Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Mathematical Association of America

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

A Faster Product for $\pi$ and a New Integral for ln $\frac{\pi}{2}$

Author(s): Jonathan Sondow


Source: The American Mathematical Monthly, Vol. 112, No. 8 (Oct., 2005), pp. 729-734
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/30037575 .
Accessed: 08/02/2015 19:57

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
The American Mathematical Monthly.

http://www.jstor.org

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions
ACKNOWLEDGMENT. The author thanks the referee for remarksthat improved this note immeasurably
both mathematicallyand linguistically.

REFERENCES

1. N. I. Akhiezer, Theoryof Approximation(trans.C. J. Hyman), FrederickUngar, New York, 1956.


2. S. N. Bernstein,D6monstrationdu thoreme de Weierstrassfondde sur le calcul de probabilit6,Proc. Math.
Soc. Kharkov13 (1912/13) 1-2.
3. M. H. Stone, The generalized Weierstrassapproximationtheorem, Math. Mag. 21 (1948) 167-184, 237-
254.
4. K. Weierstrass,Uber die analytische Darstellbarkeitsogenannterwillkiirlicher Functionen einer reellen
Vertinderlichen,Berliner Berichte (1885) 633-639, 789-805.

Departmentof Mathematics, The Universityof Toledo, Toledo, Ohio 43606


rnagise@math.utoledo.edu

A Faster Product for Trand


a New Integral for In 7
Jonathan Sondow

1. INTRODUCTION. In [17] we derived an infinite product representation of eY,


where y is Euler's constant:

=(2)1/2 22 1/3 (23 4)1/4 24 .44 )1/5


1 (1-3 1-33 1-3-5 (1)
Here the nth factor is the (n + 1)th root of the product

H (k + 1) 1)k+1(
1

k=O

In the process we noticed a strikingly similar product representationof 7:


n 2 /2 22 1/4 23 418 24 . 44 1/16
2 )1/ (i-3)1 (3 )1/ (1-36-5)1/6 (2)
In this note we give three proofs of (2). The third leads to an analog for ln(7r/2) of
integrals for ln(4/7r) [16] and y [15], [16], [17]:

In -= -x dx, (3)
2 J[o,l (1 +x)ln x

4 f 1-x
ln - - - dxdy, (4)
S n (1 + xy) Inxy
[0,1]2

S 1 -x dx dy

[0,1]2

October 2005] NOTES 729

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions
Using (3), we sketcha derivationof (1) and(2) fromthe samefunction(a formof the
polylogarithm[10]), accountingfor the resemblancebetweenthe two products.The
functionalso leadsto a productfor e (dueto J. Guillera[6]),
(2)/ 22 123 .41/3 24 44 )1/4
e= () 1-3)
3 1 33 1(3-s)5 " (5)

surprisinglyclose to product(1) for eY. (A generalizationof (5), related to the


Bolthausen-Sznitman coalescentin probabilitytheory,has been found by C. Gold-
schmidtandJ. B. Martin[5].)

2. THE ALTERNATINGZETA FUNCTION. The logarithm of product (1),


namely,
n0 1 f k/I\

y' =1-n-
n= n +1 k=0(- 1 k+lyjln(k
k + 1), (6)

remindedus of the series(see [1, eq. 5.16], [8, p. 464], and[13])

*(s) 2n+ (-1) (k + 1) (s C), (7)


n=0 k=0

whichgives the analyticcontinuationof the alternatingzeta function *(s). The latter


is definedby the Dirichletseries(see [14])
00 (-1)k-1
*(s)= ( ((s > 0). (8)
k=l

(For example,using the classic formula *(1) = In2 for the alternatingharmonic
series-for a new proof see [14]-one can derive the series ln 2 = n>,, (2"n)-1
from (7) by consideringit when s = 1.) Differentiating(7) termwiseand substitut-
ing the value of the derivative of * at s = 0,
1 r
*'(0) = - In - (9)
2 2
(see [13]), yields the series

n
ln _ 2l1(
- 1 k=0 (_l ln\k (10)

andexponentiationproducesproduct(2).

3. WALLIS'S PRODUCT AND EULER'S TRANSFORM. The pair of infinite


products(1) and (2) calls to mind anotherpair,Wallis'sproductfor nr[19] and Pip-
penger'sproductfor e [12]:
rt 2244668

e (2)1/2 (2 4)1/4 4 6 6 8)1/8


2 i1 (33) (5577) -(2

730 THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 112

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions
It is interesting to note that products (2) and (12), whose factors have exponents 1/2",
converge rapidly to numbers Jr/2 and e/2 whose irrationalityhas been proved (see, for
example, [11]), whereas product (1), with exponents 1/(n + 1), converges less rapidly
to a number eY whose (expected) irrationalityhas not yet been proved.
We give a second proof of (2) using (11) and Euler's transformationof series,
(-1)n-lan 2n+1 (-1)k ak+l, (13)

n=l n=0 k=0 k

valid for any convergent series of complex numbers [9, sec. 33B], [13]. Applying (13)
to the logarithm of Wallis's product

