FEATURE I ARTICLE

The Arithmetico-geometric Mean of Gauss

How to find the Perimeter of an Ellipse

B Sury

Elementary Calculus

As all of us learn very early in school, the length of a

circle is 71" x diameter. Somewhat later, one learns that
elementary calculus gives the length of an arc of the
unit circle to be the value of the integral s = s(x) =
The author enjoys rx dt
JO ~
. -1 ( )
= SIn x.
writing about such great
masters as: Before going further, it is worthwhile to recall how length
A person who does arouse of (part of) a curve is defined. Suppose first that the
much admiration and a curve is given on the plane and is described by the pair
million wows!
of equations x = x(t), y = y(t) in terms of a parameter
Such a mathematicians'
prince t. Intuitively, one would think of the length of any part
has not been born since. of the curve to be that of parts of a polygon which in-
I talk of Carl Friedrich creasingly approximates that part of the curve. Thus,
Gauss! one divides the interval [a, b] where t varies into finitely
many parts a = to < t1 < < tn = b. The points
Pi = (x(t i ), y(ti)) can be joined by line segments and
the sum POP1 + P1P2 + + Pn - 1Pn of the lengths of
these segments is an approximation to the length of the
part of the curve which corresponds to the parameter
t varying from a to b. This approximation might be a
crude one but it is at once clear (see Figure 1) that one
can take more points to obtain a better approximation.
At this point, calculus comes to the rescue. It is heuris-
tically clear that the best possible notion of length is ob-
tained as the limiting case as one increases the number
n of points Pi indefinitely while simultaneously letting
the lengths of all the line segments Pi P i + 1 diminish in-
definitely. In other words, if the limit exists, it can be
defined as the length of the curve. A curve for which
this limit exists is called rectifiable for evident reasons.
Clearly, circles, ellipses and hyperbolae are rectifiable

-72-----------------------------~------------------------------
RESONANCE I August 2000
 FEATURE I ARTICLE

curves.

Exercise Show that the curve defined by

y = x 2 sinl/ x 0 < x :::; 1 y(O) =0

is rectifiable while that defined by

y = xsinl/ x 0 < x :::; 1 y(O) =0

is not. What is the length of the first curve?

Is there a nice condition which is sufficient to ensure

that a given curve is rectifiable? Let us look at a curve
Figure 1.
with a parametrisation x = x(t), y = y(t) for a :::; t :::; b
where these functions are assumed to have continuous
first derivatives. This condition is indeed enough to en-
force rectifiability as we show now. Let a = to < t1 <
< tn = b be a subdivision; then the approximating
perimeter
n-1
Sn =L Pi Pi+1 =
i=O
n-1
L V[{x(ti+d - X(ti)}2 + {y(ti+1) - y(ti)}2].
i=O

By the mean-value theorem, there are ai, bi E (ti' ti+1)

such that X(t i+1)-X(ti ) = :(ai)(ti+1-t i ) and y(ti+1)-
y(ti) = 1jf(b i )(ti +1 - t i ). When we let the number of
intervals (ti' t i +1 ) tend to infinity while simultaneously
letting their lengths tend to 0, the limit is, by defini-
tion, just the integral J: V(:)2+ (¥t)2dt. We note that
the above limit exists by uniform continuity of ~~, !fit -
consequence of the assumption that the parametrising
functions x = x(t), Y = y(t) for a :::; t :::; b are functions
which have continuous first derivatives. Actually, one
can even allow the functions ~~ and !!If to have finitely
many discontinuities and still the integral makes sense.

It is clear from the change of variable formula for inte-

grals that the above notion of length is independent of

-RE-S-O-N-A-N-C-E-I-A-U-9-US-t--20-0-0-----------~-----------------------------73
 FEATURE I ARTICLE

which parametrisation we use. For this reason, if one is

given the curve in nonparametric form as y = I(x), then
one can write its length from x = a to x = b to be the in-
tegral I:
)1 + (~)2dx. If the curve is given in polar co-
ordinates as x = r(f))cosf), y = r(f))sinf), the length can,
therefore, be written as the integral IIo! Vr2 + (~~)2df)
where r( f)o)cosf)o = a, r (f)1)cosf)1 = b.
Let us look at the circle of radius ro; its polar equation
is r = ro (constant); the integral gives the length (f)1 -
f)o)ro.

