On certain trigonometrical sums and their applications in the

theory of numbers

Transactions of the Cambridge Philosophical Society, XXII, No.13, 1918, 259 276

1. The trigonometrical sums with which this paper is concerned are of the type
X 2n
cs (n) = cos ,
s

where is prime to s and not greater than s. It is plain that

X
cs (n) = n ,

where is a primitive root of the equation

xs 1 = 0.

These sums are obviously of very great interest, and a few of their properties have been
discussed already . But, so far as I know, they have never been considered from the point
of view which I adopt in this paper; and I believe that all the results which it contains are
new.
My principal object is to obtain expressions for a variety of well-known arithmetical func-
tions of n in the form of a series X
as cs (n).
s

A typical formula is

2n
 
c1 (n) c2 (n) c3 (n)
(n) = + 2 + 2 + ,
6 12 2 3

where (n) is the sum of the divisors of n. I give two distinct methods for the proof of this
and a large variety of similar formul. The majority of my formul are elementary in
the technical sense of the word they can (that is to say) be proved by a combination
of processes involving only finite algebra and simple general theorems concerning infinite
series. These are however some which are of a deeper character, and can only be proved
by means of theorems which seem to depend essentially on the theory of analytic functions.
A typical formula of this class is

c1 (n) + 12 c2 (n) + 13 c3 (n) + = 0,

See, e.g., Dirichlet-Dedekind, Vorlesungen uber Zahlentheorie, ed. 4, Supplement VII, pp. 360 370.
230 Paper 21

a formula which depends upon, and is indeed substantially equivalent to, the Prime Num-
ber Theorem of Hadamard and de la Vallee Poussin.
Many of my formul are intimately connected with those of my previous paper On certain
arithmetical functions, published in 1916 in these Transactions . They are also connected
(in a manner pointed out in 15) with a joint paper by Mr Hardy and myself, Asymptotic
Formul in Combinatory Analysis, in course of publication in the Proceedings of the
London Mathematical Society .

2. Let F (u, v) be any function of u and v, and let

X
(2.1) D(n) = F (, ),

where is a divisor of n and = n. For instance

D(1) = F (1, 1); D(2) = F (1, 2) + F (2, 1);
D(3) = F (1, 3) + F (3, 1); D(4) = F (1, 4) + F (2, 2) + F (4, 1);
D(5) = F (1, 5) + F (5, 1); D(6) = F (1, 6) + F (2, 3) + F (3, 2) + F (6, 1); . . . . . .
It is clear that D(n) may also be expressed in the form
X
(2.2) D(n) = F ( , ).

Suppose now that

s1
X 2n
(2.3) s (n) = cos ,
0
s

so that s (n) = s if s is a divisor of n and s (n) = 0 otherwise. Then

t
X 1  n
(2.4) D(n) = (n)F , ,
1

where t is any number not less than n. Now let

X 2n
(2.5) cs (n) = cos ,
s

where is prime to s and does not exceed s; e.g.

c1 (n) = 1; c2 (n) = cos n; c3 (n) = 2 cos 32 n;

[No.18 of this volume].

[No. 36 of this volume].
Xt X
[t]

is to be understood as meaning , where [t] denotes as usual the greatest integer in t.
1 1
On certain trigonometrical sums and their applications in the theory of numbers 231

c4 (n) = 2 cos 21 n; c5 (n) = 2 cos 25 n + 2 cos 54 n;

c6 (n) = 2 cos 31 n; c7 (n) = 2 cos 27 n + 2 cos 74 n + 2 cos 76 n;
c8 (n) = 2 cos 41 n + 2 cos 34 n; c9 (n) = 2 cos 92 n + 2 cos 94 n + 2 cos 89 n;
c10 (n) = 2 cos 51 n + 2 cos 35 n; . . . .

It follows from (2.3) and (2.5) that

X
(2.6) s (n) = c (n),

where is a divisor of s; and hence that

X
(2.7) cs (n) = ( ) (n),

where is a divisor of s, = s, and

X () 1
(2.8) = ,
s (s)

(s) being the Riemann Zeta-function. In particular

c1 (n) = 1 (n); c2 (n) = 2 (n) 1 (n); c3 (n) = 3 (n) 1 (n);

c4 (n) = 4 (n) 2 (n); c5 (n) = 5 (n) 1 (n); . . . . . .

But from (2.3) we know that (n) = 0 if is not a divisor of n; and so we can suppose
that, in (2.7), is a common divisor of n and s. It follows that
X
|cs (n)| ,

where is a divisor of n; so that

(2.9) c (n) = O(1)

if n is fixed and . Since

s (n) = s (n + s); cs (n) = cs (n + s),

the values of cs (n) for n = 1, 2, 3, . . . can be shewn conveniently by writing

c1 (n) = 1; c2 (n) = 1, 1; c3 (n) = 1, 1, 2;

c4 (n) = 0, 2, 0, 2; c5 (n) = 1, 1, 1, 1, 4;

See Landau, Handbuch der Lehre von der Verteilung der Primzahlen, p. 577.
232 Paper 21

c6 (n) = 1, 1, 2, 1, 1, 2 : c7 (n) = 1, 1, 1, 1, 1, 1, 6;
c8 (n) = 0, 0, 0, 4, 0, 0, 0, 4; c9 (n) = 0, 0, 3, 0, 0, 3, 0, 0, 6;
c10 (n) = 1, 1, 1, 1, 4, 1, 1, 1, 1, 4; . . . . . .

the meaning of the third formula, for example, being that c3 (1) = 1, c3 (2) = 1, c3 (3) = 2,
and that these values are then repeated periodically.
It is plain that we have also

(2.91) c (n) = O(1).

when is fixed and n .

