a1 + a2 ) + (b1 + b2 )
n
= ' (a1 ; b1 ) + ' (a2 ; b2 ) .
1 SUNY Potsdam
2 B.
Madore, H. McDonough, C. Miller, D. Van Nort, A. Rogalski, and J. Wood, “Structure Theory for Carry Groups”, Pi
Mu Epsilon Journal, Vol. 12 (Fall 2004) 1, 17-25.

4
Now, if m 2 Z, then there exist unique integers q and r such that m = nq + r and 0 r < n. Therefore
'
(q; r) 7! m so that ' is surjective. Now, if ' (a1 ; b1 ) = ' (a2 ; b2 ) = m, then na1 + b1 = na2 + b2 = m.
We know that 0 b1 ; b2 < n. The division algorithm tells us that m has a unique representation of the
form nq + r, where 0 r < n. To avoid a contradiction, we must have (a1 ; b1 ) = (a2 ; b2 ). Alternatively,
show that j(0; 1)j = 1 and Z Zn = h(0; 1)i.

m
Proposition 1.3. Z Zm Zn = Z Zn .

m
Proof. De…ne : Z Zm Zn ! Z Zn by (a; b; c) = (ma + b; c). Let (a1 ; b1 ; c1 ) ; (a2 ; b2 ; c2 ) 2
Z Zm Zn . So,

b1 + b2 c1 + c2
((a1 ; b1 ; c1 ) (a2 ; b2 ; c2 )) = a1 + a2 + + ; b1 +m b2 ; c1 +n c2
m n
b 1 + b2 c1 + c2
= m a1 + a2 + + + (b1 +m b2 ) ; c1 +n c2
m n
c1 + c2
= m a1 + a2 + + (b1 + b2 ) ; c1 +n c2
n
= (ma1 + b1 ; c1 ) (ma2 + b2 ; c2 ) = (a1 ; b1 ; c1 ) (a2 ; b2 ; c2 ) .

As in Proposition 1.2, surjectivity and injectivity both follow from the division algorithm.

m
Proposition 1.4. Z Zn = Z Zgcd(m;n) .

Proof. As m and n are positive integers, we can surely …nd r; s 2 Z such that mr+ns = gcd (m; n) =
m
d and 0 < r < n. Now de…ne : Z Zn ! Z Zd by

nx + my
(x; y) = ; rx +d sy . (2)
d
nx+my m
Note that d 2 Z since d j m and d j n. Let (a1 ; b1 ) ; (a2 ; b2 ) 2 Z Zn . We have that

b1 + b 2
((a1 ; b1 ) (a2 ; b2 )) = a1 + a2 + m ; b1 +n b2
n
!
b1 +b2
n a1 + a2 + m n + m (b1 +n b2 ) b1 + b 2
= ; r a1 + a2 + m +d s (b1 +n b2 )
d n
n (a1 + a2 ) + m (b1 + b2 ) b1 + b 2
= ; r a1 + a2 + m +d s (b1 +n b2 )
d n
n (a1 + a2 ) + m (b1 + b2 ) b1 + b2
= ; ra1 +d ra2 +d rm +d sb1 +d sb2
d n
n (a1 + a2 ) + m (b1 + b2 )
= ; ra1 +d ra2 +d sb1 +d sb2
d
na1 + mb1 na2 + mb2
= + ; (ra1 +d sb1 ) +d (ra2 +d sb2 )
d d
na1 + mb1 na2 + mb2
= ; ra1 +d sb1 ; ra2 +d sb2 = (a1 ; b1 ) (a2 ; b2 ) .
d d

5
Importantly, notice that

m n
; = (0; 1) and (s m; n r) = (1; 0) .
d d

Since is a homomorphism, we know that for (a; b) 2 Z Zd ,

m n
a (s m; n r) b ; = (a; b) :
d d
na1 +mb1 na2 +mb2
Thus, is surjective. Now, if (a1 ; b1 ) = (a2 ; b2 ), then d ; ra1 +d sb1 = d ; ra2 +d sb2 .
Hence,
n m
(a1 a2 ) + (b1 b2 ) = 0 and (3)
d d
r (a1 a2 ) +d s (b1 b2 ) = 0. (4)

We know that gcd m n m

d ; d = 1. Hence, (3) implies that d j (a1 a2 ) and nd j (b1 b2 ). It follows that
m n
k d = ja1 a2 j and k d = jb1 b2 j for some nonnegative integer k. Without loss of generality, suppose
a1 a2 = k m n
d and k d = b1 b2 . Thus, r (a1 a2 ) + s (b1 b2 ) = k r dm + s nd = k dd = k. By
(4), we know that k 0 (mod d). Since 0 b1 ; b2 < n, we know that jb1 b2 j < n. If k 6= 0, then
jb1 b2 j = k nd n as d j k; this is a contradiction. Therefore, k = 0 =) b1 = b2 =) a1 = a2 .
Hence, is injective. The proof is complete.

Corollary 1.5. Z Zm Zn = Z Zgcd(m;n) .

m
Lemma 1.6. If K is an abelian group and ' : Z Za1 ! K is an isomorphism such that ' (m; 0) = M ,
then
m m m m M M M
Z Za1 Za2 Zan = K Za2 Zan .

m m m m M M M
Proof. Let G = Z Za1 Za2 Zan and let H = K Za2 Zan . Let the operation of K
be denoted by . Let : G ! H be de…ned by

(x; x1 ; x2 ; :::; xn ) = (' (x; x1 ) ; x2 ; :::; xn ) .

If (x; x1 ; x2 ; :::; xn ) = g1 and (y; y1 ; y2 ; :::; yn ) = g2 are elements of G, then

n
!
X xi + yi
(g1 g2 ) = x+y+m ; x1 +a1 y1 ; x2 +a2 y2 ; :::; xn +an yn
i=1
ai
n
! !
X xi + yi
= ' x+y+m ; x1 +a1 y1 ; x2 +a2 y2 ; :::; xn +an yn
i=1
ai
n
! !
X xi + yi
= ' (x; x1 ) ' (y; y1 ) ' m ; 0 ; x2 +a2 y2 ; :::; xn +an yn
i=2
ai
n
! !
X xi + yi
= ' (x; x1 ) ' (y; y1 ) M; x2 +a2 y2 ; :::; xn +an yn
i=2
ai
= (' (x; x1 ) ; x2 ; :::; xn ) (' (y; y1 ) ; y2 ; :::; yn ) = (g1 ) (g2 ) .

It is very clear that is a bijection since ' is a bijection.

6
 (q;0) (q;0) (q;0) q q q
Lemma 1.7. (Z Zd ) Zb2 Zbn = Zd Z Zb2 Zbn .

(q;0) (q;0) (q;0) q q q

Proof. Let G = (Z Zd ) Zb2 Zbn and let G0 = Zd Z Zb2 Zbn . Consider
0
the mapping : G ! G , which is de…ned by

((x; x1 ) ; x2 ; :::; xn ) = (x1 ; (x; x2 ; :::; xn )) .

