Arc Length, Area, and the Arcsine Function

Andrew M. Rockett
Long Island University
Greenvale, NY 11548
Mathematics Magazine, March 1983, Volume 56, Number 2, pp. 104110.

eometric interpretations of algebraic quantities provide an essential motivation

for elementary calculus. The calculus with analytic geometry titles of many
introductory texts allude to the reliance on pictorial images in the development
of the derivative and the definite integral. While geometric considerations dominate most
presentation of the trigonometric functions and their derivatives, there is a near universal
switch to algebraic methods for the introduction and study of the inverse trigonometric
functions (see [6], [3], and so on). This note shows how the geometric ideas used in the
definitions of the definite integral, the trigonometric functions, and their derivatives may
be continued in the discussion of the corresponding inverse functions. Reference will be
made to the relation of these geometric ideas to the development of the theory of elliptic
functions and to Eulers method of finding algebraic addition theorems for circular,
hyperbolic, and lemniscate sines.
Typical calculus versus the arcsine as arc length

Taking [6] and [3] as our models, we note that in the typical modern textbook, after
the definite integral has been defined, the applications include the area between two curves
and the arc length formula. Since few integration techniques are available, the arc length
b
problems are restricted to nice curves y 5 f sxd such that the integral ea !1 1 f9 sxd2 dx
is particularly simple and sometimes an authors apology is offered for the lack of
interesting applications (see [3], p. 429).
The introduction of the trigonometric functions follows a review of radian measure as
arc length measured from the point s1, 0d on the unit circle x2 1 y2 5 1. The sine and
cosine of a real number u are defined as the coordinates of the point sx, yd on the unit
circle u radians from s1, 0d (see FIGURE 1). Then the properties of sin u and cos u are
derived from the symmetries of the circle and the other trigonometric functions are
defined in terms of the sine and cosine. The derivatives of the sine and the cosine are
found as consequences of lim ssin udyu 5 1. This limit is established by equating arc
u 0

length along the edge of the unit circle with the area of the sector determined by the
arc sin Figure 2, u 5 2 ? area AOBd and then squeezing this area between two triangular
regions.
After studying the calculus of the six trigonometric functions s"f sxd"d, the corresponding
inverse functions s"f 21 sxd"d are sought by reversing the graphs s"y 5 f sxd" becomes
"x 5 f 21syd"d, making arbitrary choices for the principal values (see [6], pp. 2956),
and then calculating Dxs f 21sxdd from the identity f s f21sxdd 5 x.

FIGURE 1

FIGURE 2

FIGURE 3

FIGURE 4

In contrast to exploiting the definitions and properties of inverse functions, the arcsine
function can be approached in a more geometric way. Since sin u was defined as the ycoordinate of the point on the unit circle an arc length of u away from s1, 0d, a
geometric attempt to invert this function would ask given that the second coordinate of
the point is y, from what arc length did it come? There are two small answers to this
question (see Figure 3) and then infinitely more separated from each other by multiples
of 2p. The principal value of the inverse sine function may be introduced naturally as
the smallest distance from the stating point, s1, 0d, that will work, and this arc length
may be called the arc sine of y and written arcsin (y) or the inverse sine of y and
written "sin21syd." We will use the arcsine notation for the remainder of our
discussion. Thus the arcsine function has arcsin syd 5 u where 2 py2 u py2 and
sinu 5 y.
Since arcsin (y) is an arc length, the arc length formula ea !1 1 f9std2 dt can be applied
to f std 5 !1 2 t2 from t 5 0 to t 5 y (see FIGURE 4) to find that
b

E
E!
y

arcsinsyd 5

!1 1 f9std2 dt

0
y

11

t2
dt
1 2 t2
2

1
!1 2 t2

dt

(1)

and then, by the Fundamental Theorem of Calculus, it follows that

Dysarcsinsydd 5

1
!1 2 y2

Since arcsins1d 5 py2, we also have a simple example of an improper integral:

1
dt 5 lim
!1 2 t2
y1

1
!1 2 t2

dt 5

p
.
2

For the inverse tangent, a similar argument yields an arctangent function

arctanswd 5 u with principal value 2 py2 < u < py2 such that arctan(w) is the
distance along the unit circle from s1, 0d to the point sx, yd with yyx 5 w. Since
x 5 !1 2 y2, we can solve yyx 5 w to find y 5 wy!1 1 w2. From the arc length
formula (1) we have

wy!11w 2

arctanswd 5

!1 2 t2

dt.