3. Substituting (2.6) in (2.4) and collecting the coefficients of c1 (n), c2 (n), c3 (n) . . . , we
find that

(3.1)
1 1
t 2t 3t
X 1  n X 1  n X 1  n
D(n) = c1 (n) F , + c2 (n) F 2, + c3 (n) F 3, + ,
2 2 3 3
1 1 1

where t is any number not less than n. If we use (2.2) instead of (2.1) we obtain another
expression, viz.

(3.2)
1 1
t 2t 3t
X 1 n  X 1 n  X 1 n 
D(n) = c1 (n) F , + c2 (n) F , 2 , +c3 (n) F , 3 + ,
1
1
2 2 1
3 3

where t is any number not less that n.

Suppose now that

F1 (u, v) = F (u, v) log u, F2 (u, v) = F (u, v) log v.

Then we have
X X
D(n) log n = F (, ) log n = F (, ) log( )

X X

= F1 (, ) + F2 (, ),

where is a divisor of n and = n.

On certain trigonometrical sums and their applications in the theory of numbers 233

Now for F1 (, ) we shall write the expression corresponding to (3.1) and for F2 (, )
P P

the expression corresponding to (3.2). Then we have

1
r 2r
X log  n X log 2  n
(3.3) D(n) log n = c1 (n) F , + c2 (n) F 2,
2 2
1 1
1
3 r t
X log 3  n X log  n 
+ c3 (n) F 3, + + c1 (n) F ,
1
3 3 1

1 1
2t 3t
X log 2  n  X log 3  n 
+ c2 (n) F , 2 + c3 (n) F , 3 + ,
2 2 3 3
1 1

where r and t are any two numbers not less than n. If, in particular, F (u, v) = F (v, u),
then (3.3) reduces to
t
1
X log  n
(3.4) 2 D(n) log n = c1 (n) F ,
1

1 1
2t 3t
X log 2  n X log 3  n
+ c2 (n) F 2, + c3 (n) F 3, + ,
2 2 3 3
1 1

where t is any number not less than n.

4. We may also write D(n) in the form

u
X r
X
(4.1) D(n) = F (, ) + F ( , ),
=1 =1

where is a divisor of n, = n, and u, v are may two positive numbers such that uv = n,
it being understood that, if u and v are both integral, a term F (u, v) is to be subtracted
from the right-hand side. Hence (with the same conventions)
u v
X 1  n X 1 n 
D(n) = (n)F , + (n)F , ,

1 1

Applying to this formula transformations similar to those of 3, we obtain

1
u 2u
X 1  n X 1  n
(4.2) D(n) = c1 (n) F , + c2 (n) F 2, +
1
1
2 2
1
v 2v
X 1 n  X 1 n 
+ c1 (n) F , + c2 (n) F , 2 + ,
2 2
1 1
234 Paper 21

where u and v are positive numbers such that uv = n. If u and v are integers then a term
F (u, v) should be subtracted from the right-hand side.
If we suppose that 0 < u 1 then (4.2) reduces to (3.2), and if 0 < v 1 it reduces to
(3.1). Another particular case of interest is that in which u = v. Then

n
X 1 n  n  n o
(4.3) D(n) = c1 (n) F , +F ,

1
1
2Xn
1 n  n n o
+ c2 (n) F 2, +F , 2 + .
1
2 2 2

If n is a perfect square then F ( n, n) should be subtracted from the right hand side.

5. We shall now consider some special forms of these general equations. Suppose that
F (u, v) = v s , so that D(n) is the sum s (n) of the sth powers of the divisors of n. Then
from (3.1) and (3.2) we have
1 1
t 2 t 3 t
s (n) X 1 X 1 X 1
(5.1) = c1 (n) + c2 (n) + c3 (n) + ,
ns 1
s+1 1
(2)s+1 1
(3)s+1

1 1
t 2 t 3 t
X X X
(5.2) s (n) = c1 (n) s1 + c2 (n) (2)s1 + c3 (n) (3)s1 + ,
1 1 1

where t is any number not less than n: from (3.3)

1
r 2 r
X X
(5.3) s (n) log n = c1 (n) s1 log + c2 (n) (2)s1 log 2 +
1 1
1

t 2 t
s
X log X log 2
+ n c1 (n) + c2 (n) + ,

1
s+1 1
(2)s+1

where r and t are two numbers not less than n: and from (4.2)
1 1
u 2 u 3 u
X X X
s1
(5.4) s (n) = c1 (n) + c2 (n) (2)s1 + c3 (n) (3)s1 +
1 1 1
1 1

v 2v 3v
X 1 X 1 X 1
+ ns c1 (n) + c2 (n) + c3 (n) + ,
s+1 (2)s+1 (3)s+1
1 1 1
On certain trigonometrical sums and their applications in the theory of numbers 235

where uv = n. If u and v are integers then us should be subtracted from the right-hand
side.
Let d(n) = 0 (n) denote the number of divisors of n and (n) = 1 (n) the sum of the
divisors of n. Then from (5.1) (5.4) we obtain
1 1
t 2 3 t t
X 1 X 1 X 1
(5.5) d(n) = c1 (n) + c2 (n) + c3 (n) + ,
1
1
2 1
3

(5.6) (n) = c1 (n)[t] + c2 (n)[ 12 t] + c3 (n)[ 31 t] + ,

1 1
t 2 t 3 t
1
X log X log 2 X log 3
(5.7) 2 d(n) log n = c1 (n) + c2 (n) + c3 (n) + ,
1
1
2 1
3
1 1

( u v
) 2 u 2 v
X 1 X 1 X 1 X 1
(5.8) d(n) = c1 (n) + + c2 (n) +
2 2
1 1 1 1
1 1

3u 3v
X 1 X 1
+ c3 (n) + +
1 3

1
3

where t n and uv = n. If u and v are integers then 1 should be subtracted from the

right-hand side of (5.8). Putting u = v = n in (5.8) we obtain
1 1
n 2X 3X n n
1
X 1 1 1
(5.9) 2 d(n) = c1 (n) + c2 (n) + c3 (n) + ,
1
1
2 1
3

unless n is a perfect square, when 12 should be subtracted from the right-hand side. It may
be interesting to note that, if we replace the left-hand side in (5.9) by
1 1 
2 + 2 d(n) ,

then the formula is true without exception.