It is easily shown that is a homomorphism; also, from the de…nition of , bijectivity quickly follows.

Now, for example, consider the carry group

Z Z22 Z6 Z4 .
'
We know that Z Z22 = Z. Moreover, there is an isomorphism ' : Z Z22 ! Z such that (1; 0) 7! 22.
Therefore, Lemma 1.6 tells us that
22 22
Z Z22 Z6 Z4 = Z Z6 Z4 .
22
Now, it follows from Proposition 1.4 that Z Z6 = Z Zgcd(22;6) = Z Z2 and there is an isomorphism
22 22 6
: Z Z6 ! Z Z2 such that (22; 0) 7! gcd(22;6) ; 0 = (lcm (22; 6) ; 0) = (66; 0). Applying Lemma 1.6
again, we have that
22 22 (66;0)
Z Z6 Z4 = (Z Z2 ) Z4 .
It follows from Lemma 1.7 and Proposition 1.4 that
(66;0)
(Z Z2 ) Z4
66
= Z2 Z Z4
= Z2 Z Zgcd(66;4)
= Z2 Z Z2 .

Actually, we can follow the same procedure with any carry group and …nd its structure. We will now
generalize this procedure.
Consider the carry group G = Z Za1 Za2 Zan . Note that Z Za1 = Z by an isomorphism that
maps (1; 0) to a1 . Therefore, Lemma 1.6 tells us that
a1 a1 a1
G=Z Za2 Zan :
a1
Proposition 1.4 tells us that Z Za2 = Z Zgcd(a1 ;a2 ) by an isomorphism that maps (a1 ; 0) to (lcm (a1 ; a2 ) ; 0).
Lemma 1.6 tells us that
(lcm(a1 ;a2 );0) (lcm(a1 ;a2 );0) (lcm(a1 ;a2 );0)
G= Z Zgcd(a1 ;a2 ) Za3 Zan .

We now apply Lemma 1.7 to get that

lcm(a1 ;a2 ) lcm(a1 ;a2 ) lcm(a1 ;a2 )
G = Zgcd(a1 ;a2 ) Z Za3 Zan :

Using the Principle of Mathematical Induction, we come upon the following theorem.

Theorem 1.8. If a1 ; a2 ; :::; an are positive integers greater than 1, then

n
!
M
Z Za1 Za2 Zan = Z Zgcd(lcm(a1 ;:::;ai 1 );ai )
.
i=2

7
 According to the theorem above, …nding the structure of any uniform carry group comes down to a few
simple calculations. Using the theorem and the same example from above, we can quickly …nd that

Z Z22 Z6 Z4
= Z Zgcd(22;6) Zgcd(lcm(22;6);4)
= Z Z2 Z2 :

In the following sections, we will discuss some more structure theory that applies to all carry groups (and
not just uniform carry groups).

8
Chapter 2

General Structure Theory for Finitely

Generated Carry Groups

During the 2006 Potsdam-Clarkson NSF-REU, with Dr. Madore, we1 further investigated carry groups
and continued the work of the 2001 REU program. There, we completely classi…ed the structure of …nitely
generated carry groups. Our classi…cation for in…nitely generated carry groups is, for the most part, complete.
We will now state and prove our main results.
a1 a2 an a01 a02 a0n
Lemma 2.1. Z Zb1 Zbn = Z Zb1 Zbn if ai a0i (mod bi ) for all i = 1; :::; n.

a1 a2 an a01 a02 a0n

Proof. Let G = Z Zb1 Zbn and let H = Z Zb1 Zbn . Suppose that ai a0i (mod bi )
for all i = 1; :::; n. Therefore, for each i = 1; :::; n,
a0i = ai + mi bi
for some mi 2 Z. Now, let ' : G ! H be de…ned by
n
!
X
' (x; x1 ; :::; xn ) = x mi xi ; x1 ; :::; xn .
i=1

If (c; c1 ; :::; cn ) and (d; d1 ; :::; dn ) are elements of G, then

' ((c; c1 ; :::; cn ) (d; d1 ; :::; dn ))
n
!
X ci + di
= ' c+d+ ai ; c1 +b1 d1 ; :::; cn +bn dn
i=1
bi
n
!
X ci + di
= c+d+ ai mi (ci +bi di ) ; c1 +b1 d1 ; :::; cn +bn dn
i=1
bi
n
!
X ci + di
= c+d+ (a0i mi b i ) mi (ci +bi di ) ; c1 +b1 d1 ; :::; cn +bn dn
i=1
bi
n
!
X ci + di
= c+d+ a0i mi (ci + di ) ; c1 +b1 d1 ; :::; cn +bn dn
i=1
bi
n
! n
!
X X
= c mi ci ; c1 ; :::; cn d mi di ; d1 ; :::; dn = ' (c; c1 ; :::; cn ) ' (d; d1 ; :::; dn ) .
i=1 i=1

Now, surjectivity and injectivity follow quickly from the de…nition of '.
1 TylerCook - SUNY Potsdam, Erica Miller - St. Edwards University, Tim Pollio - Princeton University, Zachariah Riel -
Kent State University, Annaliese Spaeth - Xavier University

9
 a1 a2 an ca1 a2 an
Lemma 2.2. If gcd (b1 ; c) = 1, then Z Zb1 Zbn = Z Zb1 Zbn .

a1 a2 an ca1 a2 an
Proof. Let G = Z Zb1 Zbn and let H = Z Zb1 Zbn . Let ' : H ! G be de…ned by
' (x; x1 ; x2 ; :::; xn ) = c (0; x1 ; 0; :::; 0) (x; 0; x2 ; :::; xn ) .
If (d; d1 ; d2 ; :::; dn ) and (e; e1 ; e2 ; :::; en ) are elements of H, then
' ((d; d1 ; d2 ; :::; dn ) (e; e1 ; e2 ; :::; en ))
n
!
d1 + e1 X di + ei
= ' d + e + ca1 + ai ; d1 +b1 e1 ; d2 +b2 e2 ; :::; dn +bn en
b1 i=2
bi
n
!
d1 + e1 X di + ei
= c (0; d1 +b1 e1 ; 0; :::; 0) d + e + ca1 + ai ; 0; d2 +b2 e2 ; :::; dn +bn en
b1 i=2
bi
d1 + e1
= c (0; d1 ; 0; :::; 0) (0; e1 ; 0; :::; 0) a1 ; 0; 0; :::; 0
b1
n
!
d1 + e1 X di + ei
d + e + ca1 + ai ; 0; d2 +b2 e2 ; :::; dn +bn en
b1 i=2
bi
n
!
X di + ei
= c (0; d1 ; 0; :::; 0) c (0; e1 ; 0; :::; 0) d+e+ ai ; 0; d2 +b2 e2 ; :::; dn +bn en
i=2
bi
= c (0; d1 ; 0; :::; 0) c (0; e1 ; 0; :::; 0) (d; 0; d2 ; :::; dn ) (e; 0; e2 ; :::; en )
= (c (0; d1 ; 0; :::; 0) (d; 0; d2 ; :::; dn )) (c (0; e1 ; 0; :::; 0) (e; 0; e2 ; :::; en ))
= ' (d; d1 ; d2 ; :::; dn ) ' (e; e1 ; e2 ; :::; en ) .
Since gcd (b1 ; c) = 1, we can …nd integers r and s such that b1 r + cs = 1. Clearly, c b1 s = 1. Let q be
the least nonnegative residue of s, modulo b1 . Now, let (p; p1 ; p2 ; :::; pn ) 2 G. It follows from ( ) that
0 1
jcj 1
X (i sgn (c) b1 q) + q sgn (c)
c (0; q; 0; :::; 0) = @ a1 ; 1; 0; :::; 0A
i=0
b1

