Interesting Series Involving the Central Binomial Coefficient Author(s): D. H. Lehmer Source: The American Mathematical Monthly, Vol.

92, No. 7 (Aug. - Sep., 1985), pp. 449-457 Published by: Mathematical Association of America Stable URL: http://www.jstor.org/stable/2322496 Accessed: 26/09/2010 05:26
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=maa. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. 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

INTERESTING SERIES INVOLVING THE CENTRAL BINOMIAL COEFFICIENT

D. H. LEHMER CA 94720 Berkeley, ofCalifornia, University ofMathematics, Department

A seriesis as follows. senseexplained is used herein a technical The adjective"interesting" its and at thesametime its nthterm for formula is a simple explicit in case there calledinteresting Thus of known constants. in terms sumcan be expressed 1+ 4 2 -++*+-+* 1-2+ 34+ 2 34 16 *-+( 256 2 ) -+* n
n4

= 2, =log2, 90

1 1

81

1 +1 .+i

series. ofinteresting are familiar examples are of-two types: The serieswe plan to discuss I
O

an 2n)

and

II.

(2n

of TypeI. of n. We beginwithseries functions simple wherethe a,1 are very we have theorem By thebinomial
(1)
2()n

series we gettheinteresting if IxI < 1/4. If we put x = 1/8, forexample, This converges {2nA (2) If x = 1/10, we get
1?1?4

1? 2 + 32 128

_ _-n

(3)
Forx= -1/8

1+.2+.06+.02+

.007++

+( n +

(4)

5 3 1 + 128 32 4 32- 12 + 1-

(-)(n) + 8 8

(2) and (4) givesus Averaging

E
n=o

( 2n

64n

3V2 + V6 6

of x leads to Thususingix instead in complex variables. Another stepalongthispathis to bring

MONTHLY.

first of theauthor's of thepublication paperin this thesixtieth anniversary Thispaper marks D. H. Lehmer: arises. time as occasion from to time ofother to submit papers The author manuscripts hopes

449

450 8n
4 E4J=3+
n=

D. H. LEHMER

[August-September

84n

+2
+

or
,

(8n\
44
n0

(15

V ? + 6 V2 + dV) + 5 60

In general one can obtaina valueforthesum (2( an+b)) n= an+b

n can+b

the by settingx = er/c (v = 0,1,2,... ,a - 1), where E = exp(2q7i/a), in (1) and forming linearcombination of these a process known as multisection (see Riordan[4]). appropriate results, If we integrate Another (1) from 0 to x we obtain path out of (1) is to applyoperators.
0n=O

(2n)ftndt 0

(1

- 4t

Dividingbothsidesby x we get
(5)

(~2n)

=1(

Theirfirst tenvaluesare The coefficients areintegers known as Catalannumbers. Cnof theseries 1, 1, 2, 5, 14, 42, 132,429,1430,4862,... formula occurfrequently in combinatorial for and they analysis (see Gould[2]).ByusingStirling's n! we see that
Cn
=

g(

n + 1)(

8n)

in (5) converges Thus(5) gives slowly likeX(1/n3/2). Fromthiswe see thatforx = 1/4 theseries series us theinteresting
1 +

64

+)

{2nA ( n

4n(n +1)

= 2.

itsfirst term to theright both We can treat side,dividing (1) a little differently by transposing Thisgivesus theimproper sidesby x and then integral integrating.
E(1

n )tn

dt

jx( t(l

- 4t)1/2

t)

That is,
(6) , (2)n = 21og( 1

x= Putting

we gettheseries
1 3 5
35

16

48

512

n4f

nlog4.

