JACOBIAN

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

The Jacobian

The Jacobian of a Transformation


In this section, we explore the concept of a "derivative" of a coordinate transformation, which is known as the Jacobian of the transformation. However, in this
course, it is the determinant of the Jacobian that will be used most frequently.
If we let u = hu; vi ; p = hp; qi, and x = hx; yi, then (x; y) = T (u; v) is given
in vector notation by
x = T (u)
This notation allows us to extend the concept of a total derivative to the total
derivative of a coordinate transformation.
Denition 5.1: A coordinate transformation T (u) is di erentiable
at a point p if there exists a matrix J (p) for which
lim

u!p

jjT (u)

T (p)
jju

J (p) (u
pjj

p)jj

=0

(1)

When it exists, J (p) is the total derivative of T (u) at p.


In non-vector notation, denition 5.1 says that the total derivative at a point
(p; q) of a coordinate transformation T (u; v) is a matrix J (u; v) evaluated at
(p; q) : In a manner analogous to that in section 2-5, it can be shown that this
matrix is given by
xu xv
J (u; v) =
yu yv
(see exercise 46). The total derivative is also known as the Jacobian Matrix of
the transformation T (u; v) :
EXAMPLE 1 What is the Jacobian matrix for the polar coordinate
transformation?
Solution: Since x = r cos ( ) and y = r sin ( ) ; the Jacobian matrix
is
xr x
cos ( )
r sin ( )
J (r; ) =
=
yr y
sin ( ) r cos ( )
If u (t) = hu (t) ; v (t)i is a curve in the uv-plane, then x (t) = T (u (t) ; v (t)) is
the image of u (t) in the xy-plane. Moreover,
dx
=
dt

dx
dt
dy
dt

dv
xu du
dt + xv dt
du
dv
yu dt + yv dt

xu
yu

xv
yv

du
dt
dv
dt

The last vector is du=dt: Thus, we have shown that if x (t) = T (u (t)) ; then
du
dx
= J (u)
dt
dt
That is, the Jacobian maps tangent vectors to curves in the uv-plane to tangent
vectors to curves in the xy-plane.

In general, the Jacobian maps any tangent vector to a curve at a given point to
a tangent vector to the image of the curve at the image of the point.
EXAMPLE 2

Let T (u; v) = u2

v 2 ; 2uv

a) Find the velocity of u (t) = t; t2 when t = 1:


b) Find the Jacobian and apply it to the vector in a)
c) Find x (t) = T (u (t)) in the xy-plane and then nd its velocity
vector at t = 1: Compare to the result in (b).
Solution: a) Since u0 (t) = h1; 2ti ; the velocity at t = 1 is u0 (1) =
h1; 2i :

b) Since x (u; v) = u2 v 2 and y (u; v) = 2uv; the Jacobian of T (u; v)


is
xu xv
2u
2v
J (u; v) =
=
yu yv
2v 2u
Since u0 = h1; 2ti ; we have
J (u; v) u0

2u
2v

2u (1) 2v (2t)
2v (1) + 2u (2t)

2u 4tv
2v + 4tu

2v
2u

1
2t

Substituting hu; vi = t; t2 yields


x0 = J (u; v) u0 =
In vector form, x0 (t) = 2t

2t 4t t2
2t2 + 4t (t)

4t3

2t
6t

4t3 ; 6t2 ; so that x0 (1) = h 2; 6i :

c) Substituting u = t; v = t2 into T (u; v) = u2


in
x (t) = t2 t4 ; 2t3
which has a velocity of x0 (t) = 2t
h 2; 6i :

v 2 ; 2uv results

4t3 ; 6t2 . Moreover, x0 (1) =

Check your Reading: At what point in the xy-plane is x0 (1) tangent to the
curve?

The Jacobian Determinant


The determinant of the Jacobian matrix of a transformation is given by
det (J) =

xu
yu

xv
yv

@x @y
@u @v

@x @y
@v @u

However, we often use a notation for det (J) that is more suggestive of how the
determinant is calculated.
@ (x; y)
@x @y
=
@ (u; v)
@u @v

@x @y
@v @u

The remainder of this section explores the Jacobian determinant and some of
its more important properties.
EXAMPLE 3

Calculate the Jacobian Determinant of


T (u; v) = u2

Solution: If we identify x = u2
@ (x; y)
@ (u; v)

=
=
=

v; u2 + v
v and y = u2 + v; then

@x @y @x @y
@u @v
@v @u
(2u) (1) ( 1) (2u)
4u

Before we consider applications of the Jacobian determinant, lets develop some


of the its properties. To begin with, if x (u; v) and y (u; v) are dierentiable
functions, then
@ (y; x)
@ (u; v)