and so 0 1
jcj 1
X (i sgn (c) b1 q) + q sgn (c)
'@ a1 ; q; 0; :::; 0A = (0; 1; 0; :::; 0) .
i=0
b1

Therefore, since ' is a homomorphism, we have that

0 1
jcj 1
X (i sgn (c) b1 q) + q sgn (c) '
p1 @ a1 ; q; 0; :::; 0A (p; 0; p2 ; :::; pn ) 7 ! (p; p1 ; p2 ; :::; pn ) .
i=0
b1

Thus, ' is surjective. Suppose that (d; d1 ; d2 ; :::; dn ) ; (e; e1 ; e2 ; :::; en ) 2 H such that ' (d; d1 ; d2 ; :::; dn ) =
' (e; e1 ; e2 ; :::; en ). Thus,
(d; 0; d2 ; :::; dn ) c (0; d1 ; 0; :::; 0) = (e; 0; e2 ; :::; en ) c (0; e1 ; 0; :::; 0) .
Surely, di = ei for i = 2; :::; n. The previous equation gives us
0 1
jcj 1
X (i sgn (c) b1 d1 ) + d1 sgn (c)
@d + a1 ;c b1 d1 ; d2 ; :::; dn A
i=0
b1
0 1
jcj 1
X (i sgn (c) b1 e1 ) + e1 sgn (c)
= @ e+ a1 ;c b1 e1 ; e2 ; :::; en A .
i=0
b1

Since cd1 ce1 (mod b1 ) and gcd (b1 ; c) = 1, we have that d1 e1 (mod b1 ). Consequently, d1 = e1 .
This forces d = e. Thus, ' is injective and the proof is complete.

10
 The following lemma gives a few minor facts about carry groups. At this point, the reader should be
familiar enough with carry groups to see (or at least have a strong suspicion) that they are true. We could
probably get away with assuming these facts without stating them. For the sake of clarity, we will state
them. However, we will leave it to the reader to prove the lemma, as it is a straightforward exercise. Note
that the second fact is a generalization of Lemma 1.6; its proof is similar to that of 1.6. The third fact is
practically identical with Lemma 1.7.

Lemma 2.3.
a1 a2 an a (1) a (2) a (n)
(a) If 2 Sn , then Z Zb1 Zbn = Z Zb (1)
Zb (n)
.
(b) Let H and K be Abelian groups. If H = K via ' : H ! K, then for any h1 ; :::; hn 2 H,

h1 h2 hn '(h1 ) '(h2 ) '(hn )

H Zb1 Zbn = K Zb1 Zbn .

(a2 ;0) (a3 ;0) (an ;0) a2 a3 an

(c) (Z Zd ) Zb2 Zbn = Zd Z Zb2 Zbn .

a1 a2 an ca1 a2 an
Lemma 2.2 tells us that if gcd (b1 ; c) = 1, then Z Zb1 Zbn = Z Zb1 Zbn . Actually, we
can generalize this lemma a bit. Now, Lemma 2.3(a) tells us that the structure of a carry group does not
change if we rearrange the …nite cyclic groups along with their respective carries. Therefore, we can apply
2.3(a) and 2.2 over and over again. Without loss of generality, if gcd (bi ; ci ) = 1 for all i = 1; :::; n, then
a1 a2 an c1 a1 c2 a2 cn an
Z Zb1 Zbn = Z Zb1 Zbn . From now on, whenever we refer to Lemma 2.2, we will
refer to it in this general sense.

Lemma 2.4. Given a; b 2 N, there exists c 2 N such that gcd(b; c) = 1 and ac gcd(a; b)(mod b).

Proof. So, let a; b 2 N. Let b = p1 1 p2 2 pl l be a prime factorization for b. It is clear that, for all
i = 1; :::; l, there exist ci ; i 2 Z such that a ci pi i (mod pi i ), where ci pi i < pi i and gcd(ci ; pi ) = 1.
Also, for all i = 1; :::; l, there exist ki ; i 2 Z such that gcd(a; b) ki pi i (mod pi i ), where ki pi i < pi i
and gcd(ki ; pi ) = 1. We claim that, for all i = 1; :::; l, i = i . First, suppose that j > j for some j.
We know that we can write gcd(a; b) = ra + sb for some r; s 2 Z. Therefore,

kj pj j gcd(a; b) ra + sb ra rcj pj j (mod pj j ).

Hence,
j j
kj rcj pj (mod pj j ).
j j
Now, this implies that kj = rcj pj + qj pj j for some qj 2 Z. Hence, pj j kj . Since pj > 1, this is
a contradiction of the fact that gcd(kj ; pj ) = 1. Thus, i i for all i. Now, suppose that j < j
for some j. So, since gcd(a; b) j a, we can write a = w gcd(a; b) for some w 2 Z. Now, we have that
j
cj pj j a w gcd(a; b) wkj pj j (mod pj j ). Therefore, we have that cj wkj pj j (mod pj j ). Since
j < j , we have that pj j cj , which contradicts the fact that gcd(cj ; pj ) = 1. Therefore, we have that
i i for all i. Since we also have that i i for all i, we conclude that i = i for all i.

Now, we can …nd a c such that c ki ci 1 (mod pi i ) for all i = 1; :::; l, which is guaranteed by the
Chinese Remainder Theorem. Also, for all i = 1; :::; l, we have that
1 1
ac aki ci ci pi i ki ci pi i ki = pi i ki gcd(a; b)(mod pi i ).

Also, by the Chinese Remainder Theorem, we have that ac gcd(a; b)(mod b).

