Multiplicative Functions and Mobius Inversion Formula

Zvezdelina Stankova
Berkeley Math Circle Director
Mills College and UC Berkeley
1. Multiplicative Functions. Overview
Denition 1. A function f : N C is said to be arithmetic.
In this section we discuss the set M of multiplicative functions, which is a subset of the set A of
arithmetic functions. Why this subset is so special can be explained by the fact that it is usually
easier to calculate explicit formulas for multiplicative functions, and not so easy to do this for
arbitrary arithmetic functions.
1
Denition 2. An arithmetic function f(n) : N C is multiplicative if for any relatively prime
n, m N:
f(mn) = f(m)f(n).
Examples. Let n N. Dene functions , , : N N as follows:
(n) = the number of all natural divisors of n = #{d > 0 | d|n};
(n) = the sum of all natural divisors of n =

d|n
d;
(n) = the product of all natural divisors of n =

d|n
d.
As we shall see below, and are multiplicative functions, while is not. From now on we shall
write the prime decomposition of n N as n = p

1
1
p

2
2
p

r
r
for distinct primes p
i
and
i
> 1. It
is easy to verify the following properties of multiplicative functions:
Lemma 1. If f is multiplicative, then either f(1) = 1 or f 0. Further, if f
1
, f
2
, ..., f
k
are
multiplicative, then the usual product f
1
f
2
f
n
is also multiplicative.
Using the prime decomposition of n N, derive the following representations of , and .
Lemma 2. The functions (n), (n) and (n) are given by the formulas:
(n) =
r

i=1
(
i
+ 1), (n) =
r

i=1
p

i
+1
i
1
p
i
1
, (n) = n
1
2
(n)
.
Conclude that and are multiplicative, while is not.
Examples. The following are further examples of well-known multiplicative functions.
(n), the Mobius function;
e(n) =
1,n
, the Dirichlet identity in A;
I(n) = 1 for all n N;
id(n) = n for all n N.
Taking the sum-functions of these, we obtain the relations: S

= e, S
e
= I, S
I
= , and S
id
= .
These examples suggest that the sum-function is multiplicative, provided the original function is
too. In fact,
1
But on a deeper level, M is special partly because it is closed under the Dirichlet product in A. For this one has
to wait until after Section 2.
Theorem 1. f(n) is multiplicative i its sum-function S
f
(n) is multiplicative.
Proof: Let f(n) be multiplicative, and let x, y N such that (x, y) = 1. Further, let x
1
, x
2
, ..., x
k
and y
1
, y
2
, ..., y
m
be all divisors of x and y, respectively. Then (x
i
, y
j
) = 1, and {x
i
y
j
}
i,j
are all
divisors of xy.
S
f
(x) S
f
(y) =
k

i=1
f(x
i
)
m

j=1
f(y
j
) =

i,j
f(x
i
)f(y
j
) =

i,j
f(x
i
y
j
) = S
f
(xy).
Hence S
f
(n) is multiplicative.
Conversely, if S
f
(n) is multiplicative, then let n
1
, n
2
N such that (n
1
, n
2
) = 1 and n = n
1
n
2
.
We will prove by induction on n that f(n
1
n
2
) = f(n
1
)f(n
2
). The statement is trivial for n = 1:
f(1) = S
f
(1)(= 1 or 0.) Assume that it is true for all m
1
m
2
< n. Then for our n
1
, n
2
we have:
S
f
(n
1
n
2
) =

d
i
|n
i
f(d
1
d
2
) =

d
i
|n
i
d
1
d
2
<n
f(d
1
d
2
) + f(n
1
n
2
) =

d
i
|n
i
d
1
d
2
<n
f(d
1
)f(d
2
) + f(n
1
n
2
).
On the other hand,
S
f
(n
1
)S
f
(n
2
) =

d
1
|n
1
f(d
1
)

d
2
|n
2
f(d
2
) =

d
i
|n
i
d
1
d
2
<n
f(d
1
)f(d
2
) + f(n
1
)f(n
2
).
Since S
f
(n
1
n
2
) = S
f
(n
1
)S
f
(n
2
), equating the above two expressions and canceling appropriately,
we obtain f(n
1
n
2
) = f(n
1
)f(n
2
). This completes the induction step, and shows that f(n) is indeed
multiplicative.
Corollary 1. The sum-function S
f
(n) of a multiplicative function f(n) is given by the formula:
S
f
(n) =
r

