You are on page 1of 23

Introduction

Johann Bernoulli’s . . .
Euler’s Solution

Home Page

Title Page

The Brachistochrone Curve


JJ II

Paige MacDonald
J I

May 16, 2014


Page 1 of 21

Go Back

Full Screen

Close

Quit
Introduction
Johann Bernoulli’s . . .
Euler’s Solution

Home Page

Abstract
Title Page
This papers explores the brachistochrone problem, which is to find the path of quickest
descent of a particle moving between two fixed points in a vertical plane. A brief history
of the problem is given, followed by different approaches to solving the problem. JJ II

J I

Page 1 of 21

Go Back

Full Screen

Close

Quit
1. Introduction
The brachistochrone problem seeks to find the curve between two points, A and B, in
Introduction
a vertical plane and not in the same vertical line, along which a particle will slide in
the shortest amount of time under the force of gravity and neglecting friction. Johann Bernoulli’s . . .

Johann Bernoulli posed the problem in 1696. Solutions were found by Gottfried Wil- Euler’s Solution

helm von Leibniz, Isaac Newton, Guillaume de l’Hopital, Jacob Bernoulli, and Johann
Bernoulli himself. All of their answers agreed, although each used different methods of Home Page
derivation.
In 1744, Euler published a work generalizing the work done by the Bernoulli brothers
and came up with what is now known as the Euler-Lagrange differential equation (which Title Page
Lagrange later independently derived) in order to minimize the value of a definite
integral over a family of functions, which led to the calculus of variations.
Lagrange provided an analytic method to solve the brachistochrone problem and JJ II
other problems of its type, and introduced partial derivatives to the equation.

J I
2. Johann Bernoulli’s Solution
A creative solution like Johann Bernoulli’s uses an indirect approach, but results in Page 1 of 21
mathematics that are much easier to understand, so this will be a good starting point.
Consider a ray of light that travels from A to P with velocity v1 and then from P to
Go Back
B with velocity v2 as in Figure 1. Using the equation
distance
time =
velocity Full Screen

and solving for time leads to the following (in terms of the notation in Figure 1):
√ p Close
a2 + x2 b2 + (c − x)2
T = +
v1 v2
Quit
Introduction

A Johann Bernoulli’s . . .
v1 Euler’s Solution

a
α1
Home Page

x P c−x
Title Page

α2 JJ II
v2
b
J I

Page 2 of 21

B
c Go Back

Figure 1: A ray of light travels from A to P with velocity v1 and then from P to B
with velocity v2 . Full Screen

Close

Quit
Taking the derivative of each side gives:
√ p !
dT d a2 + x2 b2 + (c − x)2
= + Introduction
dx dx v1 v2
Johann Bernoulli’s . . .
1 2 2 − 12 1 2 2 − 21
2 (a
+ x ) (2x) + (c − x) ) (2)(c − x)(−1)
2 (b
Euler’s Solution
= +
v1 v2
x c−x
= √ − p Home Page
v1 a2 + x2 v2 b2 + (c − x)2

In order to minimize T , set dT /dx = 0. So, Title Page

x c−x
0= √ − p
2
v1 a + x 2 v2 b + (c − x)2
2
(1) JJ II
x c−x
√ = p
2
v1 a + x 2 v2 b + (c − x)2
2
J I
Referring back to Figure 1, it can be seen that
x
sin α1 = √ (2) Page 3 of 21
a2+ x2
and
c−x Go Back
sin α2 = p (3)
b2 + (c − x)2
Substituting equations (2) and (3) into equation (1) results in the following: Full Screen