In (-1)n-1 In n (14)
n=1

gives

n 0 1 ) lnk+2
In - = ( Ink+ (15)
n=0 k=0 W /c

If we replace n with n - 1, write the last logarithm as ln(k + 2) - ln(k + 1) and the
sum on k as the difference of two sums in the first of which we replace k with k - 1,
then the recursion
(n - (n- 1 (n\
k -i)+ nki)-]\k
leads to (10), completing the second proof of (2). The first proof is basically the same,
because in [13] we use Wallis's productto evaluate (9), and we take the Euler transform
of (8) to get (7) for complex s with 9t(s) > 0.
Products (12) and (11) are linked by Stirling's asymptotic formula

n! ~ (n/e)/2nn,

which is proved in [3] using (11) and is used in [12] to establish (12). Products (1)
and (2) are linked by transformations:a hypergeometric one [17] for (1) and Euler's
for (2). (To strengthen the link, we can write series (14) and (15) as integrals of hy-
pergeometric functions-compare [17, Proof 1]-and then obtain (15) from (14) by a
hypergeometric transformationequivalent to (13).) However, this link does not explain
the remarkableresemblance between (1) and (2).
Euler's transformationaccelerates the rate of convergence of a slowly converging
series such as (14) (see [9, sec. 35B]). Thus, product (2) converges faster than prod-
uct (11), as Figure 1 (courtesy of P. Sebah) shows.

4. AVOIDING EULER. A third proof of (2) (due in part to S. Zlobin [20]) avoids
using Euler's transformationaltogether (compare the proofs avoiding hypergeometric
functions in [17]). We show that
f00 /f1 0 /1 - \n+1

I :=f j
Jo =
xy
2
dx dy = In -.
2
(16)

This implies (2), because if we factor 1 - x from the integrand and use the binomial
theorem, then termwise integration (justified since the integrand is majorized by the

October 2005] NOTES 731

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions
2 2

1.9 1.9

1.8 1.8

1.7 1.7

1.6 1.6

1.5 1.5
1.4 1.4

1.3 1.3
1 2 3 4 5 6 1 2 3 4 5 6

Figure 1. Partialproductsof Wallis's productand its Euler transform.

series 2-"-') yields (15) and, therefore, (2). To prove (16), we use the geometric
series summations

" 1-x n+l 1-x (1-x)2 00


S ^
= - ^(\- = (1 _x)2X2n
n=O 2 1+ x 1-x2 =(1
n=O (17)

to write

I= (1 - x)2xy+2"ndx dy.

The integrandis majorized by C (n + 1)-2 (because

max (1 - x)2x2n ( 1 2 n - 1
o<x< n+1 ( n+1 (n+1)2

and x y 1), so we may perform the integrationsterm by term, which by invoking (11)
gives

00 (2n + 2)2 22 42 62 27
I= ln( =InI - = ln--.
no (2n+1)(2n+3) 133.55.7 2

This proves (16) and completes the third proof of (2).

Proof of (3). Equation (16) and the first equality in (17) yield
f0 ff 1-x
xY- dx dy - In .
o o 1+x 2

Reversing the order of integration (permitted since the integrand is nonnegative), we


integrate with respect to y and arrive at formula (3).
Alternatively, one can derive (3) from (4) by making the change of variables
u = xy, v = 1 - x and integrating with respect to v: the result is In 2 minus inte-
gral (3) (with u in place of x), and equality (3) follows. U

5. RELATING THE PRODUCTS FOR ir AND e". Recall that we derived product
(2) for Jrfrom the alternatingzeta function * (s). Omitting details, we sketch a deriva-

732 THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 112

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions
tion of product (1) for eY from a generalization of *(s). This accounts for the resem-
blance between the two products. (Formulas (18) and (19) are due to J. Guillera [6].)
We generalize series (7) for *(s) by defining the function

f(t,s)= tn+l (-1)k (k+ )- s (-1 <t < 1, s C), (18)


n=0 k=0 k

so that f(1/2, s) = '*(s). Using integral (16) but replacing (1 - x)/2 with t(1 - x),
we can show that the formula obtained from (3) and (9),
1 r1 1-x

2J (1 + x)lnxdx,
extends to

f'(t, 0) = - 1t2 - dx, (19)


(1 - t (1 - x)) lnxx

where the prime ' is shorthandfor a/as.