11
 a1 a2 an
Lemma 2.4 is quite useful for our cause. Consider the carry group Z Zb1 Zbn . For each i = 1; :::; n,
it follows from Lemma 2.4 that there exists ci 2 N such that gcd(bi ; ci ) = 1 and ai ci gcd(ai ; bi )(mod bi ).
It follows from Lemma 2.2 and Lemma 2.1 that
a1 a2 an
Z Zb1 Zbn
c1 a1 c2 a2 cn an
= Z Zb1 Zbn
gcd(a1 ;b1 ) gcd(a2 ;b2 ) gcd(an ;bn )
= Z Zb1 Zbn .
a1 a2 an
Thus, without loss of generality, we can assume that if given any carry group Z Zb1 Zbn , then ai j bi
for all i = 1; :::; n. The following lemma uses this general assumption. As you will see later, the following
a1 a2 an
lemma is a necessary tool for …nding the structure of Z Zb1 Zbn .

Lemma 2.5. If ai j i, bi j i, and gcd ( i ; i) = 1 for all i = 1; :::; n, then

a1 b1 a2 b2 an bn a1 b1 a2 an bn
Z Z 1 1
Z n n
=Z Z 1
Z 1
Z n
Z n
.

a1 b1 a2 b2 an bn a1 b1 a1 b1 a2 b2 an bn an bn
Proof. Let G = Z Z 1 1
Z n n
and let H = Z Z 1 Z 1
Z n Z n
.
In G, let

y = (1; 0; 0; :::; 0)
y1 = (0; 1; 0; :::; 0)
..
.
yn = (0; 0; 0; :::; 1) .

In H, let

x = (1; 0; 0; :::; 0)
x11 = (0; 1; 0; :::; 0)
x12 = (0; 0; 1; 0; :::; 0)
..
.
xn1 = (0; 0; 0; :::; 1; 0)
xn2 = (0; 0; 0; :::; 1) .

Let ' : H ! G be de…ned by

n
' (q; q11 ; q12 ; :::; qn1 ; qn2 ) = qy ( i qi1 yi i qi2 yi ) ,
i=1

where is the operator corresponding to the operation . Let (c; c11 ; c12 ; :::; cn1 ; cn2 ) ; (d; d11 ; d12 ; :::; dn1 ; dn2 ) 2
H. Then,

' ((c; c11 ; c12 ; :::; cn1 ; cn2 ) (d; d11 ; d12 ; :::; dn1 ; dn2 ))
n
!
X ci1 + di1 ci2 + di2
= ' c+d+ ai bi + ; c11 + 1
d11 ; c12 + 1
d12 ; :::; cn1 + n
dn1 ; cn2 + n
dn2
i=1 i i
n
!
X ci1 + di1 ci2 + di2 n
= c+d+ ai bi + y i (ci1 + i
di1 ) yi i ci2 + i
di2 yi .
i i i=1
i=1

12
Note that, in the previous line,
n
i (ci1 + i
di1 ) yi i ci2 + i
di2 yi
i=1
n n
= i (ci1 + i di1 ) yi i ci2 + i
di2 yi
i=1 i=1
n n
i ci1 + i di1 i ci2 + i di2
= i ci1 yi i di1 yi ai bi y i ci2 yi i di2 yi ai bi y
i=1 i i i=1 i i
n ci1 + di1 n ci2 + di2
= i ci1 yi i di1 yi ai bi y i ci2 yi i di2 yi ai bi y
i=1 i i=1 i
n
!
X ci1 + di1 ci2 + di2 n
= ai bi + y ( i ci1 yi i di1 yi i ci2 yi i di2 yi ) .
i i i=1
i=1

Therefore,
' ((c; c11 ; c12 ; :::; cn1 ; cn2 ) (d; d11 ; d12 ; :::; dn1 ; dn2 ))
n
!
X ci1 + di1 ci2 + di2 n
= c+d+ ai bi + y i (ci1 + i
di1 ) yi i ci2 + i
di2 yi
i i i=1
i=1
n
= (c + d) y ( i ci1 yi i di1 yi i ci2 yi i di2 yi )
i=1
n n
= cy ( i ci1 yi i ci2 yi ) dy ( i di1 yi i di2 yi )
i=1 i=1
= ' (c; c11 ; c12 ; :::; cn1 ; cn2 ) ' (d; d11 ; d12 ; :::; dn1 ; dn2 ) .
So, ' is a homomorphism. For each i = 1; :::; n, we can …nd ri ; si 2 Z such that i ri + i si = 1, as
gcd ( i ; i ) = 1. For each i = 1; :::; n, let Ri be the least nonnegative residue of ri , modulo i . Also,
for each i, let Si be the least nonnegative residue of si , modulo i . For each i = 1; :::; n, there are
integers mi ; ni such that ri = Ri + mi i and si = Si + ni i . Therefore, for each i = 1; :::; n,

i ri + i si = i (Ri + mi i) + i (Si + ni i) i Ri + i Si 1(mod i i ).

Therefore, for each i = 1; :::; n,

' i Si + i Ri
Si xi1 Ri xi2 7 ! i Si yi i Ri yi = yi ai bi y
i i

and so
i Si + i Ri '
ai bi x Si xi1 Ri xi2 7 ! yi .
i i

Let (p; p1 ; :::; pn ) 2 G. Therefore,

n
i Si + i Ri '
px pi ai bi x Si xi1 Ri xi2 7 ! (p; p1 ; :::; pn ) .
i=1 i i

So, ' is surjective. Now, suppose that (c; c11 ; c12 ; :::; cn1 ; cn2 ) ; (d; d11 ; d12 ; :::; dn1 ; dn2 ) 2 H such that
' (c; c11 ; c12 ; :::; cn1 ; cn2 ) = ' (d; d11 ; d12 ; :::; dn1 ; dn2 ). Then,
n n
cy ( i ci1 yi i ci2 yi ) = dy ( i di1 yi i di2 yi ) .
i=1 i=1

Hence,
n
!
X n
i ci1 + i ci2
c+ ai bi y i ci1 + i i i ci2 yi
i i i=1
i=1
n
!
X n
i di1 + i di2
= d+ ai bi y i di1 + i i i di2 yi .
i i i=1
i=1

13
 Therefore, for each i = 1; :::; n,

i ci1 + i ci2 i di1 + i di2 (mod i i)

=) i (ci1 di1 ) + i (ci2 di2 ) = ki i i

for some ki 2 Z. Since i j ki i i and i j i (ci1 di1 ) for i = 1; :::; n, it is forced that i j i (ci2 di2 )
for i = 1; :::; n. Therefore, for all i = 1; :::; n, i j (ci2 di2 ) since gcd ( i ; i ) = 1. If ci2 6= di2 for some
i, then 0 < jci2 di2 j < i and i - (ci2 di2 ), which is a contradiction. We must have that ci2 = bi2
for i = 1; :::; n. Similarly, i j (ci1 di1 ) for i = 1; :::; n. Since jci1 di1 j < i , we must have that
ci1 = di1 for all i in order to avoid a contradiction. This forces c = d. Therefore, H = G. Now, let
a1 b1 a2 an bn
K=Z Z 1 Z 1 Z n Z n . Since bi j i for i = 1; :::; n, we have that gcd ( i ; bi ) = 1 for
i = 1; :::; n. Similarly, ai j i for all i implies that gcd (ai ; i ) = 1 for all i. It follows from Lemma 2.2
that H = K. Therefore, G = K.