Already, it might be clear that even if a sequence of

functions In defined on some interval (a, b) converges in
the interval to a function I, the corresponding integrals
I: In(x)dx need not converge to I:
I(x)dx. The same
problem persists with the integrals of their derivatives
etc.
This, for instance, provides the correct explanation of
one of the 'Think-it-over' problems: 'Where is the miss-
ing string?' posed in Resonance, Vol. 3, No.1, 1998.
The area under the string in its various positions can
get arbitrarily close to that under the diagonal string
without the arc lengths tending to the diagonal length.
Ellip.tic Integrals
After the brief digression to recall the fundamental con-
cept of length of a curve, let us come back a whole circle
viz., look at x 2 + y2 = 1. The length s = s(xo) from
(x,y) = (0,1) to (XO,yo) is given as I;o Jt-x
2 = sin-1xo.
We express this as x = sine s) and, viewed this way, the
addition formula for the sine function amounts to the
formula
{X dt {Y dt {Z dt
Jo V1 - t2 + 10 v'f=t2 = 10 V1 - t2

where z = XVI - y2 + YVI - x 2 .

What happens when we try to calculate the length of an

-74----------------------------~-----------------------------
RESONANCE I August 2000
 FEATURE I ARTICLE

arc in an ellipse Lemniscate can be

described as the
~+~=1?
locus of points, the
Such a length is given by an integral of the form product of whose
distances from two
8(XO) = fixed points is
constant.

where R is a rational function (i.e. a quotient of two

polynomials) and p is a polynomial of degree 4 in x.
Here e = ~ is the eccentricity of the ellipse.
Similarly, let us look at a lemniscate which can be de-
scribed as the locus of points, the product of whose dis-
tances from two fixed points is constant. It is given in
polar co-ordinates by the equation r2 = cos(20), the arc
length is given by the integral

dr
8(00) = Jofooo (r2 + (-)2)dO
dO
_ 10°0 cos(2B)-1/2dB
_ Io xO
(1 _ x4)-1/2dx.

The analogy to the circular length JtO(l - x2)-1/2dx

is evident. For the circle, one evaluates the integral
by rationalising the integrand i.e., by making a change
of variables which makes the integrand into a ratio-
nal function i.e., into a quotient of two polynomials.
One can rationalise the integrand by the substitution
x = 1!~2 (How?). Guided by this analogy in 1718,
Fagnano tried the substitution x = 1~:4 for the lem-
niscate. Although the integrand does not rationalise, he
noticed that J;O(l - x4)-1/2dx = J2J~O(l + t 4)-1/2dt

-RE-S-O-N-A-N-C-E-I-A-U-9-US-t--20-0-0-----------~---------------------------7-5
 FEATURE I· ARTICLE

Using a ruler and if 0 :::; x:::; 1. Following this up with the sub-
compass, one can stitution t = 1~~4' one obtains fcf°(l - x4)-1/2dx =
double the arc 2 f~O(l - u 4)-1/2du. Let us note the interesting con-
length of a sequence of this that using a ruler and compass, one can
lemniscate. double the arc length of a lemniscate - see [1] for a more
general discussion of this aspect.
A fundamental idea in the general problem of evaluat-
ing integrals of the form fcf° R(x, Vp(x) )dx is due to
Abel and Jacobi. This is simply to invert the problem!
So, one considers Xo as a function of s in the formula
s(xo) = fto R(x, Vp(x))dx. Note that the integrand is a
multi-valued function while however their inverse func-
tions arise as single-valued functions as Abel and Jacobi
realised in the 1820's. This yields addition formulae for
these functions akin to those for the trigonometric func-
tions. These integrals and, in general, the integrals of
the form s(xo) = fto R(x, V"P)dx are called elliptic inte-
grals where R is a rational function and p is a polynomial
of degree 3 or 4 in x (why did we leave out degrees 1 and
2?). The inverse functions x = x (s) are called elliptic
functions. They have properties analogous to those for
trigonometric functions; in place of the usual periodicity,
they are doublyl periodic!
Although we can't prove it her-e, the fundamental fact
involved here is that any cubic curve over the complex
numbers which is smooth i. e., has no singularities, is
biholomorphic to a complex torus. We digress now to
discuss a notion first introduced by Gauss and studied
in relation with elliptic integrals to get some beautiful
results.
The agM
Let a ~ b > 0 be two real numbers. Construct the
following sequence of arithmetic and geometric means:

al -- a+b b1 --
-2-' VrJ"b
ao, a2 -- al +b1
2 ,b
2 -- r::z:-b
Valul,
an~bn ,bn+1 = v' anb n , etc.