sin α1 sin α2
= (4)
v1 v2 Close

Quit
Introduction
Johann Bernoulli’s . . .
Euler’s Solution

Home Page
v1 α1

Title Page

v2 α2 α2
JJ II
α3
v3 α3
J I

v4 α4 Page 4 of 21

Figure 2 Go Back

Full Screen

Close

Quit
Equation (4) is known as Snell’s law of refraction and demonstrates Fermat’s prin-
ciple of least time, which says that light travels from one point to another along the
path requiring the least time. If the layers are divided as in Figure 2, the velocity
Introduction
of light in each layer is constant, but decreases from one layer to the layer below it.
Johann Bernoulli’s . . .
Applying Snell’s law gives
Euler’s Solution
sin α1 sin α2 sin α3 sin α4
= = =
v1 v2 v3 v4
Home Page
As the layers are divided into smaller and smaller sections, the path approaches a
smooth curve and the velocity decreases continuously so that
sin α Title Page
=c (5)
v
where c is a constant.
Returning to the Brachistochrone problem, assume that the particle moves along a JJ II
curve from A to B in the least possible time as in Figure 3. Due to the assumption that
there is no friction, the total energy of the particle at point A is that same as the total
J I
energy at B, which is the kinetic energy plus the potential energy. Since the particle
starts from rest,
Ki + Ui = Kf + Uf Page 5 of 21
1
0 + mgy = mv 2 + 0
2
1 Go Back
mv 2 = mgy.
2
Solving for velocity results in
Full Screen
p
v = 2gy. (6)
From Figure 3,
1 1 1 Close
sin α = cos β = =p =p . (7)
sec β 2
1 + tan β 1 + (y 0 )2
Quit
x
A Introduction
Johann Bernoulli’s . . .
Euler’s Solution

y Home Page

Title Page

β JJ II
α
B
J I
y

Figure 3 Page 6 of 21

Substituting equations (6) and (7) into equation (5) results in: Go Back

Full Screen

Close

Quit
1
√ p =c
2gy 1 + (y 0 )2
p p 1
2gy 1 + (y 0 )2 = Introduction
c Johann Bernoulli’s . . .
√ p 1
y 1 + (y 0 )2 = √ Euler’s Solution
c 2g
1
y[1 + (y 0 )2 ] = 2 Home Page
2c g
Noting that the right hand side of the equation will be replaced by some other arbitrary
constant, the differential equation of the brachistochrone is Title Page

y[1 + (y 0 )2 ] = C (8)
JJ II
Using Leibniz notation, the differential equation of the brachistochrone becomes:
"  2 #
dy J I
y 1+ =C
dx
 2 Page 7 of 21
dy C
1+ =
dx y
 2
dy C Go Back
= −1
dx y
 2
dy C −y Full Screen
=
dx y
 1/2
dy C −y
= Close
dx y

Quit
Separating variables and solving for dx results in
 1/2
y
dx = dy (9) Introduction
C −y
Johann Bernoulli’s . . .

Introducing a new variable and setting Euler’s Solution

 1/2
y
= tan φ, (10) Home Page
C −y

and then solving for y gives Title Page


y
= tan2 φ
C −y
JJ II
y = (C − y) tan2 φ
y = C tan2 φ − y tan2 φ
J I
y(1 + tan2 φ) = C tan2 φ
y sec2 φ = C tan2 φ
Page 8 of 21
y = C sin2 φ.

Taking the derivative of each side with respect to φ leads to


Go Back

dy = 2C sin φ cos φdφ. (11)


Full Screen
Next, substitute equations (10) and (11) into equation (9) to obtain

dx = (tan φ)(2C sin φ cos φdφ)


Close
dx = 2C sin2 φdφ

Quit
Integrating both sides results in
Z Z
dx = 2c sin2 φdφ
Introduction
1 − cos 2φ
Z
Johann Bernoulli’s . . .
x = 2C dφ
2 Euler’s Solution
Z Z
x = C 1dφ − C cos 2φdφ
Home Page
1
x = Cφ − c sin 2φ + c1
2
C Title Page
x = (2φ − sin 2φ) + c1
2
Using the initial condition (0, 0) in equation (10) and solving for φ gives φ = 0 when JJ II
the particle is at point A. Therefore, c1 is 0 and
C
x= (2φ − sin 2φ) (12) J I
2
and
y = C sin2 φ Page 9 of 21
 
1 − cos 2φ
y=C (13)
2
Go Back
C
y = (1 − cos 2φ).
2
Letting a = C/2 and θ = 2φ, equations (12) and (13) become Full Screen

x = a(θ − sin θ) (14)


and Close

y = a(1 − cos θ). (15)


Quit
Introduction
Johann Bernoulli’s . . .
Euler’s Solution

y
Home Page

Title Page

JJ II

a J I
θ
x
2πa
(x, y) Page 10 of 21

Figure 4
Go Back

Full Screen

Close

Quit
These are the parametric equations of the curve traced out by a point on the circum-
ference of a circle of radius a rolling along the x axis (the cycloid) with the condition
that the curve passes through B for some value of a. The result is shown in Figure 4.
Introduction
Johann Bernoulli’s . . .