i=1
_
1 + f(p
i
) + f(p
2
i
) + + f(p

i
i
)
_
.
2. Dirichlet Product and M

obius Inversion
Consider the set A of all arithmetic functions, and dene the Dirichlet product of f, g A as:
f g(n) =

d
1
d
2
=n
f(d
1
)f(d
2
).
Note that f g is also arithmetic, and that the product is commutative, and associative:
(f g) h(n) = f (g h)(n) = f g h(n) =

d
1
d
2
d
3
=n
f(d
1
)f(d
2
)f(d
3
).
With respect to the Dirichlet product, the identity element e A is easy to nd:
e(n) =
_
1 if n = 1
0 if n > 1.
Indeed, check that e f = f e = f for any f A. If we were working with the usual product of
functions f g(n) = f(n)g(n), then the identity element would have been
I(n) = 1 for all n N,
2
because I f = f for all functions f. In the set A, however, I is certainly not the identity element,
but has the nice property of transforming each function f into its so-called sum-function S
f
. Dene
S
f
(n) =

d|n
f(d),
to be the sum-function of f A, and note that S
f
is also arithmetic. Check that:
I f = f I = S
f
for all f A.
It is interesting to nd the Dirichlet inverse g A of this simple function I in our set A, i.e. such
that g I = I g = e. This naturally leads to the introduction of the so-called Mobius function :
Denition 2. The Mobius function : N C is dened by
(n) =
_
_
_
1 if n = 1
0 if n is not square-free
(1)
r
if n = p
1
p
2
p
r
, p
j
distinct primes.
Lemma 3. The Dirichlet inverse of I is the Mobius function A.
Proof: The lemma means that I = e, i.e.
(1)

d|n
(d) =
_
1 if n = 1
0 if n > 1.
This follows easily from the denition of . Indeed, for n = p

1
1
p

2
2
p

r
r
> 1 we have

d|n
(d) =

d|n,dsq.free
(d) = (1) +
r

k=1

1i
1
<<i
k
r
(p
i
1
p
i
k
).
Here the last sum runs over all square-free divisors of n. For combinatorial reasons,

d|n
(d) =
r

k=0
_
r
k
_
(1)
k
= (1 1)
r
= 0.
To summarize, we have shown that the Dirichlet inverse of the function I(n) is the Mobius
function (n): I = I = e. Unfortunately, not all arithmetic functions have Dirichlet inverses
in A. In fact, show that
Lemma 4. An arithmetic function f has a Dirichlet inverse in M i f(1) = 0.
2
The right notion to replace inverses in A turns out to be sum-functions, which is the idea
of the Mobius inversion theorem.
Theorem 2 (Mobius inversion theorem). Any arithmetic function f(n) can be expressed in terms
of its sum-function S
f
(n) =

d|n
f(d) as
f(n) =

d|n
(d)S
f
(
n
d
).
2
For those who care about group theory interpretations, this implies that M is not a group with the Dirichlet
product, but the subset M

of it consisting of all arithmetic functions f with f(1) = 0 is a group.

3
Proof: The statement is nothing else but the Dirichlet product f = S
f
in A:
S
f
= (I f) = ( I) f = e f = f.
Note: Here is a more traditional proof of the Mobius Inversion Formula:

d|n
(d)S
f
(
n
d
) =

d|n
(
n
d
)S
f
(d) =

d|n
(
n
d
)

d
1
|d
f(d
1
)
=

d
1
|n
f(d
1
)

d
1
|d|n
(
n
d
) =

d
1
|n
f(d
1
)

d
2
|m
(
m
d
2
),
where m = n/d
1
, d
2
= d/d
1
. By property (1), the second sum is non-zero only when m = 1, i.e.
d
1
= n, and hence the whole expression equals f(n).
This proof, however, hides the product structure of A under the Dirichlet product. It is natural
to asks the opposite question: given the Mobius relation f(n) =

d|n
(d)g(
n
d
) for two arithmetic
functions f and g, can we deduce that g is the sum-function S
f
of f? The answer should be obvious
from the product structure of A:
f = g I f = I ( g) S
f
= (I )g = e g = g,
and indeed, g is the sum-function of f. If you prefer more traditional proofs, you can recover one
from the above by recalling that multiplying f by I is the same as taking the sum-function of f.
Corollary 2. For two arithmetic functions f and g we have the following equivalence:
g(n) =