1985]

INTERESTING

SERIES INVOLVING THE CENTRAL BINOMIAL COEFFICIENT

451

we get thesignsof thisseries, If we alternate

-

1 2

---+---

16'

5 48

+ ~ 512

35

(-))?(n)
n4f
n

I lo =l

+I V?
g2]

forthetwosums series interesting we can derive thesetwoseries Addingor subtracting

d n odd

n )2n4n

or

n even n>0

n )n4n

(6) and obtain We can integrate (7)

n x,n(n +1 ) n= 2xlog xnY2

+x

+ x4(log4-1) 2 2xlg41-

series x = 1/4,we obtaintheinteresting If we substitute 1 1 -+--~ 4 16 7 5 (n) +~+ ++**++**=log4 192 512 n(n + 1)4

-1.

of (1). integrations by repeated of thissortcan be obtained of examples Anynumber Let 6 be theoperator of thiskindis differentiation. operator Another
=

xd

dx

If we apply0 to (1), we get (8)

If we set x
-,

00

n=1

n n= (2,~l) n

(=3

l~4x

we get 1

4+ +125+ 4 16 125

15

512 +=

35

~~

_ n_

2-

againby , we obtain Operating

n=1

n2( 2n xn nfl

2(1

4x)-1/2

2x(2x + 1)

(1

-x5/

x Setting

we find 4
1
-+-+~

128

45

128
(o n=O
n

35

~++

n+

8192

1575
+

n2(2n)

bothsides,we get integrate If in (1) we replacex by x2 and then

2n)

2n +

x2

x (arcsin2x)

and for x

4 00

n=O

n (2n + 1)16's

(2n)

in the We turnnow to seriesof type II whichinvolvethe centralbinomialcoefficient In an informal and less well understood. These series are more mysterious denominator.

452

D. H. LEHMER

[August-September

[5] refers to four of '(3), A. J.van der Poorten of Apery's proofof theirrationality discussion was quoted and its proofby Z. A. Melzak [3] was theorem seriesof typeII. The following differently proof. as "not quiteappropriate". We givean entirely described
THEOREM. If (9)

lxi< 1,
x xarcsin ~~~~~~~~2
/ -x 2 2=E
-

(9)

2x

(2) m=i M(2mn

Gregory series Proof. We use thefamiliar (10) and set t

X/ l1
-

tarctant=
X2,

m=1

00

(_j)m-1t2m

2m-1

x. Then(10) becomes so thatarctan t = arcsin

arcsin
2 () m-2m

acm-E

(2m -1)(1mm-1 m-1

_2)m

Z
j=O

_0 Exr
r=1

L
m=1 (m-

_)-(r-)
)!(r-

m)!(2m

1)

side of (9). Thatis, of X2r is halfof thaton theright to showthatthecoefficient It suffices (11) r) ~~~~~r(
(7/2

v!(r - v - 1)!(2v + 1) =22rl

2r- 1dO

To provethiswe use Wallis'integral

(sin Josln,

da

2 46 .. (2r - 2)_ 1) 1 3 5 ...(2r-

The left-hand sideof (11) can be written as follows.

r( 2r)E

r____

(1)

(r

1)

v+1

= r(2 r)l
-

E
V

l(rY2))

2-dy y

r(2rr)|(1

dy

r~~~~~r rK2r)I / (sinG)2r dO

we find y = cos 0. UsingWallis'integral by substituting 22 (2rr rk

)r(r (r -

(2r

1)!

1)! = 22r-

is established. This proves(11) and hencethetheorem x= If we substitute

2,

we gettheresult [5],p. 202) (van derPoorten

1985] (12)

INTERESTING

SERIES INVOLVING THE CENTRAL BINOMIAL COEFFICIENT

453

m=1 m( m )

series of Type II. If we wish to alternatethe signs in this series,we merely as our first interesting put x = i/2 in (9). This gives us

m -1

m( 2 m )

)/

we obtain If we divide both membersof (9) by 2x and thenintegrate,

(13) 2(arcsinX)2
=

m=1 M(2)

_(2X)2m

fromwhich we get, forx (14)

-2

00

T2
1

Also, where ?(s) is Riemann's zeta function.