6. So far our work has been based on elementary formal transformations, and no questions
of convergence have arisen. We shall now consider the equation (5.1) more carefully. Let
us suppose that s > 0. Then
t/k    
X 1 X 1 1 1 1
s+1
= +O = (s + 1) + O .
(k) (k)s+1 kts ks+1 kts
1 1
236 Paper 21

The number of terms in the right-hand side of (5.1) is [t]. Also we know that c (n) = O(1)
as n . Hence
t t
( )  
s (n) X c (n) 1 X1 X c (n) log t
= (s + 1) +O s = (s + 1) +O .
ns s+1 t s+1 ts
=1 =1 1

Making t , we obtain
 
s c1 (n) c2 (n) c3 (n)
(6.1) s (n) = n (s + 1) + s+1 + s+1 + ,
1s+1 2 3

if s > 0. Similarly, if we make t in (5.3), we obtain

1
r 2 r
X X
s1
s (n) log n = c1 (n) log + c2 (n) (2)s1 log 2 +
1 1

( )
X log X log 2
+ns c1 (n) + c2 (n) + .
1
s+1 1
(2)s+1

But

X log k log k 1
s+1
= s+1 (s + 1) s+1 (s + 1).
(k) k k
1

It follows from this and (6.1) that

t
(s + 1)
  X
(6.2) s (n) + log n = c1 (n) s1 log
(s + 1) 1
1
2t
X
+ c2 (n) (2)s1 log 2 +
1
 
s c1 (n) log 1 c2 (n) log 2 c3 (n) log 3
+ n (s + 1) + + + ,
1s+1 2s+1 3s+1

where s > 0 and t n. Putting s = 1 in (6.1) and (6.2) we obtain

2
 
c1 (n) c2 (n) c3 (n)
(6.3) (n) = n + 2 + 2 + ,
6 12 2 3

(2) 2
   
c1 (n) c2 (n)
(6.4) (n) + log n = n 2
log 1 + 2 log 2 +
(2) 6 1 2
+ c1 (n)[t] log 1 + c2 (n)[ 12 t] log 2 +
+ c1 (n) log[t]! + c2 (n) log[ 21 t]! + ,
On certain trigonometrical sums and their applications in the theory of numbers 237

where t n.

7. Since

(7.1) s (n) = ns s (n),

we may write (6.1) in the form

s (n) c1 (n) c2 (n) c3 (n)

(7.2) = s+1 + s+1 + s+1 + ,
(s + 1) 1 2 3

where s > 0. This result has been proved by purely elementary methods. But in order
to know whether the right-hand side of (7.2) is convergent or not for values of s less than
or equal to zero we require the help of theorems which have only been established by
transcendental methods.
Now the right-hand side of (7.2) is an ordinary Dirichlets series for
1
s (n) .
(s + 1)

The first factor is a finite Dirichlets series and so an absolutely convergent Dirichlets series.
It follows that the right-hand side of (7.2) is convergent whenever the Dirichlets series for
1/(s + 1), viz.
X (n)
(7.3) ,
n1+s
is convergent. But it is known that the series (7.3) is convergent when s = 0 and that its
sum is 0.
It follows from this that

(7.4) c1 (n) + 12 c2 (n) + 13 c3 (n) + = 0.

Nothing is known about the convergence of (7.3) when 12 < s < 0. But with the assump-
tion of the truth of the hitherto unproved Riemann hypothesis it has been proved that
(7.3) is convergent when s > 12 . With this assumption we see that (7.2) is true when
s > 21 . In other words, if 12 < s < 12 , then
 
c1 (n) c2 (n) c3 (n)
(7.5) s (n) = (1 s) + 1s + 1s +
11s 2 3
 
s c1 (n) c2 (n) c3 (n)
= n (1 + s) + 1+s + 1+s + .
11+s 2 3

Landau, Handbuch, p. 591.

Littlewood, Comptes Rendus, 29 Jan. 1912.
238 Paper 21

8. It is known that all the series obtained from (7.3) by term-by-term differentiation with
respect to s are convergent when s = 0, and it is obvious that the derivatives of s (n)
with respect to s are all finite Dirichlets series and so absolutely convergent. It follows
that all the derivatives of the right-hand side of (7.2) are convergent when s = 0; and so
we can equate the coefficients of like powers of s from the two sides of (7.2). Now
1
(8.1) = s s2 + ,
(s + 1)
where is Eulers constant. And
X X X
s (n) = s = 1s log + ,

where is a divisor of n. But

X X X
1
log = log = 2 log( ) = 12 d(n) log n,

where = n. Hence

(8.2) s (n) = d(n) 21 sd(n) log n + .

Now equating the coefficients of s and s2 from the two sides of (7.2) and using (8.1) and
(8.2), we obtain

(8.3) c1 (n) log 1 + 12 c2 (n) log 2 + 31 c3 (n) log 3 + = d(n),

(8.4) c1 (n)(log 1)2 + 12 c2 (n)(log 2)2 + 13 c3 (n)(log 3)2 + = d(n)(2 + log n).

9. I shall now find an expression of the same kind for (n), the number of numbers prime
to and not exceeding n. Let p1 , p2 , p3 , . . . be the prime divisors of n, and let

(9.1) s (n) = ns (1 ps s s
1 )(1 p2 )(1 p3 ) ,

so that 1 (n) = (n). Suppose that

F (u.v) = (u)v s .