d|n
f(d) f(n) =

d|n
(d)g(
n
d
).
This incidentally shows that every arithmetic function g is the sum-function of another arithmetic
function f: simply dene f by the second formula as f = g.
Notice that Theorem 1 does not use the Dirichlet product at all, but instead it hides a more
general fact. For an arithmetic function f, its sum-function is S
f
= I f, and by Mobius inversion,
f = S
f
. Thus, Theorem 1 simply proves that if f M, then the product of the two multi-
plicative functions I and f is also multiplicative, and that if S
f
M, then the product of the two
multiplicative functions S
f
and is also multiplicative. This naturally suggest the more general
Theorem 3. The set M of multiplicative functions is closed under the Dirichlet product: f, g
M f g M.
Proof: Let f, g M, (a, b) = 1, and h = f g. Then
h(a)h(b) =
_
f g(a)

_
f g(b)

d
1
|a
f(d
1
)g(
a
d
1
)

d
2
|b
f(d
2
)g(
b
d
2
)
=

d
1
|a,d
2
|b
f(d
1
)f(d
2
)g(
a
d
1
)g(
b
d
2
) =

d
1
|a,d
2
|b
f(d
1
d
2
)g(
ab
d
1
d
2
)
=

d|ab
f(d)g(
ab
d
) = f g(ab) = h(ab).
Thus, h is also multiplicative, so that f g M.
For the group theory fans: nd the explicit Dirichlet inverse of any multiplicative function f 0,
and conclude
4
Lemma 5. The set M{f 0} is a group under the Dirichlet product.
3. Warm-up Problems
Problem 1. Find m, n N such that they have no prime divisors other than 2 and 3, (m, n) = 18,
(m) = 21, and (n) = 10.
Problem 2. Find n N such that one of the following is satised: (n) = 2
3
3
6
, (n) = 3
30
5
40
,
(n) = 13 31, or (n) = 13 31.
Problem 3. Dene
k
(n) =

d|n
d
k
. Thus,
0
(n) = (n) and
1
(n) = (n). Prove that
k
(n) is
multiplicative for all k N, and nd a formula for it.
Problem 4. Show that (n) is odd i n is a perfect square, and that (n) is odd i n is a perfect
square or twice a perfect square.
Problem 5. If f(n) is multiplicative, f 0, then show

d|n
(d)f(d) =
r

i=1
_
1 f(p
i
)
_
.
Problem 6. If f(n) is multiplicative, then show that h(n) =

d|n
(
n
d
)f(d) is also multiplicative.
Conclude that every multiplicative function is the sum-function of another multiplicative function.
4. The Euler Function (n)
Denition 4. The Euler function (n) assigns to each natural number n the number of the integers
d between 1 and n which are relatively prime to n (set (1) = 1.)
Lemma 6. (n) is multiplicative.
Proof: Consider the table of all integers between 1 and ab, where a, b N, (a, b) = 1.
1 2 i a 1 a
a + 1 a + 2 a + i 2a 1 2a
2a + 1 2a + 2 2a + i 3a 1 3a

ja + 1 ja + 2 ja + i (j + 1)a 1 (j + 1)a

(b 2)a + 1 (b 2)a + 2 (b 2)a + i (b 1)a 1 (b 1)a
(b 1)a + 1 (b 1)a + 2 (b 1)a + i ba 1 ba
Note that if ja + i is relatively prime with a, then all numbers in its (i-th) column will be
relatively prime with a. In the rst row there are exactly (a) numbers relatively prime with a, so
their (a) columns will be all numbers in the table which are relatively prime with a.
As for b, each column is a system of remainders modulo b because (a, b) = 1 (check that in
each column the b integers are distinct modulo b.) Thus, in each column we have precisely (b)
relatively prime integers to b. Hence, the total number of elements in this table relatively prime to
ab is exactly (ab) = (a)(b), and (n) is multiplicative.
5
We can use multiplicativity of the Euler function to derive a formula for (n). For a prime p,
all numbers d [1, p
k
] such that (d, p
k
) = 1 are exactly those d divisible by p: d = p c with
c = 1, 2, ..., p
k1
, i.e. p
k1
in number. Hence (p
k
) = p
k
p
k1
. Therefore,
(n) =
r

i=1
(p

i
i
) =
r

i=1
(p

i
i
p

i
1
i
) =
r

i=1
p

i
i
(1
1
p
i
) = n(1
1
p
1
) (1
1
p
r
).
Lemma 7. The sum-function S

(n) of the Euler function (n) satises:

d|n
(n) = n.
Proof: The multiplicativity of the Euler function (n) implies that S