00 ( ( + 1 )2 m logz - -2~~~~~~~

we obtain If we divide both sides of (13) by 2x and integrate,

E

(2

x)2m

4X

(arcsin

y)2

m=1

m3( 2m )

This integral is a "higher transcendent".It is closely connected with Spence's transcendent, function.For x = 2 we obtain Clausen's integral,and the trigamma L 1
m

= 4f1 /(arcsiny)2d

=-2f

x log(2 sin ) dx

=- 3

-72

+3)

(3);

where +(x) is the trigammafunction.Van der Poorten [6] rejects this evaluation as being series non-instructive. However, he does give the interesting 1 + -+ 1 - + 1 -_ - - 1 2 48 540 4480 as well as Comtet's [1] remarkable
( 1 +iI + *(-1)~ m3(2 m) =2(3)

00 m1m

177T4

m(4)

3240

454

D. H. LEHMER

[August-September

Thereare no known interesting series of theform

mk)

fork > 4. If we apply the operator0k to both sides of (9), we get an almostunlimited number of interesting series. To beginwithO we obtain (15)
E m=(

(2X)2m_
2rn)
m

x2
1-X2

x arcsin x
(1-x2)3/2

Fromthiswe obtain, forexample, 1 2 1 6 1 20 1 70 1

(2n)

9 + 2'nV3 27

If we replacex by ix in (15) and set x = a/b, we getanother form of (15)

E

1)m1(2a/b)2m (2m h)

ab2[log a+ h )

ah]

m=1

whereh=

a2 + b2. For example, if a= 1, b 72, h =/,

1 1 1 - 1 - 70 + 2 - 6 + #20 ( 1)n1

we get 1 + 5 5_

+ ( (2n) ( 2) )

4J r

og 25 25og

x/S? +1 2

For a

23660, b = 23661,we have h = 33461and we obtain 00 ~

nEl(1l)

n-1

/23660 (23661 (2n) n

57121 23660+ (23661) 33461 log23661] (33461)2

23660

= .811587506.... This seriesis of courseconvergent. If one examines theratioof consecutive terms one finds that theterms increase in absolute valueuntiln = 11830whentheterms are as largeas 117.After that decreaseto zero.It is wellthatwe haveour formula they forthesum. If we operateon (9) byhigher powers of 4, theseries we obtainare of theform 00 mk -2 (2x )2m ( k > 0) (16) M=1 {2m 2m and thevalue of thesumdepends, as a function of k, on twosequences ofpolynomials (t) and Vk as follows: Wk (t) defined recursively V1(t)=1, w1(t) = O, (17) (18)
Vk+l(t) = {(2k - 2)t + 1} Vk(t) + 2(1 t)4Vk(t),

Wk+,(t) = ((2k - 4) t + 2} WK(t) + 2(1 - t) 4Wk(t) + Vk(t). The first fewpolynomials are
V2= 1, W1= 1,

1985]

INTERESTING

SERIES INVOLVING THE CENTRAL BINOMIAL COEFFICIENT W3

455

V3 = 1 + 2t,

= 3,

V4= 1 + 10t+ 4t2,

V5 =1

W4= 7+ 8t,
W5 = 15 + 70t + 20t2.

+ 36t + 60t2 + 8t3,

of Vk and Wk by explicitly in terms The value of thesum(16) can be given

k-2(l

_ X2)k-1/2 Iarcsinxkk(x)+

- x2Wk(x2)]

of arcsinx and x 1 If we operateon thisby , and collectthecoefficients of thisresult is, withx = sinG, form (17) and (18). A trigonometric formulas
00
Em
mk-24m (sinG0)2m_

we obtainthe

(2rn)

sin2G
sin

[2GVk(sin2') + sin2OWk(sin?)].

series If we replacex by ix we getforthealternating

(o
m=2

1) m-1mk-24m (sinhZ)2m

sinh2z

[2log{sinhz

+ coshz}

(2rz)

(2cosh2z)k X Vk(-sinh2z)
+ sinh2zWk(

- sinh2z)I.

as ofinteresting series 0 anqd k yielda widevariety twoparameters their with These twoformulas from 1 to oo. The sumsextend We listonlythefollowing. examples.