We now derive product (1) by evaluating the integral fo t- f'(t, 0) dt in two dif-
ferent ways. On the one hand, a glance at (18) reveals that this integral equals the right
side of (6). On the other hand, substituting (19) into the integral and reversing the order
of integration gives
f'(t,) f1 f1 t(l-x) 1 1 1
Sf(t-O)dt = - dtdx= +- dx.
Jo t (1 - t (1 - x)) Inx J Inx 1- x

The last is a classical integral for Euler's constant [2, sec. 10.3], [17], and (6) follows,
implying (1).
Otherproducts can be derived in the same way. For example, exponentiating the in-
tegral f1 t-2 f'(t, 0) dt = 1 gives product (5) for e, which converges more slowly than
Pippenger's product for e, because of the exponents 1/n in (5), versus 1/2" in (12).
In order to identify the function f(t, s), we reverse the order of summation in (18)
and sum the resulting series on n. We then replace k with k - 1, obtaining

f (t, s) - k + 1) - kl ks (20)

for t satisfying -1 < t < 1/2 and for suitable s. Therefore, f(t, s) and '*(s) are
related to the function

tk
F(t, s) = - (- 1 t < 1,( () > 0)
k=l

by the formulas

f (t, s) = -F(t/(t - 1), s), (21)


*(s) = -F(-1, s),

for appropriatet and s. (With t = 1/2, equations (18) and (20) verify that formulas (7)
and (8) for *(s) agree.) The function F(t, s), a special case of the Lerch transcendent

October 2005] NOTES 733

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions
P(z,s, v) (see [4, sec. 1.11], [18, Sec. 64]), is the polylogarithmLis(t) when s is an
integer [10, p. 189], [18, secs. 25, 64]. Relations(18) and (21) lead to an analytic
continuation of F(t, s), and thus of the polylogarithm [7].

REFERENCES

1. P. Biane, J. Pitman, and M. Yor,Probabilitylaws related to the Jacobi theta and Riemann zeta functions,
and Brownianexcursions, Bull. Amer.Math. Soc. 38 (2001) 435-465.
2. G. Boros and V. Moll, Irresistible Integrals: Symbolics,Analysis, and Experimentsin the Evaluation of
Integrals, CambridgeUniversity Press, Cambridge,2004.
3. A. J. Coleman, A simple proof of Stirling's formula, this MONTHLY 58 (1951) 334-336.
4. A. Erd6lyi, et al., Higher TranscendentalFunctions, The Bateman ManuscriptProject, vol. 1, McGraw-
Hill, New York, 1953.
5. C. Goldschmidtand J. B. Martin,Randomrecursivetrees and the Bolthausen-Sznitmancoalescent, Elec-
tron. J. Probab. (to appear);also available at http://arxiv.org/abs/math/0502263.
6. J. Guillera,personal communication,25 July 2003.
7. J. Guilleraand J. Sondow, Double integralsand infinite productsfor some classical constantsvia analytic
continuationsof Lerch's transcendent(2005, preprint);available at http://arxiv.org/abs/math/0506319.
8. H. Hasse, Ein Summierungsverfahrenfuirdie Riemannsche (-Reihe, Math. Z. 32 (1930) 458-464.
9. K. Knopp, Theoryand Application of InfiniteSeries, Dover, New York, 1990.
10. L. Lewin, Polylogarithmsand Associated Functions, Elsevier North-Holland,New York, 1981.
11. A. E. Parks, rt,e, and other irrationalnumbers,this MONTHLY 93 (1986) 722-723.
12. N. Pippenger,An infinite productfor e, this MONTHLY 87 (1980) 391.
13. J. Sondow, Analytic continuationof Riemann's zeta function and values at negative integers via Euler's
transformationof series, Proc. Amer Math. Soc. 120 (1994) 421-424.
14. - , Zeros of the alternatingzeta function on the line 9(s) = 1, this MONTHLY 110 (2003) 435-437.
15. - , Criteriafor irrationalityof Euler's constant,Proc. Amer Math. Soc. 131 (2003) 3335-3344.
16. - , Double integralsfor Euler's constant and ln(4/r) and an analog of Hadjicostas's formula, this
MONTHLY112 (2005) 61-65.
17. - , An infinite productfor ev via hypergeometricformulasfor Euler's constant, y (2003, preprint);
available at http://arXiv.org/abs/math/0306008.
18. J. Spanierand K. B. Oldham,An Atlas of Functions, Hemisphere,New York, 1987.
19. J. Wallis, Computationof n by successive interpolations,in A Source Book in Mathematics,1200-1800,
D. J. Struik, ed., Princeton University Press, Princeton, 1986, pp. 244-253; reprintedin Pi: A Source
Book, 2nd ed., L. Berggren, J. Borwein, and P. Borwein, eds., Springer-Verlag,New York, 2000, pp.
68-77.
20. S. Zlobin, personal communication,26 May 2003.

209 West97th Street,New York,NY 10025


jsondow @alumni.princeton.edu

Generalizationsof Fermat's Little Theorem


via Group Theory
I. M. Isaacs and M. R. Pournaki

1. INTRODUCTION. As is well known,Fermat'sso-called little theoremasserts


that if p is prime, then aP = a (mod p) for all integers a. This is usually proved by
consideringtwo cases:theresultis trivialif a = 0 (modp), andit followsfromthe fact
thataP-'1 1 (modp) if a is notdivisibleby p. Thislatterfactis generallyestablished
by an appealto the theoremof elementarygrouptheorythatassertsthataIGI= 1 for
everyelementa of a finitegroupG.

734 THE MATHEMATICALASSOCIATIONOF AMERICA [Monthly 112

This content downloaded from 128.235.251.160 on Sun, 8 Feb 2015 19:57:14 PM


All use subject to JSTOR Terms and Conditions

You might also like