3. Euler’s Solution Euler’s Solution

Euler’s approach to solving this problem is more analytic and involves minimizing a
Home Page
function of a function, called a functional. Problems of this type belong to The Calculus
of Variations, which Euler’s work on this problem has been credited with leading to.
This is similar to optimization problems in elementary calculus, except in calculus the
Title Page
quantity that varies when finding extrema of f (x) is a simple variable, x. In calculus
of variations, the quantity that varies is itself a function of the form f (x, y, y 0 ). The
functional assigns a numerical value (time, in the case of the brachistochrone problem) JJ II
to each function in the family of candidate functions.
For the brachistochrone problem, referring to Figure 5, suppose two points A and
B are in a vertical plane. An infinite number of curves can be drawn connecting the J I
points, but which of these curves minimizes the time for a frictionless particle to move
from point A to point B under gravity? The problem, then, is to find a function that
minimizes the family of functions that are differentiable and whose derivatives and Page 11 of 21

second derivatives are continuous and pass through the points A and B-these are the
candidate, or admissable, functions. If the coordinates of points A and B are (x1 , y1 )
Go Back
and (x2 , y2 ), as in Figure 6, consider the family of functions y = y(x) that satisfy the
conditions y(x1 ) = y1 and y(x2 ) = y2 . The function that will minimize time is of the
form Z x2 Full Screen
I(y) = f (x, y, y 0 )dx. (16)
x1

Assuming a function y(x) that minimizes the integral in equation (16) exists, consider Close

a function η(x) that “disturbs” y(x) slightly. If y(x) minimizes I, then I will inrease
Quit
Introduction
Johann Bernoulli’s . . .
y Euler’s Solution

Home Page

A
Title Page

B
JJ II

J I

Page 12 of 21
x

Go Back

Figure 5
Full Screen

Close

Quit
as a result.
Let η(x) be a function with η 00 (x) continous and η(x1 ) = η(x2 ) = 0. If α is a small
parameter, then
Introduction
ȳ = y(x) + αη(x) (17)
Johann Bernoulli’s . . .
repesents a family of admissable functions of one-parameter with a vertical deviation Euler’s Solution
from the minimizing curve y(x), as seen in Figure 6. For any choice of the function
η(x), the function y(x) belongs to the family and corresponds to α = 0. Substituting
ȳ(x) = y(x) + αη(x) and ȳ 0 (x) = y 0 (x) + αη 0 (x) into the integral (16) gives a function Home Page

of α. Z x2
I(α) = f (x, ȳ, ȳ 0 )dx Title Page
x1
Z x2 (18)
I(α) = f [x, y(x) + αη(x), y 0 (x) + αη 0 (x)]dx.
x1 JJ II
When α = 0, equation (17) gives ȳ(x) = y(x). Since y(x) minimizes the integral (16),
I(α) must have a minimum when α = 0, which means that the derivative I 0 (α) when J I
α = 0 is 0, or I 0 (0) = 0. To compute the derivative I 0 (α), differentiate (18) under the
integral sign. Z x2

I 0 (α) = f (x, ȳ, ȳ 0 )dx (19) Page 13 of 21
x1 ∂α
Using the chain rule
Go Back

∂ ∂f ∂x ∂f ∂ ȳ ∂f ∂ ȳ 0
f (x, ȳ, ȳ 0 ) = + + 0 (20)
∂α ∂x ∂α ∂ ȳ ∂α ∂ ȳ ∂α Full Screen
0 0
Now, ∂x/∂α = 0, ∂ ȳ/∂α = η(x) and ∂ ȳ /∂α = η (x), so equation (20) becomes
Close
∂f ∂f ∂f
f (x, ȳ, ȳ 0 ) = η(x) + 0 η 0 (x)
∂α ∂ ȳ ∂ ȳ
Quit
Introduction
y Johann Bernoulli’s . . .
Euler’s Solution
(x1 , y1 )
ȳ(x) = y(x) + αη(x)
Home Page

αη(x)
y(x) Title Page
(x2 , y2 )