Then it is easy to see that

D(n) = s (n).
Hence, from (3.1), we have
1
t 2 t
s (n) X () X (2)
(9.2) s
= c1 (n) s+1
+ c2 (n) + ,
n 1
1
(2)s+1

Landau, Handbuch, p.594
On certain trigonometrical sums and their applications in the theory of numbers 239

where t is any number not less than n. If s > 0 we can make t , as in 6. Then we
have

s (n) X () X (2)
(9.3) s
= c1 (n) s+1
+ c2 (n) + .
n (2)s+1
1 1

But it can easily be shewn that

X (n) (n)
(9.4) = ,
1
s (s)(1 ps
1 )(1 ps s
2 )(1 p3 )

where p1 , p2 , p3 , . . . are the prime divisors of n. In other words

X (n) (n)ns
(9.5) = .
s s (n)(s)
1

It follows from (9.3) and (9.5) that

s (n)(s + 1) (1)c1 (n) (2)c2 (n) (3)c3 (n)

(9.6) s
= + + + .
n s+1 (1) s+1 (2) s+1 (3)

In particular

2 c2 (n) c3 (n) c5 (n)

(9.7) (n) = c1 (n) 2 2 2
6n 2 1 3 1 5 1
c6 (n) c7 (n) c10 (n)
+ 2 2 + 2 .
(2 1)(3 1) 7 1 (2 1)(52 1)
2

10. I shall now consider an application of the main formul to the problem of the number
of representations of a number as the sum of 2, 4, 6, 8, . . . squares. We require the following
preliminary results.
(1) Let
X 1s1 x 2s1 x2 3s1 x3
(10.1) D(n)xn = X1 = + + + .
1+x 1 x2 1 + x3
We shall choose
F (u, v) = v s1 , u 1( mod 2),
F (u, v) = v s1 , u 2( mod 4),
F (u, v) = (2s 1)v s1 , u 0( mod 4).
240 Paper 21

Then from (3.1) we can shew, by arguments similar to those used in 6, that

D(n) = ns1 (1s + 3s + 5s + ){1s c1 (n) + 2s c4 (n) + 3s c3 (n)

(10.11)
+4s c8 (n)5s c5 (n) + 6s c12 (n) + 7s c7 (n) + 8s c16 (n) + }.

if s > 1.
(2) Let
X 1s1 x 2s1 x2 3s1 x3
(10.2) D(n)xn = X2 = + + + .
1x 1 + x2 1 x3
We shall choose
F (u, v) = v s1 , u 1( mod 2),
F (u, v) = v s1 , u 2( mod 4),
F (u, v) = (1 2s )v s1 , u 0( mod 4).
Then we obtain as before
D(n) = ns1 (1s + 3s + 5s + ){1s c1 (n) 2s c4 (n) + 3s c3 (n)
(10.21)
4s c8 (n) + 5s c5 (n) 6s c12 (n) + 7s c7 (n) 8s c16 (n) + }.

(3) Let
X 1s1 x 2s1 x2 3s1 x3
(10.3) D(n)xn = X3 = + + + .
1 + x2 1 + x4 1 + x6
We shall choose
F (u, v) = 0, u 0( mod 2),
F (u, v) = v s1 , u 1( mod 4),
F (u, v) = v s1 , u 3( mod 4).
Then we obtain as before

(10.31) D(n) = ns1 (1s 3s + 5s ){1s c1 (n) 3s c3 (n) + 5s c5 (n) }.

(4) We shall also require a similar formula for the function D(n) defined by
X 1s1 x 3s1 x3 5s1 x5
(10.4) D(n)xn = X4 = + .
1x 1 x3 1 x5
The formula required is not a direct consequence of the preceding analysis, but if, instead
of starting with the function
X 2n
cr (n) = cos ,
r

On certain trigonometrical sums and their applications in the theory of numbers 241

we start with the function

1 2n
(1) 2 (1) sin
X
sr (n) = ,
r

where is prime to r and does not exceed r, and proceed as in 23, we can shew that

(10.41) D(n) = 21 ns1 (1s 3s + 5s ){1s s4 (n) + 2s s8 (n) + 3s s12 (n) + }.

It should be observed that there is a correspondence between cr (n) and the ordinary -
function of the one hand and sr (n) and the function

(s) = 1s 3s + 5s

on the other. It is possible to define an infinity of systems of trigonometrical sums such as

cr (n), sr (n), each corresponding to one of the general class of L-functions of which (s)
and (s) are the simplest members.
We have shewn that (10.31) and (10.41) are true when s > 1. But if we assume that the
Dirichlets series for 1/(s) is convergent when s = 1, a result which is precisely of the
same depth as the prime number theorem and has only been established by transcendental
methods, then we can shew by arguments similar to those of 7 that (10.31) and (10.41)
are true when s = 1.

11. I have shewn elsewhere that if s is a positive integer and

X
1+ rs (n)xn = (1 + 2x + 2x4 + 2x9 + )s ,

then
r2s (n) = 2s (n) + e2s (n),
where e2s (n) = 0 when s = 1, 2, 3 or 4 and is of lower order than 2s (n) in all cases; that
if s is a multiple of 4 then
X s
(11.1) (1s + 3s + 5s + ) 2s (n)xn = X1 ;
(s 1)!
if s is of the form 4k + 2 then
X s
(11.2) (1s + 3s + 5s + ) 2s (n)xn = X2 ;
(s 1)!
if s is of the form 4k + 1 then
X s
(11.3) (1s 3s + 5s ) 2s (n)xn = (X3 + 21s X4 );
(s 1)!

See Landau, Handbuch, pp. 414 et seq.

Transactions of the Cambridge Philosophical Society Vol.XXII, 1916, pp. 159 184. [No.18 of this
volume; see in particular 24 28, pp. 202 208].

For a more precise result concerning the order of e2s (n) see 15.
242 Paper 21

except when s = 1; and if s is of the form 4k + 3 then

X s
(11.4) (1s 3s + 5s ) 2s (n)xn = (X3 21s X4 ),
(s 1)!