--------~--------
76 RESONANCE August 2000 1
 FEATURE I ARTICLE

For example, look at a = 2, b = 1. We have the following The fundamental

sequence of values correct to eight decimal places: fact involved here
is that any cubic
n an bn
curve over the
0 2 1
complex numbers
1 1.5 1.41421356
which is smooth is
2 1.45710678 1.45647531
biholomorphic to a
3 1.45679105 1.45679101
4 1.45679103 1.45679103 complex torus.

It is natural to guess that the two sequences approach a

common number. This is indeed true in general and is
seen quite easily as follows. From the familiar inequality
which asserts that the arithmetic mean is at least as
big as the geometric mean, one has an ;::: bn . In fact,
a ;::: al ;::: a2 ;::: ;::: an ;::: bn ;::: bn - 1 ;::: ;::: bI ;::: b.

Can one say more?

Indeed, it is easy to see that any of the a's is at least as
big as any of the b's (why?) i.e.,

;::: bI ;::: b.

Moreover, by mathematical induction, it follows that

an - bn < a2~b for all n ~ 1. What do these two observa-
tions give us? The first implies that the limits 1imn~ooan
and limn~oobn exist. The second implies that these two
limits are actually equall This common limit, denoted
by AI (a, b) is called the arithmetico-geometric mean and
denoted by agM, a notation introduced by Gauss him-
self. Some of the evident properties of the agM are:

JvJ(a, a) =a
M(a, b) = JvJ(a n , bn ) Vn
AI(ta, tb) = tM(a, b)

-RE-S-O-N-A-N-C-E-I--AU-9-U-st--2-0-00-----------~------------------------------n
 FEATURE I ARTICLE

agM and the Perimeter

Gauss proved the following beautiful result on the perime-

ter of an ellipse:
Theorem

Gauss's proof is ingenious and goes as follows. If J(a,b)

denotes the integral above, the key step is to show that
J(a, b) = J(ab bd for then, repeating this, and going to
the limit, one will get J(a, b) = J(M(a, b), M(a, b)) =
M(a, b)-l~, which is the assertion. The proof of the key
step is achieved in the following manner by making the
substitution
. 0
run = a + b +2asin</>
(a - b)sin2</>
.
This is now called Gauss's t.ransformation.
The change of variables gives easily that
2
2 20 b . 0 2 (a+b-(a-b)sin </»2
a cos + 2sIn 2 = a
(a + b + (a - b) sin 2</> ) 2
Let us differentiate both sides of
a + b - (a - b) sin 2 </>
(a 2cos 20 + b2sin20)1 / 2 = a (~)
a + b + (a - b) sin2</>
to compute the integral in terms of the new variable </>.
Derivative of the left hand side of (~) is
(a 2cos 20 + b2sin 20)-1/2(b 2 - a2)sinOcosOdO.
Differentiating the right hand side of (~), one gets the
expression
2 2
4a(b - a )sin¢cos</> d</>.
(a + b + (a - b)sin 2</»2

-78--------------------------~----------R-E-S-O-NA-N-C-E-I--AU-9-US-t--20-0-0
 FEATURE I ARTICLE

Substituting for the denominator, this further simplifies

to 4a(b 2 - a2)sin¢cos¢ 4:J~i~<pd¢ i.e., to the expression

On the other hand, using a + b = 2a1 and ab = bi, one

has

2 2).4 _ (a+b)2-2(a 2 +b 2)sin 2¢+(a-b)2sin4¢

( a 1 -b 1 SIn ¢ - 4

(a + b + (a - b) sin 2¢ ) 2 - 4a 2sin 2¢ a2cos 2Bsin 2¢

4 sin 2B
Thus,

1/2 sinBcos¢
(a 2cos 2¢ + b21sin 2¢)- = ..
1 acosBsIn¢

Mutiplying this with (b 2 - a2)sinBcosBd¢ yields just the

derivative of the right hand side of (~). In other words,
we have verified that

(a 2cos 2B + b2sin 2B)-1/2dB =

(aicos 2¢ + bisin2 ¢ )-1/2d¢.