(n) is multiplicative, so that

S

(n) = S

(p

1
1
) S

(p

r
r
). This reduces the problem to a prime power n = p
a
, which together
with (p
j
) = p
j
p
j1
easily implies S

(p
a
) = p
a
.
Remark: We can prove the lemma also directly by considering the set of fractions
_
1
n
,
2
n
, ...,
n 1
n
,
n
n
_
=
_
a
1
b
1
,
a
2
b
2
, ...,
a
n1
b
n1
,
a
n
b
n
_
where (a
i
, b
i
) = 1. The set D of all denominators {b
i
} is precisely the set of all divisors of n. Given
a divisor d of n, d appears in the set D exactly (d) times:
a
i
b
i
=
a
i
d
=
a
i
n
d
n
,
where (a
i
, d) = 1 and 1 a
i
d. Counting the elements in D in two dierent ways implies
n =

d|n
(d).
We can use this Remark to show again multiplicativity of the Euler function: indeed, this follows
from the fact that its sum-function S

(n) = n is itself obviously multiplicative.

5. Warm-up Problems
Problem 7. Show that (n
k
) = n
k1
(n) for all n, k N.
Problem 8. Solve the following equations:
(2
x
5
y
) = 80.
(n) = 12.
(n) = 2n/3.
(n) = n/2.
((n)) = 2
13
3
3
.
Problem 9. Show that
(n)(m) = ((n, m))([n, m]);
(nm)((n, m)) = (n, m)(n)(m).
6
6. Applications to Problems
Problem 10. For two sequences of complex numbers {a
0
, a
1
, ..., a
n
, ...} and {b
0
, b
1
, ..., b
n
, ...} show
that the following relations are equivalent:
a
n
=
n

k=0
b
k
for all n b
n
=
n

k=0
(1)
k+n
a
k
for all n.
Problem 11. Solve the equation ((2
n
)) = 2
n
.
Problem 12. Let f(x) Z[x] and let (n) be the number of values f(j), j = 1, 2, ..., n, such that
(f(j), n) = 1. Show that (n) is multiplicative and that (p
t
) = p
t1
(p). Conclude that
(n) =

p|n
(p)/p.
Problem 13. Find closed expressions for the following sums:

d|n
(d)(d)

d|n
(d)(d)

d|n
(d)(d)

d|n

2
(d)
2
(d)

d|n
(d)(
n
d
)

d|n
(d)
(d)

d|n
(d)(
n
d
)

d|n
(
n
d
)ln d

d|n
(d)
d

(t,n)=1
1t<n
t
Problem 14. Consider the function (s) =

n=1
1
n
s
, the so-called Riemann zeta-function. It con-
verges for s > 1. In fact, one can extend it to an analytic function over the whole complex plane
except s = 1. The famous Riemann Conjecture claims that all zeros of (s) in the strip 0 Re s 1
lie on the line Re s = 1/2. For (s) show the following formal identities:
(s) =

p
1
1 p
s
(s)
2
=

n=1
(n)
n
s
(s)
1
=

n=1
(n)
n
s
(s)(s 1) =

n=1
(n)
n
s
Problem 15. Suppose that we are given innitely many tickets, each with one natural number
on it. For any n N, the number of tickets on which divisors of n are written is exactly n. For
example, the divisors of 6, {1, 2, 3, 6}, are written in some variation on 6 tickets, and no other ticket
has these numbers written on it. Prove that any number n N is written on at least one ticket.
Problem 16. Let f(n) : N N be multiplicative and strictly increasing. If f(2) = 2, then
f(n) = n for all n.
7
7. Problems with [x] and Multiplicative Functions
Problem 17. Prove that
n

k=1
(k) =
n

k=1
_
n
k
_
and
n

k=1
(k) =
n

k=1
k
_
n
k
_
.
Problem 18. Prove that for (p, q) = 1:
q1

k=1
_
kp
q
_
=
p1

k=1
_
kq
p
_
.
Problem 19. Prove that for n N:

1k<n
_
(k + 1)
k
_
= n 1
n

k=1
__
(k 1)! + 1
k
__
.
Problem 20. Let S(m, n) =
_
k Z

m(mod k) + n(mod k) k
_
. Find

kS(m,n)
(k).
8. Hints and Solutions to Selected Problems
Hint 10. Dene a function a : N C by
a(n) =
_
_
_
a
0
if n = 1
a
r
if n = p
1
p
2
p
r
0 if n not sq.free.
Solution 11. We have (2
n+1
1) = 2
n
. If 2
n+1
1 = 1, then n = 0 and (1) = 1. Otherwise,
2
n+1
1 = p

1
1
p

2
2
p

r
r
> 1 with all p
i
s odd. From a previous formula,
(p

1
1
p

2
2
p

r
r
) = 2
n
= p

1
1
1
p

2
1
2
p

r
1
r
(p
1
1)(p
2
1) (p
r
1).
Therefore, all
i
= 1, 2
n
= (p
1
1)(p
2
1) (p
r
1), and all p
i
= 2
s
i
+ 1 for some s
i
1. It is
easy to see that if s
i
has an odd divisor > 1, then 2
s
i
+ 1 will factor, making p
i
non-prime. Hence
p
i
= 2
2
q
i
+ 1 for some q
i
0.
We have reduced the problem to nding all sets of primes p
i
of the above type such that
p
1
p
2
p
r
+ 1 = 2
n+1
= 2(p
1
1)(p
2
1) (p
r
1).
Order the primes p
1
< p
2
< < p
r
. Check consecutively via modulo appropriate powers of 2,
that, if they exist, the smallest primes must be: p
1
= 3, p
2
= 5, p
3
= 17, p
4
= 257. However, if p
5
also exists, then p
5
= 2
2
4
+ 1, which is not prime. Hence, r 4, and n = 1, 2, 3, 4.
Hint 12. Modify the table proof for multiplicativity of the Euler function (n).
Hint 13, lnd. Note that lnd is not multiplicative. Consider the function (n) dened by:
(n) = ln p if n is a power of a prime p, and (n) = 0 otherwise, and use Mobius inversion.
Solution 16. We have f(1) = 1 and f(2) = 2. Let f(3) = 3 + m for m N {0}.
f(6) = f(2) f(3) = 6 + 2m f(5) 5 + 2m f(10) 10 + 4m
f(9) 9 + 4m f(18) 18 + 8m f(15) 15 + 8m.
But f(15) = f(3) f(5) (3 + m) (5 + m) = 15 + 8m + m
2
. Hence, m
2
0, i.e. m = 0 and
f(3) = 3.
8
Now assume by induction on k that f(s) = s for s = 1, 2, ..., 2k 1, where k 2. Then
f(4k 2) = 2f(2k 1) = 4k 2. Since f(2k 1) = 2k 1 and the function is strictly increasing,
f(s) = s for all s [2k1, 4k2]. In particular, f(2k) = 2k and f(2k+1) = 2k+1 (4k2 > 2k+1).
This completes the induction step.
Solution 17a. Set B
n,k
=
_
n
k
_

_
n 1
k
_
for k = 1, 2, ..., n. Then B
n,k
=
_
1 when k |n,
0 when k|n,
and
(n) =
n

k=1
B
n,k
. Summing up, we obtain
n

k=1
(k) =
n

k=1
k

m=1
B
m,k
=
n

k=1
k

m=1
__
k
m
_

_
k 1
m
__
=
=
n

k,m=1
_
k
m
_

1mn
1kn1
_
k
m
_
=

1mn
k=n
_
k
m
_
.
Solution 17b. From the previous problem, kB
n,k
=
_
k when k |n,
0 when k|n,
and (n) =
n

k=1
kB
n,k
.

k=1
(k) =
n

k=1
k

m=1
mB
m,k
=
n

k=1
k

m=1
_
m
_
k
m
_
m
_
k 1
m
__
=
=
n

k,m=1
m
_
k
m
_

1mn
1kn1
m
_
k
m
_
=

1mn
k=n
m
_
k
m
_
.
Hint 18. Use the identity
_
kp
q
_

_
(q k)p
q
_
= p 1 for k = 1, 2, ..., q 1, and show that both
sums equal
(p 1)(q 1)
2
.
Hint 19. Both sides count the number of primes p n.
Hint 20. The given sum equals

k
__
m + n
k
_

_
m
k
_

_
n
k
_
_
(k) = (m + n) (m) (n),
where (l) =

k
_
l
k
_
(k). Further,
(l) =

1ql

d|q
(d) =
_
l + 1
2
_
.
9