X1 , X2 , X3 , X4 being the same as in 10.

In the case in which s = 1 it is well known that

x3 x5
 
P n x
2 (n)x = 4 +
1 x 1 x3 1 x5
(11.5)
x2 x3
 
x
= 4 + + +
1 + x2 1 + x4 1 + x6

It follows from 10 that, if s is a multiple of 4 then

s ns1 s
2s (n) = {1 c1 (n) + 2s c4 (n) + 3s c3 (n) + 4s c8 (n)
(11.11) (s 1)!
+5s c5 (n) + 6s c12 (n) + 7s c7 (n) + 8s c16 (n) + };

if s is of the form 4k + 2 then

s ns1 s
2s (n) = {1 c1 (n) 2s c4 (n) + 3s c3 (n) 4s c8 (n)
(11.21) (s 1)!
+5s c5 (n) 6s c12 (n) + 7s c7 (n) 8s c16 (n) + };

if s is of the form 4k + 1 then

s ns1 s
2s (n) = {1 c1 (n) + 2s s4 (n) 3s c3 (n) + 4s s8 (n)
(11.31) (s 1)!
+5s c5 (n) + 6s s12 (n) 7s c7 (n) + 8s s16 (n) + };

except when s = 1; and if s is of the form 4k + 3 then

s ns1 s
2s (n) = {1 c1 (n) 2s s4 (n) 3s c3 (n) 4s s8 (n)
(11.41) (s 1)!
+5s c5 (n) 6s s12 (n) 7s c7 (n) 8s s16 (n) + };

From (11.5) and the remarks at the end of the previous section, it follows that

r2 (n) = 2 (n) = {c1 (n) 13 c3 (n) + 15 c5 (n) }

(11.51)
= { 21 s4 (n) + 41 s8 (n) + 61 s12 (n) + },

but this of course not such an elementary result as the preceding ones.
On certain trigonometrical sums and their applications in the theory of numbers 243

We can combine all the formul (11.11) (11.41) in one by writing

s ns1 s
2s (n) = {1 c1 (n) + 2s c4 (n) + 3s c3 (n) + 4s c8 (n)
(11.6) (s 1)!
+5s c5 (n) + 6s c12 (n) + 7s c7 (n) + 8s c16 (n) + },

where s is an integer greater than 1 and

cr (n) = cr (n) cos 12 s(r 1) sr (n) sin 21 s(r 1).

12. We can obtain analogous results concerning the number of representations of a number
as the sum of 2, 4, 6, 8, . . . triangular numbers. Equation (147) of my former paper is
equivalent to

X
(12.1) (1 2x + 2x4 2x9 + )2s = 1 + 2s (n)(x)n
1

f 4s (x) X f 24n (x2 )

+ Kn (x)n ,
f 2s (x2 ) f 24n (x)
1
1n 4 (s1)

where Kn is a constant and

f (x) = (1 x)(1 x2 )(1 x3 ) .

Suppose now that

x = e , x = e2/ .
Then we know that
1
(12.2) (1 2x + 2x4 2x9 + ) = 2x 8 (1 + x + x 3 + x 6 + ),

q
1 1 1 1
(12.3) ( 21 )x 24 f (x) = x 12 f (x 2 ), x 12 f (x2 ) = x 24 f (x ).
P
Finally 1 + 1 2s (n)(x)n can be expressed in powers of x by using the formula

12s1 22s1 32s1

 
1
s 2 (1 2s) + + + +
e2 1 e4 1 e6 1
(12.4)
12s1 22s1 32s1
 
1
= ()s 2 (1 2s) + 2 + + + ,
e 1 e4 1 e6 1

Loc. cit., p.181 [p. 204 of this volume].
244 Paper 21

where = 2 and s is an integer greater than 1; and

12s 22s 32s
 
1
s+ 2
(2) + + +
e + e e2 + e2 e3 + e3
(12.5)
12s 32s 52s
 
s 1
p
= () (2) 2 (2s) + + ,
e 1 e3 1 e5 1
where = 2 , s is any positive integer, and (s) is the function represented by the series
1s 3s + 5s and its analytical continuations.
It follows from all these formul that, if s is a positive integer and
X X X
(12.6) (1 + x + x3 + x6 + )2s =
r2s (n)xn =
2s (n)xn + e2s (n)xn ,

then
X f 4s (x2 ) X f 24n (x)
e2s (n)xn = Kn (x)n ,
f 2s (x) f 24n (x2 )
1
1n 4 (s1)

where Kn is a constant, and f (x) is the same as in (12.1);

(n)xn =
( 21 )s 1 s
(1s + 3s + 5s + )
P
2s x 4
(s 1)!
(12.61)  s1
2s1 x2 3s1 x3

1 x
+ + +
1 x2 1 x4 1 x6
if s is a multiple of 4;

(n)xn =
2( 14 )s 1 s
(1s + 3s + 5s + )
P
2s x 4
(s 1)!
(12.62) 1 3 5

s1 s1 s1
1 x + 3 x + 5 x +
2 2 2

1x 1 x3 1 x5

if s is of the form 4k + 2;
X 2( 18 )s 1 s
(12.63) (1s 3s + 5s )
2s (n)xn = x 4
(s 1)!
1 3 5 1 3 5
!
1s1 x 4 3s1 x 4 5s x 4 1s1 x 4 3s1 x 4 5s1 x 4
1 + 3 + 5 + + 1 3 + 5
1 + x2 1 + x2 1 + x2 1 x2 1 x2 1 x2
if s is of the form 4k + 1 (except when s = 1); and
X 2( 18 )s 1 s
(12.64) (1s 3s + 5s )
2s (n)xn = x 4
(s 1)!
1 3 5 1 3 5
!
1s1 x 4 3s1 x 4 5s x 4 1s1 x 4 3s1 x 4 5s x 4
1 + 3 + 5 + 1 + 3 5 +
1 + x2 1 + x2 1 + x2 1 x2 1 x2 1 x2
On certain trigonometrical sums and their applications in the theory of numbers 245

if s is of the form 4k + 3. In the case in which s = 1 we have

1 3 5
!
41 x4 x4 x4
2 (n)xn
P
= x 1 + 3 + 5 +
1 + x2 1 + x2 1 + x2
(12.65) 1 3 5
!
14 x4 x4 x4
= x 1 3 + 5 .
1 + x2 1 + x2 1 + x2

It is easy to see that the principal results proved about e2s (n) in my former paper are also
true of e2s (n), and in particular that

e2s (n) = 0

when s = 1, 2, 3 or 4, and

r2s (n) 2s (n)
for all values of s.