JJ II

J I

η(x) Page 14 of 21
η(x)
x
x1 x x2 Go Back

Full Screen
Figure 6

Close

Quit
and equation (19) can be written as
Z x2  
0 ∂f ∂f 0
I (α) = η(x) + 0 η (x) dx. (21) Introduction
x1 ∂ ȳ ∂ ȳ
Johann Bernoulli’s . . .
0 0 0
Setting α = 0 and recalling that when α = 0, I (α) = 0, ȳ = y, and ȳ = y ,equation (21) Euler’s Solution
gives Z x2  
∂f ∂f 0
η(x) + 0 η (x) dx = 0 (22) Home Page
x1 ∂y ∂y
Integrating the the second term by parts to eliminate η 0 (x) yields
Z x2  x2 Z x2   Title Page
∂f 0 ∂f d ∂f
0
η (x)dx = η(x) − η(x) dx
x1 ∂y ∂y 0 x1 x1 dx ∂y 0
Z x2   JJ II
d ∂f
=− η(x) dx
x1 dx ∂y 0
J I
due to the previously stated conditions that η(x1 ) = η(x2 ) = 0. Placing the result back
into equation (22) gives
Z x2    Page 15 of 21
∂f d ∂f
η(x) − η(x) dx = 0
x1 ∂y dx ∂y 0
Z x2    (23) Go Back
∂f d ∂f
η(x) − dx = 0.
x1 ∂y dx ∂y 0
Full Screen
The integral in equation (23) must vanish for every choice of the function η(x), which
implies that the expression in brackets must also vanish. Therefore,
Close
 
d ∂f ∂f
− = 0. (24)
dx ∂y 0 ∂y
Quit
Equation (24) is Euler’s equation. So, if y(x) is an admissable function that min-
imizes the integral (16), then y(x) satisfies Euler’s equation. However, the converse
is not true. That is, if y(x) is an admissable function satisfying Euler’s equation, it
Introduction
does not necessarily minimize the integral. The conditions for distinguishing admiss-
Johann Bernoulli’s . . .
able solutions of Euler’s equation, called stationary functions or stationary curves, are
Euler’s Solution
complicated and will not be explained here. Rather, the geometry of the problem will
be used to determine whether a particular stationary function minimizes the integral.
By expanding the first term in (24) according to the chain rule, Home Page

∂f dy 0
     
∂ ∂f ∂ ∂f dy ∂ ∂f
+ + − = 0,
∂x ∂y 0 ∂y ∂y 0 dx ∂y 0 ∂y 0 dx ∂y Title Page

Euler’s equation becomes


JJ II
dy dy 0
fy0 x + fy0 y + fy0 y0 − fy = 0
dx  dx
dy d dy J I
fy0 x + fy0 y + fy0 y0 − fy = 0 (25)
dx dx dx
d2 y dy
fy 0 y 0 2 + fy 0 y + fy0 x − fy = 0 Page 16 of 21
dx dx

where fy refers to ∂f ∂
∂y and fyx refers to ∂x fy . Go Back
An equation like (25) is usually unsolvable, but there are special cases where it is
solvable. Namely, if x and y are missing from the function f , if y is missing from
the function f , or if x is missing from f , then equation (25) is solvable. For the Full Screen
brachistochrone problem, the case of interest is the latter case, where x is missing from
the function f .
Close

Quit
3.1. Special Case
If x is missing from f , equation (25) can be integrated to
Introduction
∂f
0 Johann Bernoulli’s . . .
y − f = c. (26)
∂y 0 Euler’s Solution

This equation is known as Beltrami’s Identity. To show that this is true, begin by
rearranging Euler’s equation (24) and multiplying by y 0 to obtain Home Page

∂f d ∂f
y0 = y0 . (27)
∂y dx ∂y 0 Title Page

Now, according to the chain rule,


JJ II
df ∂f 0 ∂f 00 ∂f
= y + 0y + .
dx ∂y ∂y ∂x
J I
Rearranging this result yields
∂f 0 df ∂f ∂f
y = − 0 y 00 − . Page 17 of 21
∂y dx ∂y ∂x

Substituting this into the left hand side of (27) gives


Go Back
df ∂f 00 ∂f d ∂f
− y − = y0
dx ∂y 0 ∂x dx ∂y 0
Full Screen

and, rearranging this result leads to


df ∂f ∂f d ∂f Close
− 0 y 00 − − y0 = 0. (28)
dx ∂y ∂x dx ∂y 0
Quit
Noting that, by the product rule
   
d ∂f d d ∂f
f − y0 0 = f− y0 0 Introduction
dx ∂y dx dx ∂y
  Johann Bernoulli’s . . .
d ∂f df ∂f d ∂f
f − y0 0 = − y 00 0 − y 0 , Euler’s Solution
dx ∂y dx ∂y dx ∂y 0

then,   Home Page


0d ∂f df ∂f d ∂f
y = − y 00 0 − f −y 0
.
dx ∂y 0 dx ∂y dx ∂y 0
Title Page
Plugging this result for the last term on the left hand side of (28) gives
  
df ∂f ∂f df ∂f d ∂f
− 0 y 00 − − − y 00 0 − f − y0 0 =0 JJ II
dx ∂y ∂x dx ∂y dx ∂y
 