Given a carry group, what are the necessary and su¢ cient conditions for it to be isomorphic to Z? The
following theorem tells us precisely when a carry group is isomorphic to Z.

a1 a2 an
Theorem 2.6. Let G = Z Zb1 Zbn . The following are equivalent:

(a) The integers b1 ; :::; bn are pairwise relatively prime and gcd (ai ; bi ) = 1 for all i = 1; :::; n.
(b) G = Z.

a1 a2 an
Proof. [(b) =) (a)] Let G = Z Zb1 Zbn . Assume that G = Z. We wish to show that the
integers b1 ; :::; bn are pairwise relatively prime and gcd (ai ; bi ) = 1 for all i = 1; :::; n. Suppose that it
isn’t true. Then, there exist 1 i1 < i2 n such that gcd (bi1 ; bi2 ) > 1 or there exists 1 j n such
that gcd (ai ; bi ) > 1.
aj
(i) Suppose that there exists 1 j n such that gcd (aj ; bj ) > 1. We know that Z Zbj < G. By
aj aj
Proposition 1.4, Z Zbj = Z Zgcd(aj ;bj ) . Therefore, there is a corresponding element of Z Zbj (in
G) that has order gcd (aj ; bj ). Thus, G has an element of …nite order. This contradicts the fact that
G = Z and Z has no elements of …nite order. Therefore, we must have that gcd (ai ; bi ) = 1 for all
i = 1; :::; n.
(ii) Suppose that there exist 1 i1 < i2 n such that gcd (bi1 ; bi2 ) = d > 1. We know that
ai1 ai2 ai1 ai2
Z Zbi1 Zbi2 < G. Consider the element q 2 Z Zbi1 Zbi2 given by

b i1 b i2
q= ai (0; 1; 0) ai (0; 0; 1) .
d 2 d 1

I claim that q 6= (0; 0; 0). Otherwise, if the third coordinate of q is zero, then we would have to conclude
b
that di2 ai1 is a multiple of bi2 ; this means that d j ai1 . So, gcd (ai1 ; bi1 ) d > 1 since d j bi1 . This
violates the fact that gcd (ai ; bi ) = 1 for all i = 1; :::; n. Hence, q 6= (0; 0; 0). We now get that

b i1 b i 2 bi 1 bi 2
b i 1 b i2 q = bi1 ai2 (0; 1; 0) bi2 ai1 (0; 0; 1)
d d
b i1 b i 2 b i1 b i 2
= ai2 (ai1 ; 0; 0) ai1 (ai2 ; 0; 0)
d d
b i1 b i 2 b i 1 b i2
= ai1 ai2 (1; 0; 0) ai1 ai2 (1; 0; 0) = (0; 0; 0) .
d d

We have found that q is an element of …nite order. We obtain the contradiction that G has elements
of …nite order. We must have that b1 ; :::; bn are all pairwise relatively prime.

14
 [(a) =) (b)] Suppose that we’re given positive integers b1 ; :::; bn and integers a1 ; :::; an such that
a1 a2 an
b1 ; :::; bn are pairwise relatively prime and gcd (ai ; bi ) = 1 for all i = 1; :::; n. Let G = Z Zb1 Zbn .
Using Lemma 2.2, Theorem 1.8, and the hypothesis that b1 ; :::; bn are pairwise relatively prime, we …nd
that

G = Z Zb1 Zbn
= Z Zgcd(b1 ;b2 ) Zgcd(lcm(b1 ;b2 );b3 ) Zgcd(lcm(b1 ;b2 ;:::;bn 1 );bn )

= Z Zgcd(b1 ;b2 ) Zgcd(b1 b2 ;b3 ) Zgcd(b1 b2 bn 1 ;bn )

= Z Z1 Z1 Z1 = Z.

We have a method for …nding the structure of any …nitely generated carry group, which we will now
present. Suppose that we’re given the carry group
a1 a2 an
G=Z Zb1 Zbn .

According to the lemmas (2.4, 2.2, and 2.1), without loss of generality, we can assume that ai j bi for all i.
We can now rewrite this as
1m1 2m2
p1111 p1212 p1m p2121 p2222 p2m pn1n1 pn2n2 nmn
pnm
1 2 n
G=Z Z 1m Zp n1 n2 nmn ,
p1111 p1212 p1m 1 n1 pn2 pnm n
1

im
where pi1i1 pi2i2 pimi i is the prime factorization of bi . For each 1 i n, ij ij for all 1 j mi .
Now, by Lemma 2.5,
nmn
p1111 p1212 p1313 pnmn
G=Z Zp 11 Zp 12 Zp nmn . (#1 )
11 12 nmn

This may look a bit ambiguous. To be clear, the group written in (#1 ) consists of m1 + + mn = N
carry “components,” whereas the original group G consisted of n carry “components.” We say that a direct
sum of …nite cyclic groups with any group of this form is “reduced.” We will now show that the group in
(#1 ) can be transformed from an N -component group to an (N 1)-component reduced group. Choose any
prime
p 2 fpij : 1 i n and 1 j mi g .
Let
= min f ij : p = pij for some 1 i n and 1 j mi g .
For each 1 i n and each 1 j mi , let

1 if pij = p;
ij =
p if pij 6= p.

Applying Lemma 2.2, we get that

11 12 13 nmn
11 p11 12 p12 13 p13 nmn pnmn
G=Z Zp 11 Zp12 Zp nmn .
11 12 nmn

ij
Note that p j ij pij for all 1 i n and all 1 j mi . In light of Lemma 2.3(a), we can assume that
p
p = p11 and = 11 . Further, by Proposition 1.4, there is an isomorphism :Z Zp 11 !Z Zp , where
(p ; 0) = p 11 ; 0 . Using Lemma 2.3(b), we have that
nmn
( 12
12 p12 ;0 ) ( 13
13 p13 ;0 ) ( nmn pnmn ;0 )
G = (Z Zp ) Zp 12 Zp nmn .
12 nmn

15
Notice that, for each 1 i n, ij pij
ij
;0 = ij pij
ij
p( 11 )
; 0 for all 1 j mi (why?). So now, it
follows from Lemma 2.3(c) that
12 ( 11 ) 13 ( 11 ) nmn ( 11 )
( 12 p12 p ;0) ( 13 p13 p ;0) ( nmn pnmn p ;0)
G = (Z Zp ) Zp 12 Zp nmn
12 nmn
!
12 ( 11 ) 13 ( 11 ) nmn ( 11 )
12 p12 p 13 p13 p nmn pnmn p
= Zp Z Zp 12 Zp nmn . (#2 )
12 nmn