13. It follows from (12.62) that, if s is of the form 4k + 2, then

(1s + 3s + 5s + )2s

(n)

is the coefficient of xn in
1 3
!
2( 14 )s 1 s 1s1 x 2 2s1 x 3s1 x 2
(13.1) x 4 + + + .
(s 1)! 1x 1 x2 1 x3

Similarly from (12.63) and (12.64) it follows that, if s is an odd integer greater than 1, then
(n) is the coefficient of xn in
(1s 3s + 5s )2s
1 1 3
!
4( 18 )s 1 s 1s1 x 4 2s1 x 2 3s1 x 4
(13.2) x 4 1 + + 3 + .
(s 1)! 1 + x2 1+x 1 + x2

Now by applying our main formul to (12.61) and (13.1) and (13.2) we obtain:

( 12 )s
(n)
2s = (n + 41 s)s1
(13.3) (s 1)!
{1s c1 (n + 14 s) + 3s c3 (n + 14 s) + 5s c5 (n + 14 s) + }

if s is a multiple of 4;

( 21 )s
(n) =
2s (n + 14 s)s1
(13.4) (s 1)!
{1s c1 (2n + 12 s) + 3s c3 (2n + 21 s) + 5s c5 (2n + 12 s) + }
246 Paper 21

if s is twice an odd number; and

( 12 )s

2s (n) = (n + 41 s)s1
(13.5) (s 1)!
{1s c1 (4n + s) 3s c3 (4n + s) + 5s c5 (4n + s) }
if s is an odd number greater than 1.
Since the coefficient of xn in (1 + x + x3 + )2 is that of x4n+1 in
1 4

2
2 + x + x + ,

it follows from (11.51) that

(13.6) {c1 (4n + 1) 31 c3 (4n + 1) + 15 c5 (4n + 1) }.
r2 (n) = 2 (n) =
4
This result however depends on the fact that the Dirichlets series for 1/(s) is convergent
when s = 1.
(n) may be arrived at by another method.
14. The preceding formul for s (n), 2s (n), 2s
We understand by
sin n
(14.1)
k sin(n/k)
the limit of
sin x
k sin(x/k)
when x n. It is easy to see that, if n and k are positive integers, and k odd, then (14.1)
is equal to 1 if k is a divisor of n and to 0 otherwise.
When k is even we have (with similar conventions)
sin n
(14.2) = 1 or 0
k tan(n/k)
according as k is a divisor of n or not. It follows that
(   !
sin n sin n
s1 (n) = ns1 1s + 2s
sin n tan 12 n
(14.3) ! ! )
sin n sin n
+3s + 4s + .
sin 31 n tan 14 n
(n) we find that
Similarly from the definitions of 2s (n) and 2s
s ns1
  
s s s s s sin n
(14.4) {1 + (3) + 5 + (7) + }2s (n) = 1
(s 1)! sin n
! ! ! )
sin n sin n sin n
+2s +3s + 4s +
sin( 21 n + 12 s) sin( 13 n + s) sin( 14 n + 32 s)
On certain trigonometrical sums and their applications in the theory of numbers 247

if s is an integer greater than 1;

(  ! ! )
sin n 1 sin n 1 sin n
r2 (n) = 2 (n) = 4 3 + 5
sin n sin 13 n sin 51 n
(14.5) ( ! ! ! )
1 sin n 1 sin n 1 sin n
=4 2 4 + 6 ;
cos 12 n cos 14 n cos 16 n

( 21 )s
(1s + 3s + 5s + )2s
(n) = (n + 14 s)s1
(s 1)!
( ! !
sin(n + 41 s) sin(n + 41 s)
(14.6) 1s + 3s
sin(n + 41 s) sin 13 (n + 41 s)
! )
s sin(n + 14 s)
+5 +
sin 15 (n + 41 s)
if s is a multiple of 4;
( 12 )s
(1s + 3s + 5s + )2s
(n) = (n + 14 s)s1
(s 1)!
( ! !
s sin(2n + 12 s) sin(2n + 12 s)
(14.7) 1 + 3s
sin(2n + 12 s) sin 31 (2n + 12 s)
! )
s sin(2n + 21 s)
+5 +
sin 15 (2n + 21 s)
if s is twice an odd number;
(14.8)
( 12 )s
(1s 3s + 5s )2s
(n) = (n + 14 s)s1
(s 1)!
(   !
sin(4n + s) sin(4n + s)
1s 3s
sin(4n + s) sin 13 (4n + s)
! )
s sin(4n + s)
+5
sin 15 (4n + s)
if s is an odd number greater than 1; and
  !
sin(4n + 1) 1 sin(4n + 1)
r2 (n) = 2 (n) = 3
sin(4n + 1) sin 31 (4n + 1)
(14.9) ! )
sin(4n + 1)
+ 15 .
sin 51 (4n + 1)
248 Paper 21

