You are on page 1of 2

L.

Panaitopol

A formula for ( x)

NAW 5/1 nr.1 maart 2000

55

L. Panaitopol
Facultatea de Matematic a , Universitatea Bucure sti Str. Academiei nr.14, RO-70109 Bucharest 1, Romania pan@al.math.unibuc.ro

A formula for ( x) applied to a result of Koninck-Ivi c


We are going to give an approximate formula for ( x) which is x better than the well known ( x) log x , or than the more precise x x formula from [2]: ( x) log x1 , meaning that ( x) = log x , ( x) where lim ( x) = 1. We will prove
x

for m = 1, 2, . . . , n. For y > 0 we consider f ( y) = ( and we have


n

i! yi +1

)( y 1

i =0

ki ), i i =1 y

Theorem 1. ( x) = x log x 1
k1 log x k2 log2 x kn (1+n ( x)) logn x

f ( y) = 1 +

...

2! 1! k1 3! 2! 1!k1 k2 + +... y2 y3 n ! ( n 1 ) ! k1 ( n 2 ) ! . . . kn1 1 + + O ( n+1 ) yn y

where k1 , k2 , . . . , kn are given by the recurrence relation kn + 1!kn1 + 2!kn2 + . . . + (n 1)!k1 = n n!, and lim n ( x) = 0.
x

for y . It follows that f ( y) = 1 + O( yn1 +1 ), i.e. n = 1, 2, 3, . . .

i +1 i =0 y

i!

1 + O( yn1 +1 ) y1
n i =1 ki yi

1 y1
n i =1 ki yi

+ O(

1 ). yn+2

Proof. The following asymptotic formula ( x) = Li( x) + O( x exp( a log x) ), where a and are positive constants and < 3 5 is well known [3]. Integrating by parts and taking into account that x exp( a log x) = o( where n 1, it follows that ( x) = x 1 1! n! + +...+ 2 log x log x logn+1 x x logn+2 x

We denote y = log x, and using the relations of type (1) it follows that x x ( x) = + O( ) n n+2 (2) ki log ( x) log x 1 (log x)i
i =1

),

Consider ( x) = x log
n+2

x log x 1
k1 log x

k2 log2 x

...

kn (1+n ( x)) logn x

+O

x (1)

1 Combining this formula with (2) yields knn ( x) = O( log x ), from which it follows that lim n ( x) = 0. x

We dene the constants k1 , k2 , . . . , kn by the recurrence km + 1!km1 + 2!km2 + . . . + (m 1)!k1 = (m + 1)! m!,

Remark 2. It can be shown immediately that k1 = 1, k2 = 3, k3 = 13, k4 = 71.

56

NAW 5/1 nr.1 maart 2000

A formula for ( x)

L. Panaitopol

We give now a formula for km (although not suitable for a direct computation). Theorem 3. The coefcient km is given by the relation: m m! (m1)(m1)! . . km = det . 2 2! 1 1! 1! 0! . . . 0 0 2! 1! . . . 0 0 (m 1)! (m 2)! . . . 1! 0!

where lim k( x) = 3, and it follows that


x

log n 1 1 k(n) 1 = , (n) n n n log n n log2 n for n 2. Therefore we get that 1 log n 1 1 k(n) (n) = n n n log n n log2 n . n=2 n=2 n=2 n=2 n =2 For x e, f ( x) =
log x x

[ x]

[ x]

[ x]

[ x]

[ x]

is decreasing and thus log x log k dx , x k

Proof. The recurrence relations giving the coefcients km are: km + km1 1! + . . . + k1 (m 1)! = m m! km1 + . . . k1 ( m 2 ) ! = ( m 1 ) ( m 1 ) !

log(k + 1) k+1

k +1

for k 3. It follows immediately that log n log t log x dt + O( ), n = t x n =3 n=3 and so log n 1 = log2 x + O(1) . n 2 n =2
[ x] [ x]

k2 + k1 1! = 2 2! k1 = 1 1! The determinant of this linear system is 1 and the result follows by Cramers rule.

[ x]

As an application of the above results we are going to improve the following approximation, due to J.-M. de Koninck and A. Ivi c, [1]: 1 1 (n) = 2 log2 x + O(log x). n =2 Using Theorem 1 we are going to prove Theorem 4. 1 1 (n) = 2 log2 x log x log log x + O(1). n =2
[ x] [ x]

Similar arguments lead us to the relations 1 = log x + O(1) , n n =2

[ x]

and 1 = log log x + O(1) . n log n n =2

[ x]

As there exists M > 0 with | k( x) | M, and vergent, it follows that plete.


k(n) 2 n=2 n log n

1 2 n=2 n log n

is con-

[ x]

= O(1), and the proof is com-

Proof. It is enough to take ( x) = x log x 1


1 log x

k( x) log2 x

References
1 J.-M. de Koninck, A. Ivi c, Topics in arithmetical functions, North-Holland, Amsterdam, New York, Oxford, 1980. 2 J. B. Rosser, L. Schoenfeld, Approximate formulas for some functions of prime numbers, Illinois J. Math., 6 (1962), 64-94. 3 I. M. Vinogradov, A new estimate for (1 + it), Izv. Akad. Nauk. SSSR, Ser. Mat. 22 (1958).