Making the change of variable t 5 uy!1 1 u2, so that t 5 0 when u 5 0 and

t 5 wy!1 1 w2 when u 5 w, we find that

E
E

arctanswd 5

0
w

!1 1 u2

1
du
!1 1 u2s1 1 u2d

1
du,
1 1 u2

and so
Dwsarctanswdd 5

1
.
1 1 w2

| |

For the inverse secant of w (where w 1d, if we seek the smallest positive arc length
from s1, 0d to the point with first coordinate 1/w, we obtain an arcsecant function
arcsecswd 5 u with principal value 0 u < py2 and py2 < u p. For w > 1, the arc
length u is py2 2 f (see FIGURE 5) so we have
arcsecswd 5

p
2 arcsins1ywd
2

and so
Dwsarcsecswdd 5 2
5

1
!1 2 s1ywd2

21
w2

1
for w > 1.
w!w2 2 1

For w < 21, we have by symmetry that

arcsecswd 5 p 2 arcsecs2wd

and so
Dwsarcsecswdd 5 2Dwsarcsecs2wdd
and thus
1
w !w2 2 1

Dwsarcsecswdd 5

| |

for w > 1.

| |

(We omit discussion of the formula

arcsecswd 5

!w 2 21yw

1
dt
!1 2 t2

for w 1

FIGURE 5

FIGURE 6

as it results in a subtle improper integral calculation.)

The formulas for the inverse cosine, cotangent, and cosecant can be obtained by
similar arguments.
Area and the arcsine
Let us now study the arcsine as an area. Let 0 < y < 1 be given and let A denote the
area bounded by the line xstd 5 s!1 2 y2yyd ? t, the circle xstd 5 !1 2 t2, and the
x-axis t 5 0 (see FIGURE 6). Since arc length along the edge of the unit circle equals
twice the area of the sector determined by the arc, we have that
arcsinsyd 5 2 ? A

E1
?E
y

52?
52

!1 2 t2 2

0
y

!1 2 y2

2 dt

?t

!1 2 t2 dt 2 y!1 2 y2.

(2)

Differentiating (2), we have

Dysarcsinsydd 5 !1 2 y2 1
5

1
,
!1 2 y2

y2
!1 2 y2
4

as before.
Alternatively, our integral expression (2) for arcsin(y) provides a motivating example for
the integration by parts formula eu ? dv 5 u ? v 2 ev ? du. Letting u 5 !1 2 t2 and
dv 5 dt, we have

E !1 2 t dt 5 t!1 2 t 2 E t !12t2 t dt
s!1 2 t d 2 1 dt
5 t!1 2 t 2 E
!1 2 t
1
5 t!1 2 t 2 E !1 2 t dt 1 E
!1 2 t
2

2 2

dt

so

2 !1 2 t2 dt 5 t!1 2 t2 1
But then

E !112 t

dt.

arcsinsyd 5 2 ?

!1 2 t2 dt 2 y!1 2 y2

5 y!1 2 y 1
2

1
!1 2 y2
2 dt 2 y
0 !1 2 t

1
2 dt,
0 !1 2 t

as before.
This development of the arcsine as the area bounded by the x-axis, the circle
xstd 5 !1 2 t2 from t 5 0 to t 5 y, and the line from the origin to the point

s!1 2 y2, yd may be adapted to the hyperbola xstd 5 !1 1 t2 to find the inverse

hyperbolic sine sinh21syd 5 ln !1 1 y2 1 y . However, unlike the circle, there is no

simple relation between arc length and area for the hyperbola (see[4], 6 and 7; for a
representation of arc length on the hyperbola, see [1], 60, and the corresponding
geometric construction is 61). On the other hand, a similar procedure to develop a
lemniscate sine using the lemniscate instead of the circle must be carried out in terms
of arc length (see [4], 8).
Rationalization of the arcsine
We now turn to the problem of expressing the integrand in (1) as a rational function; that
is, we wish to express 1 2 t2 as the square of a rational function. If, in the two-variable
representation of Pythagorean triples,
(a2 2 b2d2 1 s2abd2 5 sa2 1 b2d2,

(3)