Zm2 ( m)
m 2m

2 ( q7Tr + 9)
2

(2m)

81
=

__

m3

4052 11 ? + 2 1375T

m)

=125[2a

(~1) ml12

+ 15]

(a =

log I
+

2=1.076022352)

(2m) ( - 1)m[ m3

1251

]
-

m2(2m3)

2m

625

2(2mn)
2mm

2
7T

456 = =- + 1 2

D. H. LEHMER

[August-September

2mA mJ m2m

(Trn)

(2m)
m2 m {2mA

1
5

32m

77+ 2

2m~~~~ m4

{2m

m42

2m

= 113,r + 355 = 229093376w + 719718067.

(2m) mJ

In general,

M=
to v. whereb/a is a close approximation
3
m2 2m)
2 q72

?? mk2m
1(2m)

+ b, = aiT

m(2m) E

T3 3m

= v = 3.627598728

2)=2v+

3m

(2m
m

= lOv + 18

3~ = 2(43v + 78)

2m
E ( J
=

p73,

p = VYlog(2 + where

4)=

2.281037989

m(2m) mJ
2 + 3

( 1)m-12m=

(,2m)
( 1) m-1 2m

(2m)

1985]

THE LOGIC OF GRAPH-THEORETIC

DUALITY

457

(-1)"

m1m22m

(2m) m
E

27

( - 1)rnm 32m )m2 (2rn m

_1=

[ p + 81/?]

15]

(2 -I2)tm
______ -

3 - 2(2

+4) ~~( 7TV2

15 16

(1) m-I32m
4m( 2m)

l (og2?-+ 125

48

2n1 (2 -

7T2 ft3)tm

m2 2m(3)

References 1. 2. 3. 4. 5. 6. 1974,p. 89. Dreidel,Dordrecht, Louis Comtet,AdvancedCombinatorics, 243 titles. containing of theEuler-Fuss-Segner-Catalan Sequence,manuscript H. W. Gould, Bibliography vol. 1, Wiley,1973,p. 108. Z. A. Melzak, Companionto ConcreteMathematics, Wiley,1968,pp. 131-135. Identities, JohnRiordan,Combinatorial vol. 1 (1979) 195-203. A proofthatEuler missed,Math. Intelligencer, A. J. van der Poorten, __, formulas, Queen's Papersin Pure and Applied Math.,54 (1979) 269-286. Some wonderful

THE LOGIC OF GRAPH-THEORETIC DUALITY

T. A. McKEE Dayton, OH 45435 Wright State University, and Statistics, ofMathematics Department

Someoptimists areprized wherever they occurin mathematics. principles Dualitiesand duality see Others in a theory. (also optimists) thenumber of results doubling see themas mechanically intoeach. butpacking moresubstance of results, thenumber themas halving in graphtheory as it is in from beingas powerful keep duality which Thereare limitations text[24]is RobinWilson'sintroductory theme. manyotherareas,but it is stilla majorunifying As evidence of the and it is also central to manyarea of applications. aroundduality, organized with Graph Theory & Johnson's in bookssuchas Johnson on duality theemphasis latter, consider [19], and Networks Graphs monograph research operations [8],Price's Applications Engineering [16]. Integral and theFeynman Theory Graph andNakanishi's with between circuits and cutsets, centers on therelationship duality Classicalgraph-theoretic thisduality and its of thispapersurveys a basicrole.The first section trees also playing spanning A key feature of thisdualityis the role played by a as a syntactical principle. formulation
in of Wisconsin-Madison logic fromthe University a Ph.D. in mathematical A. McKee: Afterreceiving Terry I have sincegrown increasingly and geometry. of logic to subjectssuch as topology 1974, I workedwithapplications My otherinterests and now inhabita nichewhichmay be describedas "graph metatheory." fond of graph theory fiction. themethodologies evidencedin vintagedetective include lap swimming and studying