So, we have transformed the N -component group to an (N 1)-component reduced group as desired. Of
course, this iterated process can be used to express G as a direct sum of Z with …nite cyclic groups. So, this
method yields the structure of any …nitely generated carry group. However, this can be improved.
Fix any 1 i n and 1 j mi . Recall that ij = p whenever pij 6= p. Therefore, if pij 6= p, then
ij (
ij pij p
11 )
= pijij p 11 . In (#2 ), if pij 6= p, then the ij th carry “component” has the form

pijij p 11

Z .
pijij

We know that gcd p 11 ; pijij = 1, since pij 6= p. Thus, we can apply Lemma 2.2 to rewrite (#2 ), where
th
the ij carry “component” becomes
pijij
Z .
pijij

This is the ij th carry “component” that we started with in (#1 ). Recall that when we chose the prime
p = p11 , we transformed the carry “component”

p1111
Zp 11
11

into a direct summand. In general, the argument above shows that the transformation of a carry “component”
into a direct summand does not a¤ect the carry components whose bases are powers of a prime di¤erent
from p.
Therefore, this problem amounts to reducing groups of the form
p 1 p 2 p n
H=Z Zp 1 Zp n ,

where p is a prime, i i for i = 1; :::; n, and (without loss of generality) 1 2 n. We now

present our main structure theorem for …nitely generated carry groups.

Theorem 2.7. Suppose that p is a prime. If i i for i = 1; :::; n and 1 2 n, then

n
!
p 1 p 2 p n M
Z Zp 1 Zp n =Z ZpMi ,
i=1

where M1 = 1 and Mi = min i; i + i 1 i 1 for i > 1.

Proof. Suppose that p is a prime. Let 1 ; :::; n and 1 ; :::; n be integers such that i 0, i > 0,
i i , and i i+1 for each i = 1; :::; n. Let M1 = 1 and Mi = min i; i + i 1 i 1 for
each i > 1. We claim that, for each 1 k n,
k
!
M p k+1 + k k p k+2 + k k p n+ k k
H= ZpMi Z Zp k+1 Zp n .
i=1

16
 p 1
By Lemma 1.4, there is an isomorphism 1 : Z Zp 1 !Z Zp 1 such that (p 1 ; 0) maps to p 1 ; 0 .
Lemma 2.3(b) and Lemma 2.3(c) tell us that

p 1 (p 2 ;0) (p 3 ;0) (p n ;0)

H = Z Zp 1 Zp 2 Zp n

(p 2+ 1 1 ;0 ) (p 3+ 1 1 ;0 ) (p n+ 1 1 ;0 )
= (Z Zp 1 ) Zp 2 Zp n

p 2+ 1 1 p 3+ 1 1 p n+ 1 1
= Zp 1 Z Zp 2 Zp n .

Our claim is true for k = 1. Suppose that, for some 1 l < n, our claim is true for k = l. Thus,
l
!
M p l+1 + l l p l+2 + l l p n+ l l
H= ZpMi Z Zp l+1 Zp n .
i=1

p l+1 + l l
Applying Lemma 1.4, there is an isomorphism 2 : Z Zp l+1 ! Z ZpMl+1 such that
l+1 + l
p l
; 0 maps to p l+1 ; 0 . It follows from Lemma 2.3 that

l
!
M p l+1 + l l p l+2 + l l p n+ l l
H = ZpMi Z Zp l+1 Zp n

i=1
l
! !
M (p l+2 l+1 + l+1 ;0 ) (p n l+1 + l+1 ;0 )
= ZpMi Z ZpMl+1 Zp n

i=1
l+1
!
M p l+2 l+1 + l+1 p l+3 l+1 + l+1 p n l+1 + l+1
= ZpMi Z Zp l+2 Zp n .
i=1

Therefore, our claim is true for k = l + 1 whenever it is true for k = l. By induction, it follows that
n
!
M
H=Z ZpMi .
i=1

6 10 5
Example. Let G = Z Z9 Z14 Z7 . Notice that gcd (2; 9) = 1, gcd (3; 14) = 1, and gcd (3; 7) = 1.
Thus, Lemma 2.2 tells us that
6 10 5
G = Z Z9 Z14 Z7
62 10 3 53
= Z Z9 Z14 Z7
12 30 15
= Z Z9 Z14 Z7 .

Also, note that 12 3 (mod 9), 30 2 (mod 14), and 15 1 (mod 7). Therefore, by Lemma 2.1,
3 2 1
G = Z Z9 Z14 Z7 .

Now, note that gcd (7; 2) = 1. Therefore, Lemma 2.5 tells us that
3 2 1 1
G = Z Z9 Z2 Z7 Z7 .

17
 Continuing on, using Theorem 2.7, we get that

3 2 70 70
G = Z Z9 Z2 Z71 Z71
2 70 70
= Z3 Z Z2 Z71 Z71

70 70
= Z3 Z2 Z Z71 Z71

= Z3 Z2 (Z Z70 Z7minf1;0+1 0g )
= Z3 Z2 (Z Z1 Z7 )
= Z Z3 Z2 Z7 .

At this point, our classi…cation of …nitely generated carry groups is complete. Now, we will explore
in…nitely generated Carry groups.

18
Chapter 3

In…nitely Generated Carry Groups

We will now consider groups of the form

a1 a2 a3 a4
G=Z Zb1 Zb2 Zb3 .

Note that the proofs of the lemmas in Chapter 2 did not depend on the fact that the groups were …nitely
generated (perhaps except for Lemma 2.3(a)). Therefore, without loss of generality, we assume that ai j bi
and bi is a prime power for all i 2 N. We will now give some de…nitions.

De…nitions
a1 a2 a3 a4
Suppose that we are given G = Z Zb1 Zb2 Zb3 , where ai j bi and bi is a prime power for all i.
ai
1. Portions of the expression of G of the form Zbi are referred to as components.
2. We will de…ne bi =ai to be the depth of the component.
3. The collection of all components of G whose bi are powers of a …xed prime, p, is called the p-system
of G. We denote a p-system of G as Gp and we denote its associated subgroup of G as Z Gp .
4. We de…ne the depth of a p-system to be the supremum of the depths of its components.
5. A p-system is said to be bounded if it has …nite depth.
ai
6. A bounded p-system is said to be simple if it has some component Zbi where ai = 1 and bi is equal
to the depth of the system.
7. The in…nitely generated carry group G is said to have rational form if there exist H Q and a
sequence n1 ; n2 ; ::: of natural numbers such that
1
!
M
G=H Zni .
i=1

8. An element x 2 G where x = (m; m1 ; m2 ; :::) is said to be an integer if mi = 0 for all i. We may refer
to x as m is this case.
9. For the in…nitely generated carry group G, we de…ne G : G ! Q, the projection homomorphism,
by
1
X ai
G (x; x1 ; x2 ; :::) = x + xi .
i=1
bi

19
 10. De…ne Gt = fx 2 G : x has …nite orderg. Gt is called the torsion subgroup of G. Furthermore, we
say that G is torsion-free if Gt = f0G g.