we let a 5 1 and b 5 u then

s1 2 u2d2 5 s1 1 u2d2 2 s2ud2

5

or
1 2 u2
1 1 u2

512

2u 2
,
1 1 u2

and so the substitution t 5 2uys1 1 u2d will result in

!1 2 t2 5

1 2 u2
1 1 u2

and
dt 5

s1 1 u2ds2d 2 s2uds2ud
1 2 u2
du
5
2
du.
2
s1 1 u2d
s1 1 u2d2

Then the indefinite integral e1y!1 2 t2 dt becomes

E 11 12 uu

1 2 u2

du 5 2 ?
2 ? 2
s1 1 u2d2

E 1 11 u

du

and the integrand has been rationalized. Solving t 5 2uys1 1 u2d for u, we find that
u 5 s1 !1 2 t2dyt. Equating t with sin u, we see that !1 2 t2 5 cos u and the
expression for u with the minus sign becomes
u 5 s1 2 cos udysin u 5 tansuy2d
and we have derived the "u 5 tansuy2d" substitution for the rationalization of
trigonometric integrals. (Compare this development with the typical it has been
discovered that ... treatment in [6], p. 368.)
Eulers sine sum formula
In the introductory section of [5], Siegel describes Fagnanos study of arc length on the
lemniscate (which follows our development of the arcsine of the circle) and speculates
that Fagnanos 1718 discovery of a geometric construction to double arc length on the
lemniscate resulted from his attempt to rationalize the integrand of the lemniscate sine.
Thirty-five years later, Euler extended Fagnanos doubling theorem to an algebraic
addition theorem for the lemniscate sine and he shortly thereafter generalized his
discovery to elliptic integrals. Siegel ([5], p. 10) describes the aim of his first chapter to
be the fuller understanding of Eulers result from the viewpoint of analytic functions on
their full domain of definition. We conclude our discussion of the arcsine with Eulers
algebraic addition theorem for the arcsine adapted from 585 and 586 of [2] (see also
[2], Caput VI for Eulers study of elliptic integrals; [4], Chapter 4; and [5], 2).
Let 2 py2 < t < py2 be a fixed angle and let 2 py2 < u, f < py2 be any two angles
with u 1 f 5 t. Since the sum u 1 f is constant, dsu 1 fd 5 0, and if we set
u 5 sin u and v 5 sin f, we can rewrite dsu 1 fd 5 0 as

1E

dt
1
!1 2 t2

dt
50
2
0 !1 2 t

and so we have the differential equation

du
dv
1
5 0.
2
!1 2 u
!1 2 v2

(4)

We seek an algebraic solution of (4) subject to the condition that u 1 f 5 t.

Eulers main observation was that if we begin with the symmetric second-order equation
u2 1 2auv 1 v2 5 K 2,

(5)

where a and K are constants, we can complete the square on the left side in either u or v.
In the first case, we have
u2 1 2auv 1 a2v2 5 K 2 1 sa2 2 1dv2,
so that
u 1 av 5 !K 2 1 sa2 2 1dv2,

(6)

while in the second, we have

v2 1 2auv 1 a2u2 5 K 2 1 sa2 2 1du2,
so that
v 1 au 5 !K 2 1 sa2 2 1du2.

(7)

If we differentiate equation (5), we find

2udu 1 2avdu 1 2audv 1 2vdv 5 0,
which becomes, after collecting terms and using (6) and (7),
!K 2

du
dv
1
5 0.
2
2
2
!K 1 sa2 2 1dv2
1 sa 2 1du

(8)

If we set a2 2 1 5 2K 2 then equation (8) is the same as (4) and so a 5 !1 2 K 2.

If we can solve (5) for K using this value of a, we will find an algebraic solution of
(4) that expresses the constant value sin(u 1 fd in terms of sin u and sin f.
Substituting our value for a into (5) and rearranging, we have

su2 1 v2d 2 K 2 5 22!1 2 K 2uv

and squaring both sides gives

su2 1 v2d2 2 2K 2su2 1 v2d 1 K4 5 4s1 2 K 2du2v2.

(9)

By setting a 5 u and b 5 v in equation (3) and solving for su2 1 v2d2, we can rewrite
(9) as

su2 2 v2d2 2 2K 2ssu2 2 v2d 2 2u2v2d 1 K4 5 0.

(10)

Since (10) is a quadratic in K 2, we can complete the square with respect to K 2 to obtain

sK 2 1 s2u2v2 2 su2 1 v2ddd2 5 ssu2 1 v2d 2 2u2v2d2 2 su2 2 v2d2.

(11)

By expanding, collecting terms, and removing common factors, the right side of (11)
may be rewritten as 4u2v2su2v2 2 su2 1 v2d 1 1d. Taking square roots, we find
K 2 5 su2 1 v2d 2 2u2v2 1 2uv!s1 2 u2ds1 2 v2d.

(12)
7

Regrouping the expression su2 1 v2d 2 2u2v2 as su2 2 u2v2d 1 sv2 2 v2u2d, we see
that the right side of (12) is a perfect square and so
K 5 u!1 2 v2 1 v!1 2 u2.

(13)

Changing back to angles, K is the constant sinsu 1 fd while !1 2 v2 5

!1 2 sin2f 5 cos f and !1 2 u2 5 cos u so we have the sine sum formula
sinsu 1 fd 5 sinu cos f 1 sin f cos u.

(14)

Since the fixed value of t played no part in our calculations, we have established
(14) for any t.
References
[1] Arthur Cayley, An Elementary Treatise on Elliptic Functions, 2nd ed., 1895.
Reprinted by Dover Pub., New York, 1961.
[2] Leonhard Euler, Institutiones Calculi Integralis, Volumen Primum, 1768. Reprint as
Leonhardi Euleri Opera Omnia, Series I, Opera Mathematica, Volumen XI, Engel et
Schlesinger, eds., B.G. Teubneri, Leipzig, 1913.
[3] Louis Leithold, The Calculus with Analytic Geometry, 4th ed., Harper & Row, New
York, 1981.
[4] A.I. Markushevich, The Remarkable Sine Functions (translated by Scripta Technica,
Inc.), Amer. Elsevier, New York, 1966.
[5] C.L. Siegel, Topics in Complex Function Theory, vol. I: Elliptic Functions and
Uniformization Theory (translated by Shenitzer and Solitar), Wiley-Interscience,
New York, 1969.
[6] George B. Thomas and Ross L. Finney, Calculus and Analytic Geometry, 5th ed.,
Addison-Wesley, Reading, 1979.