From this, the key step follows immediately and so does
Gauss's theorem.

The agM M (a, b) can be expressed in terms of one pa-

rameter, viz., the eccentricity e = Va 2 - b2 / a of the
ellipse. Indeed, M (1 + e, 1 - e) = M (1, V1 - e 2 ) =
M(1,~) = lvI(a, b)/a. Therefore, the theorem can be
restated as
7r
M(1+e,1-e)= ()
2K e

-RE-S-O-N-A-N-CE--I-A-U9-U-st--2-00-0----------~--------------------------7-9
 FEATURE I ARTICLE

There is a transformation similar to the one used by

Gauss in the proof above. It is due to a little-known
English mathematician J Landen and has turned out to
be very fruitful in evaluating elliptic integrals. Landen's
transformation is the substitution esinO = sin(2</> - 0).
Exercise. Verify that this transformation can also be
used to prove the above theorem of Gauss.
Yet another result that can be proved by a repeated
usage of Landen's transformation is the following. For
2
a > b > 0, denote by J(a, b), the integral Jo7r/ (a 2cos 20 +
b2sin 20)1/2dO. Note that J(a, b) = aE(e) where E(e) =
J;/2(1 - e2sin 20)1/2dO. Then, we have:
Proposition
00

J(a,b) = (a 2 - L 2n-l(a~ - b~))I(a,b).

n=O

We shall return to this proposition in the next section.

For some other applications of Landen's transformation,
the interested reader may see [2].
In the case of a lemniscate, the total perimeter is

10" cos(20)-1/2dO = 4fo"/\2cos2a + sin 2a)-1/2da

from the change of variables

cos(20) = cos 2 o:.

So, in this case, the constants a and b in Gauss's theorem

are vf2 and 1, respectively. Therefore, one has

= ~ where w =
1
M(V2, 1)
2w
r (1-z
Jo
4
)-1/2dz. (0 )

Gauss's 92nd entry made in 1798 in his famous math-

ematical diary reads "1 have obtained most elegant re-
sults concerning the lemniscate, which surpasses all ex-
pectation - indeed, by methods which open an entirely

--------~--------
80 RESONANCE I August 2000
 FEATURE I ARTICLE

new field to us" The 98th entry made in May 1799 The fascination of
says that" demonstration of this fact (the identity 0) calculating
will open an entirely new field of analysis" What is approximations to
this 'new field of analysis' that is being talked about?
7t is said to have
When a and b are complex numbers, the square-root
existed among
causes trouble i.e., at each stage bn has two possible
mathematicians for
choices. Thus, M(a, b) is a multi-valued function and
as long as 4000
the various different values are related. Determining the
years - even
relations between them requires the so-called theory of
laymen have had
modular forms of weight 1 for congruence subgroups of
their fingers in the
SL(2, Z). This is what Gauss refers to as 'the new field
of analysis.' 'Tt!

Ways to Approximate 7r

The fascination of calculating approximations to 7r is

said to have existed among mathematicians for as long
as 4000 years - even laymen have had their fingers in the
7r! The study of 7r has led to deep questions - an exam-
ple is the ancient Greek problem of squaring the circle-
see the discussion of this problem and other related ones
in [3]. It was only at the end of the 19th century that
Lindemann (1852-1939) finally proved that 7r is tran-
scendental i.e., that there is no nonconstant polynomial
with rational coefficients that has 7r as a root.
Several great mathematicians have given'formulae' or
approximations. Perhaps, the earliest was Archimedes
(287-212 B.C.) who gave us the approximation 2;~ <
7r <2;. The notation 7r is also due to him. Aryabhatta
(476-550 A.D.) gave the approximation 7r rv ~~~~~ =
3.1416. Leibniz (1646-1716) a~d Euler (1707-1783) gave
the respective formulae:
7r 111
4=1-3"+5-"7+
2
7r 1 1 1
"8 = 1 + 32 + 52 + 72 +
Legendre gave the following formula for 7r in terms of the
elliptic integrals ([4] has a new and rather easy proof):