In all these equations the series on the right hand are finite Dirichlets series and therefore
absolutely convergent.
But the series (14.3) is (as easily shewn by actual multiplication) the product of the two
series
1s c1 (n) + 2s c2 (n) +
and
ns1 (1s + 2s + 3s + ).
We thus obtain an alternative proof of the formul (7.5). Similarly taking the previous
expression of 2s (n). viz. the right-hand side of (11.6), and multiplying it by the series

1s + (3)s + 5s + (7)s +

we can shew that the product is actually the right-hand side of (14.4). The formul for
(n) can be disposed off similarly.
2s

15. The formul which I have found are closely connected with a method used for another
purpose by Mr. Hardy and myself . The function
X
(15.1) (1 + 2x + 2x4 + 2x9 + )2s = r2s (n)xn

has every point of the unit circle as a singular point. If x approaches a rational point
exp(2pi/q) on the circle, the function behaves roughly like

s (p,q )s
(15.2) ,
{(2pi/q) log x}s

where p,q = 1, 0, or 1 according as q is of the form 4k + 1, 4k + 2 or 4k + 3, while if q is

of the form 4k then p,q = 2i or 2i according as p is of the form 4k + 1 or 4k + 3.
Following the argument of our paper referred to, we can construct simple functions of x
which are regular except at one point of the circle of convergence, and there behave in a
manner very similar to that of the function (15.1); for example at the point exp(2pi/q)
such a function is

s (p,q )s X s1 2npi/q n
(15.3) n e x .
(s 1)!
1

The method which we used, with particular reference to the function

1 X
(15.4) = p(n)xn ,
(1 x)(1 x2 )(1 x3 )

Asymptotic formul in Combinatory Analysis, Proc. London Math. Soc., Ser.2, Vol. XVII, 1918, pp.
75 115 [No.36 of this volume].
On certain trigonometrical sums and their applications in the theory of numbers 249

was to approximate to the coefficients by means of a sum of a large number of the coefficients
of these auxiliary functions. This method leads, in the present problem, to formul of the
type
1
r2s (n) = 2s (n) + O(n 2 s ),
the first term on the right-hand side presenting itself precisely in the form of the series
(11.11) etc.
It is a very interesting problem to determine in such cases whether the approximate formula
gives an exact representation of such an arithmetical function. The results proved here shew
that, in the case of r2s (n), this is in general not so. The formula represents not r2s (n) but
(except when s = 1) its dominant term 2s (n), which is equal to r2s (n) only when s = 1, 2, 3
or 4. When s = 1 the formula gives 22 (n) .

16. We shall now consider the sum

(16.1) s (1) + s (2) + + s (n).

Suppose that
X  sin{(2n + 1)/r} 
1
Tr (n) = 2 1 ,
sin(/r)

(16.2)
X sin{(2n + 1)/r}
1
Ur (n) = 2 ,
sin(/r)

where is prime to r and does not exceed r, so that

Tr (n) = cr (1) + cr (2) + + cr (n)

and
Ur (n) = Tr (n) + 21 (r),
where (n) is the same as in 9. Since cr (n) = O(1) as r , it follows that

(16.21) Tr (n) = O(1), Ur (n) = O(r),

as r . It follows from (7.5) that, if s > 0, then

s (1) + s (2) + + s (n) = (s + 1)

(16.3)
 
T2 (n) T3 (n) T4 (n)
n + s+1 + s+1 + s+1 + .
2 3 4

The method is also applicable to the problem of the representation of a number by the sum of an odd
number of squares, and gives an exact result when the number of squares is 3, 5, or 7. See G.H. Hardy, On
the representation of a number as the sum of any number of squares, and in particular of five or seven,
Proc. London Math. Soc. (Records of proceedings at meetings, March 1918). A fuller account of this paper
will appear shortly in the Proceedings of the National Academy of Sciences (Washington, D.C.) [loc. cit.,
Vol.IV, 1918, 189-193].
250 Paper 21

Since

X (n) (s)
=
1
s+1 (s + 1)
if s > 1, (16.3) can be written as
(16.31)
s (1)s (2) + + s (n) = (s + 1)
 
U2 (n) U3 (n) U4 (n)
1
n + 2 + s+1 + s+1 + s+1 + 12 (s),
2 3 4
if s > 1. Similarly from (8.3), (8.4) and (11.51) we obtain

d(1) + d(2) + + d(n)

(16.4)
= 21 T2 (n) log 2 13 T3 (n) log 3 14 T4 (n) log 4 ,

d(1) log 1 + d(2) log 2 + + d(n) log n

(16.5)
= 21 T2 (n){2 log 2 (log 2)2 } + 31 T3 (n){2 log 3 (log 3)2 } + ,

(16.6) r2 (1) + r2 (2) + + r2 (n) = {n 13 T3 (n) + 15 T5 (n) 71 T7 (n) + }.

Suppose now that

X 2 4 2n

s
Tr,s (n) = 1 cos + 2s cos + + ns cos ,
r r r

where is prime to r and does not exceed r, so that

Tr,s (n) = 1s cr (1) + 2s cr (2) + + ns cr (n).

Then it follows from (7.5) that

s (1) + s (2) + + s (n)

(16.7)
 
s s s T2,s (n) T3,s (n) T4,s (n)
= (s + 1) (1 + 2 + + n ) + s+1 + s+1 + s+1 +
2 3 4
if s > 0. Putting s = 1 in (16.3) and (16.7), we find that

(n 1)1 (1) + (n 2)1 (2) + + (n n)1 (n)

(16.8) 2 n(n 1) 2 (n) 3 (n) 4 (n)
 
= + + + + ,
6 2 22 32 42
where
X  sin2 (n/r) 
1
r (n) = 2 n ,

sin2 (/r)
On certain trigonometrical sums and their applications in the theory of numbers 251

being prime to r and not exceeding r.

It has been proved by Wigert , by less elementary methods, that the left-hand side of
(16.8) is equal to
X
2 2 1 1 n 1 () p
(16.9) n 2 n( 1 + log 2n) 24 + J1 {4 (n)},
12 2 1

where J1 is the ordinary Bessels function.