Note that the mapping given in de…nition 9, G , is actually a homomorphism; it is an easy exercise to
show that G is a homomorphism. We leave this to the reader.

Lemma 3.1. For any x 2 G, there exists n 2 N such that nx is an integer.

Proof. Let x = (x0 ; x1 ; x2 ; :::) 2 G. Remember that xi = 0 for all except …nitely many i. Now, let
Y
n= bi .
xi 6=0

We know that bi j n whenever xi 6= 0. Hence, bi j nxi whenever xi 6= 0; in this case, the ith coordinate
of nx is 0. Therefore, nx is zero in every coordinate except possibly the …rst one. Therefore, nx is an
integer by de…nition.

Lemma 3.2. For the in…nitely generated carry group G,

ker G = Gt .

Proof. Let x = (c; c1 ; c2 ; :::) 2 G. By the previous lemma, there exists n 2 N such that nx is an
integer; say, nx = (b; 0; 0; :::).
First of all, suppose that x 2 Gt ; then mx = 0G for some m 2 N. Therefore, m (nx) = mnx =
n (mx) = 0G and nx 2 Gt . Since nx is an integer, we must have that nx = 0G ; otherwise, mnx =
m (b; 0; 0; :::) = (mb; 0; 0; :::) 6= 0G since mb would be nonzero. Therefore, G (nx) = G (0G ) = 0.
Hence, n G (x) = 0 so that G (x) = 0 and x 2 ker G . Thus, Gt ker G .
Conversely, let x 2 ker G . Still, nx is an integer; also, nx 2 ker G , since G (nx) = n G (x) = 0. We
know that nx = 0G , because G (nx) = G (b; 0; 0; :::) = b. Since nx = 0G , we conclude that x 2 Gt .
Thus, ker G Gt . Hence, ker G = Gt .

Corollary 3.3. For the in…nitely generated carry group G, G=Gt = G (G).
L1
Notice that if G has rational form H ( i=1 Zni ), then Corollary 3.3 implies that G (G) = H. In this
case, it follows that G (G) is a direct summand of G.

Lemma 3.4. If Gp is a p-system of G, then there are no in…nite chains of additive pth roots in G.

a1 a2 a3 a4
Proof. Let Gp be a p-system of G = Z Zb1 Zb2 Zb3 . Let x 2 G and suppose that x has
an in…nite chain of additive pth roots. Then, for every n 2 N, there exists xn 2 G such that pn xn = x.
Thus, for every n 2 N, G (xn ) = G (x) =pn 2 Q. Therefore, for some m 2 N, G (xm ) = ab , where
gcd (a; p) = 1 and p j b. It is true that there is at least one component of the p-system, where the
corresponding coordinate of xm is nonzero. Suppose otherwise. That is, suppose xm is zero in all the
0
coordinates corresponding to components of the p-system. Then G (xm ) = ab0 where p - b0 . Hence,
ab = a b. Since p - b , p - a, and p is a prime, we have that p - ab . This is a contradiction since p j a0 b.
0 0 0 0

Thus, there is at least one component of the p-system, where the corresponding coordinate of xm is
p
nonzero. Assume, without loss of generality, that this component is Zp and that the corresponding
coordinate of xm is 6= 0. We know that p xm+ = xm . No matter what is in the Zp -coordinate
of xm+ , we have that the Zp -coordinate of p xm+ = xm is 0. This is a contradiction. Therefore,
x 2 G does not have an in…nite chain of pth roots.

20
Lemma 3.5. If some p-system of G is not bounded, then G does not have rational form.

Proof. Suppose that Gp is an unbounded system of G. Also, suppose that G has rational form with
p 1 p 2 p 3
the associated H Q. So, Z Gp has the form Z Zp 1 Zp 2 . Since Gp is not bounded,
we know that i i can take arbitrarily large values. We claim that

n
G (Z Gp ) =
: n 2 Z; m 2 N H Q.
pm
n o n o
n n n1
It is obvious that G (Z Gp ) pm : n 2 Z; m 2 N . Let y 2 pm : n 2 Z; m 2 N ; then y = pm1
for some n1 2 Z, m 2 N. Let = min fi 2 N : i i > m1 g. Therefore,

m1 n1 m1
y=p = n1 p G (0; 0; :::; 1; 0; 0; :::) ,
p
n o
where the 1 is in the Z -coordinate. Thus, y 2 (Z Gp ) and pnm : n 2 Z; m 2 N
G G (Z Gp ).
n o
n
The claim is proved. Therefore, an isomorphic copy of pm : n 2 Z; m 2 N exists as a subgroup of
n o
G. This is impossible, because the set pnm : n 2 Z; m 2 N has in…nite chains of additive pth roots,
whereas G does not. This is a contradiction. Therefore, G cannot have rational form.

a1 a2 a3 a4
Theorem 3.6. Let G = Z Zb1 Zb2 Zb3 . The following are equivalent:

(a) The integers b1 ; b2 ; ::: are pairwise relatively prime and gcd (ai ; bi ) = 1 for all i 2 N.
(b) G is torsion-free.
(c) G is isomorphic to a subgroup of Q.

Proof. [(a)
Qn=) (b)] Let x = (x0 ; x1 ; x2 ; :::) 2 Gt . There exists n0 2 N such that xi = 0 for all i > n0 .
0
Let n1 = i=1 bi . By the proof of Lemma 3.1, n1 x is an integer. So, we must have that n1 x = 0G .
Otherwise, n1 x has in…nite order. Hence, G (n1 x) = n1 G (x) = 0. This means that
(x) = 0
G
Xn0
xi ai
=) x0 + = 0.
i=1
bi
Pn0 xi ai
We know that i=1 bi must be an integer, since x0 is an integer. The common denominator for this
sum of rational numbers is n1 . Therefore,
0 1
n0
X Y
B C
n1 j Bxi ai bj C
@ A.
i=1 1 j n0
j6=i

Fix 1 k n0 . We know that 0 1

n0
X Y
B C
bk j Bxi ai bj C
@ A.
i=1 1 j n0
j6=i

Furthermore, whenever 1 i n0 and i 6= k, we have that

Y
bk j xi ai bj .
1 j n0
j6=i

21
 So, bk divides the ith summand for all i 6= k and bk divides the whole sum. To avoid a contradiction,
bk must divide the k th summand; that is,
Y
bk j xk ak bj .
1 j n0
j6=k
Q
Since all the bi are pairwise relatively prime, we know that bk and 1 j n 0 bj are relatively prime.
j6=k
Hence, bk j xk ak . Since gcd (bk ; ak ) = 1, we have that bk j xk . Since 0 xk < bk , we must have that
xk = 0. We chose k arbitrarily. Hence, for all 1 i n0 , we have that xi = 0. In light of the equation
G (x) = 0, it is forced that x0 = 0. Hence, x = 0G . We conclude that G is torsion-free.