-RE-S-O-N-A-N-C-E-I-A-U-9-US-t--20-0-0-----------~-----------------------------81
 FEATURE I ARTICLE

Legendre's Formula

Let 0 < e < 1 and e' = y'1 - e 2 . Then

7r
K(e)E(e') + K(e')E(e) - K(e)K(e') = -.
2
Let us use this formula with the special value e = e' ==
1/ V2 to give an approximation for 7r which involves the
agM. This will be seen to be reminiscent of Archimedes's
method. His method was to approximate the unit cir-
cle by inscribed and circumscribed regular polygons.
Then, one has Pn < 27r < Qn where Pn , Qn are, re-
spectively, the perimeters of the inscribed and the cir-
cumscribed regular polygons of n sides. In fact, the
numbers an := I/Q2 n and bn = 1/ P 2n satisfy the re-
cursion an+l = (an + bn)/2, bn+1 = y'an+lbn. This is
very similar to the defining recursion for M(I/2,1/4).
However, Archimedes's sequences converge much more
slowly.
Now, Legendre's formula gives

On the other hand, the proposition above implies (with

a = 1, b = 1/V2) that

00

E(I/V2) = (1 - L: 2n-l(a~ - b~))K(l/V2).

n=O

Combining these two with Gauss's theorem, we have

00

7r(1 - L: 2i+l(a; - b;)) = 4M(1, I/Vi)2

i=l

Thus, one has the approximations

7r n = 4a~+1/ L~=l 2i+l (a; - b;) which converge quadrat-
ically to 7r. In other words, the number of correct digits
doubles at each step.

--------~--------
82 RESONANCE I August 2000
 FEATURE I ARTICLE

Finally, we end with a formula due to Ramanujan (see

[5]). Ramanujan used the theory of modular functions
(this was alluded to at the end of the last section) to give
some amazing formulae for 7r as well as for the perime-
ters of ellipses. One of these is:

_1__ ~ 1103 + 26390n (4n)!

27rV2 - ~ 99 4n+2 44n( n!)4 .

Using just the first term, one gets the approximation

0.112539536 ... to 2rrJ2 = 0.112539539 ...

Suggested Reading
Address for Correspondence
[1] B Sury, Resonance, Vol.4, No.12, p.48, 1999.
B Sury
[2] B Bemdt,Ramanujan's Notebooks, Part III, Springer-Verlag, 1991.
Statistics & Mathematics Unit
[3) Shashidhar Jagadeesan,Resonance, Vol. 4, No.3, 1999.
Indian Statistical Institute
[4) G Almkvist and B Bemdt, Amer. Math. Month(y,Vol.94 585-608, 1987.
Bangalore 560050, India.
[5] J H Borwein and P B Borwein,Pi and theAGM, Wiley, Canada, 1987.
Email: [email protected]
[6] D Cox,L~nseignementMathematique, VoUO, 275-330,1984.

~
The Square Root by Infinite Descent

One of the first instances of a proof by contradiction that one learns

I II
in school is that of the irrationality of v'2. Here is a proof due to
Tannenbaum by 'infinite descent' - the method employed by Fermat
and by Brahmagupta even earlier in other contexts. Suppose 2 = ~
where a, b are natural numbers. In the figure, ABCD is a square of
side a, PBQV and STRD are both squares of side b. As a 2 = 2b 2 ,
one has Area(TUVW) = Area(PWSA) + Area( QCRU). This means
2
(2b - a) = 2 (a - b)2 i.e., 2 = 12:~:r22. On the other hand, it is clear from
the figure that 0 < 2b - a < a which produces a smaller numerator than
the original one setting in a process of infinite descent. The resultant
contradiction forced on us proves the irrationality of v'2.

;gPTB
U Q
V
D R C Kanakku Puly

--------~--------
I
RESONANCE August 2000 83