17. We shall now find a relation between the functions (16.1) and (16.3) which enables us
to determine the behaviour of the former for large values of n. It is easily shewn that this
function is equal to

n  h n is  X n h i X n
s n
X
s s s
(17.1) 1 + 2 + 3 + + + [ n] s.
=1
=1
=1

Now
(k + 12 )s+1
1s + 2s + + ks = (s) + + O(ks1 )
s+1
for all values of s, it being understood that

(k + 12 )s+1
(s) +
s+1

denotes + log(k + 21 ) when s = 1. Let

hni n
= 21 + , [ n] = t = n 1
2 + .

Then we have
ns1
h n is  
s s 1  n s+1  n s
1 + 2 + + = (s) + + +O
s+1 s1

and hni
s= n s1 12 s + s .

It follows from these equations and (17.1) that
t 
X 1  n s+1
s (1) + s (2) + + s (n) = (s) + + n s1
(17.2) s+1
=1
s (n + ) s + O ns1
 o
n s


+ + s1 .

Acta Mathematica, Vol.XXXVII, 1914, pp. 113 140(p.140).
252 Paper 21

Changing s to s in (17.2) we have

(17.21) ns {s (1) + s (2) + + s (n)}
t 
X n s1  n s+1  n s
= ns (s) + + + s +
1s
=1
 s+1 
 n s
( n + ) +O .
n
It follows that
(17.3) ns {s (1) + s (2) + + s (n)} {s (1) + s (2) + + s (n)}
t 
X s  n s+1 s
= ns (s) (s) + + n s1 + ( n + ) s
=1
1+s 1s
 s1 s+1


 n s n
( n + ) +O s1
+ .
n

Suppose now that s > 0. Then, since varies from 1 to t, it is obvious that
s+1 ns1
< s1
n
and so
s+1 ns+1
   
O =O .
n s1
. Also
t
X
{ns (s) (s)} = ( n 1
2 + ){ns (s) (s)};
=1
t
X s  n s+1 sns+1 ns+1 1
= (1 + s) ( n + )s + O(n 2 s );
1+s 1+s s+1
=1
t
s ns n 1
( n + )s + O(n 2 s );
X
n s1 = (1 s) +
=1
1s 1s 1s
t
( n + )2+s 1
+ O(n 2 s );
X
s
( n + ) = ( n + )(s) +
=1
1+s
t
 n s ns 1
( n + )2s + O(n 2 s );
X
( n + ) = ns ( n + )(s) +
1s
=1

and
t
ns1
X  
O = O(m),
s1
=1
On certain trigonometrical sums and their applications in the theory of numbers 253

where
1
(17.4) m = n 2 s (s < 2), m = n log n(s = 2), m = ns1 (s > 2).

It follows that the right-hand side of (17.3) is equal to

sn1+s sn 1 s ( n + )2+s ns+1 ( n + )s
(1 + s) + (1 s) 2 n (s) +
1+s 1s 1+s

n( n + ) n ( n + )2s
s s
+ + O(m).
1s
But

( n + )2+s ns+1 ( n + )s 1 1
= 2n 2 (1+s) + O(n 2 s );
1+s

n( n + ) ns ( n + )2s
s 1 1
= 2n 2 (1+s) + O(n 2 s ).
1s
It follows that
s (1) + s (2) + + s (n) = ns {s (1) + s (2) + + s (n)}
(17.5) sn1+s sn
(1 + s) + 12 ns (s) (1 s) + O(m)
1+s 1s
if s > 0, m being the same as in (17.4). If s = 1, (17.5) reduces to

(n 1)1 (1) + (n 2)1 (2) + + (n n)1 (n)

(17.6) 2
= 12 n2 12 n( 1 + log 2n) + O( n) .

From (16.2) and (17.5) it follows that

n1+s
(17.7) s (1) + s (2) + + s (n) = (1 + s) + 21 ns (s)
1+s
 
sn s T2 (n) T3 (n) T4 (n)
+ (1 s) + n (1 + s) + s+1 + s+1 + + O(m),
s1 2s+1 3 4
for all positive values of s. If s > 1, the right-hand side can be written as
ns
(1 s) + ns (1 + s)
s1
(17.8)  
n 1 U2 (n) U3 (n) U4 (n)
+ + s+1 + s+1 + s+1 + + O(m).
1+s 2 2 3 4

This result has been proved by Landau. See his report on Wigerts memoir in the Gottingische gelehrte

Anzeigen, 1915, pp. 377 414 (p.402). Landau has also, by a more transcendental method, replaced O( n)
2
by O(n 5 ) (loc.cit., p.414).
254 Paper 21

Putting s = 1 in (17.7) we obtain

2 2 1
1 (1) + 1 (2) + + 1 (n) = n + 2 n( 1 + log 2n)
12
(17.9)
2 n T2 (n) T3 (n) T4 (n)
 
+ + + + + O( n).
6 22 32 42

Additional note to 7 (May 1, 1918).

From (7.2) it follows that

1 X X
{1s 1r (1) + 2s 1r (2) + } = 1s mr cm (1) + 2s mr cm (2) + ,
(r)
1 1

or

(s)(r + s 1) X X cm (n)
= r ns
,
(r) 1 1
m
from which we deduce
X cm (1) cm (2) cm (3)
(s) ()1s = + + + ,
1s 2s 3s

being a divisor of m and its conjugate. The series on the right-hand side is convergent
for s > 0 (except when m = 1, when it reduces to the ordinary series for (s)).
When s = 1, m > 1 we have to replace the left-hand side by its limit as s 1. We find
that

(18) cm (1) + 21 cm (2) + 13 cm (3) + = (m),

(m) being the well-known arithmetical function which is equal to log p if m is a power of
a prime p and to zero otherwise.