[(b) =) (c)] Suppose that G is torsion-free. Since Gt = ker G and Gt = f0G g, we have that
ker G = f0G g. This implies that G is an injective homomorphism. Therefore, G = G (G) Q.
[(c) =) (a)] The exact same argument from the …rst half of the proof of Theorem 2.6 su¢ ces.

As in the previous section, if we reduce components whose bases are powers of a prime p, then the carries
of other components are not a¤ected whenever their bases are powers of a di¤erent prime. Therefore, the
problem of reducing an in…nitely generated carry group comes down to reducing its p-systems. The following
theorem tells us how bounded p-systems are reduced.

p 1 p 2 p 3
Lemma 3.7. Suppose that Gp is a bounded p-system, where Z Gp = Z Zp 1 Zp 2 . Then,
1
M
p 1
Z Gp = Z Zp 1 Zp i .
i=2

Proof. Suppose that G is an in…nitely generated carry group, where Gp is a bounded p-system of G.
p 1 p 2 p 3
So, Z Gp is of the form Z Zp 1 Zp 2 . Since the depth of Gp is …nite, this depth must be
p 1
achieved by one of the components. Without loss of generality, assume that this component is Zp 1 .
p 1 L1
Let H = Z Zp 1 i=2 Zp i . In H, let

y = ((1; 0) ; 0; :::; 0)
y1 = ((0; 1) ; 0; :::; 0)
y2 = ((0; 0) ; 1; :::; 0)
..
.

Let ' : Z Gp ! H be de…ned by

" 1
! # 1
X M
1+
' (x; x1 ; x2 ; :::) = xy x1 xi p 1 i i y1 xi yi .
i=2 i=2

It is straightforward (but quite tedious) to show that ' is a homomorphism. Surjectivity and injectivity
follow quickly from the de…nition of '.

Lemma 3.8. If all of the p-systems of G are bounded and if all but …nitely many of the systems of
G are simple, then G has rational form.

22
 Proof. Apply Lemma 3.7 to all of the p-systems of G. Notice that the repeated application of Lemma
c1 c2 c3
3.7 reduces G to a direct sum of …nite cyclic groups and a group of the form H0 = Z Zd1 Zd2 ,
where the di are pairwise relatively prime and ci = 1 for all but …nitely many i. For the (…nitely
many) components that satisfy ci 6= 1, we can repeatedly apply Proposition 1.4 and transform them
into direct summands. After transforming all such components into direct summands, G is reduced to
a direct sum of …nite cyclic groups and a group of the form H1 = Z Zd01 Zd02 , where the d0i are
pairwise relatively prime. Therefore, H1 satis…es Theorem 2.6(a) or Theorem 3.6(a). Hence, in any
case, H1 = H for some H Q. Thus, G has rational form.

Lemma 3.9. If G has in…nitely many non-simple systems, then G does not have rational form.

Proof. Suppose that G has in…nitely many non-simple systems and that G has rational form. Then,
essentially, G (G) is a direct summand of G. Therefore, there is a subgroup G1 G such that
G1 = G (G). Let x 2 G1 be the element that corresponds to 1 2 G (G). Let n be a number so that
nx is an integer. Let p be a prime such that p relatively prime to nx and Gp is a non-simple system of
G; note that such a p exists by hypothesis. Assume that Gp is bounded; otherwise, we have our result
by Lemma 3.5. Let pd be the depth of Gp . Since the depth is …nite, it is a maximum and must be
achieved by some component of Gp . Therefore, p d 2 G (G). Then, certainly, x and nx both have
pd-th roots. Say, pd y = nx. Let

z = (1; 0; 0; :::; 0)
z1 = (0; 1; 0; :::; 0)
z2 = (0; 0; 1; :::; 0)
..
.

be the generators for G. Then, y must be a linear combination of z; z1 ; z2 ; :::. Assume that zi has
ai
a 1 in the coordinate corresponding to the Zbi component. Given ci , we know that either pd ci zi
is not an integer or pd ci zi = pd ci bii z. We claim that p j pd ci abii if pd ci zi = pd ci abii z. Assume that
a

pd ci zi = pd ci abii z and p - pd ci abii . Then, pd j abii . Since pd is the depth of Gp , we have that abii pd
and, therefore, pd = abii . Since Gp is not simple, we have that p j ai . Therefore, pd+1 j bi . Therefore,
p j ci ; otherwise, pd+1 - pd ci =) bi - pd ci and so pd ci zi cannot be an integer. Since p j ci , we have
that pd ci abii = pd ci p d = ci and p j pd ci abii ; this contradicts our assumption that p - pd ci abii . Thus, if
pd ci zi = pd ci abii z, then p j pd ci abii . Since y is a linear combination of elements of the form cz and ci zi , we
know that pd y is a linear combination of elements of the form pd cz and pd ci zi . These elements must be
multiples of p; otherwise, they wouldn’t be integers and, thus, pd y = nx wouldn’t be an integer. Since
the elements of the linear combination are multiples of p, we must have that pd y = nx is a multiple
of p. This is a contradiction of the fact that p was chosen to be relatively prime to nx. Therefore, we
must have that G does not have rational form.

We know that every countably generated carry group satis…es the hypothesis of at least one of Lemma
3.5, Lemma 3.8, or Lemma 3.9. Therefore, we have a complete characterization of which groups have a
rational form. Therefore, in sum, we conclude with the following theorem.

Theorem 3.10. For any carry group G, G has rational form if and only if every system of G is
bounded and all but …nitely many systems of G are simple.

23
Bibliography

[1] J. L. KING, The generic transformation has roots of all orders. Dedicated to the memory of Anzelm
Iwanik. Colloq. Math. 84/85 part 2, pp 521-547, 2000.
[2] B. F. MADORE, Rank-one group actions with simple mixing Z subactions, Ph.D. Thesis, University of
Toronto, 2000.
[3] T. W. HUNGERFORD, "Algebra," Springer-Verlag, New York, 1975.
[4] L. FUCHS, "In…nite Abelian Groups," Academic Press, New York, 1970-73.
[5] N. JACOBSON, "Basic Algebra I - Second Edition," W.H. Freeman and Company, New York, 1985.
[6] D.S. DUMMIT and R.S. FOOTE, "Abstract Algebra, 2nd Edition," J.W. Wiley and Sons, New York,
1999.
[7] B. MADORE, H. McDONOUGH, C. MILLER, D. VAN NORT, A. ROGALSKI, and J. WOOD, Structure
Theory for Carry Groups, Pi Mu Epsilon Journal, Vol. 12 (Fall 2004) 1, 17-25.

[8] T. COOK, B. MADORE, E. MILLER, T. POLLIO, Z. RIEL, and A. SPAETH, Carry Groups, in prepa-
ration.

24