∂f d ∂f
− − f − y0 0 = 0 (29)
∂x dx ∂y J I
 
d ∂f ∂f
f − y0 0 = − .
dx ∂y ∂x
Page 18 of 21

Since ∂f /∂x = 0 for this case, integrating (29) yields

∂f Go Back
f − y0 =c
∂y 0

which is equation (26). Full Screen

Now, referring back to equation (16), the integral we wish to minimize is


Z x2 Close
I(y) = f (x, y, y 0 )dx.
x1

Quit
It is useful to invert the y-axis and place point A at the origin, as in Figure 3. The
velocity of the particle is√v = ds/dt, so the total time of descent p is the integral of ds/v.
From equation (7), v = 2gy and length of the curve s = 1 + (y 0 )2 , so the integral
Introduction
to be minimized becomes Z x2 p Johann Bernoulli’s . . .
1 + (y 0 )2
√ dx Euler’s Solution
x1 2gy
p √
Since f = 1 + (y 0 )2 / 2gy is missing x, referring to Case C and Beltrami’s identity,
Euler’s equation becomes Home Page
p ! p
∂ 1 + (y 0 )2 0 1 + (y 0 )2
0
√ y − √ =c
∂y 2gy 2gy Title Page
p
1 + (y 0 )2
 
1 1 0 2 −1/2 0 0
√ [1 + (y ) ] 2y y − √ =c
2gy 2 2gy JJ II
p
(y 0 )2 1 + (y 0 )2 p
√ p − √ = c 2g
y 1 + (y ) 0 2 y J I
(y ) − (1 + (y 0 )2 )
0 2 p
√ p = c 2g
y 1 + (y 0 )2 Page 19 of 21
−1 p
√ p = c 2g
y 1 + (y ) 0 2
Go Back
√ p 1
y 1 + (y 0 )2 = − √
c 2g
1 Full Screen
y[1 + (y 0 )2 ] = 2
2c g
and, replacing 1/2c2 g with an arbitrary constant, Close
0 2
y[1 + (y ) ] = C
Quit
which is the differential equation (8) solved in the first section. The resulting curve is
the cycloid generated by a circle with radius a rolling under the x axis whose first arch
passes through the point (x2 , y2 ).
Introduction
Johann Bernoulli’s . . .

4. Conclusion Euler’s Solution

The brachistochrone problem’s significance lies in the fact that it led to the calculus
Home Page
of variations, which is a branch of analysis that is applicable to a large number of
problems. It has greatly influenced mechanics, leading to Hamilton’s principle, as well
as being used in Einstein’s theory on general relativity. Shrodinger used it to discover Title Page
the wave equation, a cornerstone of quantum mechanics.

JJ II
References
[1] A special thanks to Dave Arnold for all his help and support. J I
[2] Till Tantau. The Tikz and PGF Packages.
[3] George Simmons. Differential Equations with Applications and Historical Notes. Page 20 of 21

[4] David Arnold. Writing Scientific Papers in Latex.


Go Back
[5] Douglas S. Shafer. 2007 The Brachistochrone: Historical Gateway to the Calculus
of Variations.
Full Screen
[6] Nils P. Johnson. 2009 The Brachistochrone Problem
[7] J J O’Connor and E F Robertson 2002 History Topic: The Brachistochrone Prob- Close
lem http://www.history.mcs.stand.ac.uk/PrintHT/Brachistochrone.html

Quit
[8] Unknown author. The Brachistochrone Problem: http://www.math.utk.edu/
~freire/teaching/m231f08/m231f08brachistochrone.pdf
[9] Frank Porter. Calculus of Variations: http://www.hep.caltech.edu/~fcp/math/ Introduction

variationalCalculus/variationalCalculus.pdf Johann Bernoulli’s . . .


Euler’s Solution

Home Page

Title Page

JJ II

J I

Page 21 of 21

Go Back

Full Screen

Close

Quit

You might also like