=
=
=

@y @x @y @x
@u @v
@v @u
@x @y @x @y
@u @v
@v @u
@ (x; y)
@ (u; v)

from which it follows immediately that


@ (x; x)
@ (y; y)
=
=0
@ (u; v)
@ (u; v)
Similarly, if f (u; v) ; g (u; v) ; and h (u; v) are dierentiable, then
@ (f + g; h)
@ (u; v)

=
=
=
=

@ (f + g) @h @ (f + g) @h
@u
@v
@v
@u
@f @h @g @h @f @h @g @h
+
@u @v
@u @v
@v @u @v @u
@f @h @f @h
@g @h @g @h
+
@u @v
@v @u
@u @v
@v @u
@ (f; h) @ (g; h)
+
@ (u; v) @ (u; v)
4

The remaining properties in the next theorem can be obtained in similar fashion.
Theorem 5.2: If f (u; v) ; g (u; v) ; and h (u; v) are dierentiable functions
and k is a number, then
@(g;f )
@(f;g)
@(u;v) =
@(u;v)
@(f;f )
@(u;v) = 0
@(f;g)
@(kf;g)
@(u;v) = k @(u;v)

@(f +g;h)
@(f;h)
@(g;h)
@(u;v) = @(u;v) + @(u;v)
@(f g;h)
@(f;h)
@(g;h)
@(u;v) = @(u;v) + @(u;v)
@(f g;h)
@(f;h)
@(g;h)
@(u;v) = @(u;v) g + f @(u;v)

These and additional properties will be explored in the exercises.


EXAMPLE 4

Verify the property


@ (f g; h)
@ (f; h)
@ (g; h)
=
g+f
@ (u; v)
@ (u; v)
@ (u; v)

Solution: Direct calculation leads to


@ (f g; h)
@ (u; v)

=
=
=
=

@ (f g) @h @ (f g) @h
@u @v
@v @u
@f
@f
@g @h
@g @h
g+f
g+f
@u
@u @v
@v
@v @u
@f @h @f @h
@g @h @g @h
g+f
@u @v
@v @u
@u @v
@v @u
@ (g; h)
@ (f; h)
g+f
@ (u; v)
@ (u; v)

Check Your Reading: If k is constant and f (u; v) is dierentiable, then what


is
@ (k; f )
?
@ (u; v)

The Area Dierential


Let T (u; v) be a smooth coordinate transformation with Jacobian J (u; v) ; and
let R be the rectangle spanned by du = hdu; 0i and dv = h0; dvi : If du and dv
are su ciently close to 0, then T (R) is approximately the same as the parallelogram spanned by
dx = J (u; v) du = hxu du; yu du; 0i
dy = J (u; v) dv = hxv dv; yv dv; 0i

If we let dA denote the area of the parallelogram spanned by dx and dy; then
dA approximates the area of T (R) for du and dv su ciently close to 0.

The cross product of dx and dy is given by


dx

dy =

xu
yu

0; 0;

xv
yv

dudv

from which it follows that


dA = jjdx

dyjj = jxu yv

xv yu j dudv

(2)

Consequently, the area di erential dA is given by


dA =

@ (x; y)
dudv
@ (u; v)

(3)

That is, the area of a small region in the uv-plane is scaled by the Jacobian
determinant to approximate areas of small images in the xy-plane.
EXAMPLE 5 Find the Jacobian determinant and the area dierential of T (u; v) = u2 v 2 ; 2uv at (u; v) = (1; 1) ; What is the
approximate area of the image of the rectangle [1; 1:4] [1; 1:2]?
Solution: The Jacobian determinant is
@ (x; y)
@ (u; v)

=
=
=

@x @y @x @y
@u @v
@v @u
(2u) (2u) ( 2v) (2v)
4u2 + 4v 2

Thus, the area dierential is given by


dA =

@ (x; y)
dudv = 4u2 + 4v 2 dudv
@ (u; v)
6

On the rectangle [1; 1:4] [1; 1:2], the variable u changes by du = 0:4
and v changes by dv = 0:2. We evaluate the Jacobian at (u; v) =
(1; 1) and obtain the area
dA = (4 12 + 4 12 ) 0:4 0:2 = 0:32
which is the approximate area in the xy-plane of the image of [1; 1:4] [1; 1:2]
under T (u; v) :

Lets look at another interpretation of the area dierential. If the coordinate


curves under a transformation T (u; v) are su ciently close together, then they
form a grid of lines that are "practically straight" over short distances. As a
result, su ciently small rectangles in the uv-plane are mapped to small regions
in the xy-plane that are practically the same as parallelograms.

11 6 d
Consequently, the area dierential dA approximates the area in the xy-plane of
the image of a rectangle in the uv-plane as long as the rectangle in the uv-plane
is su ciently small.

EXAMPLE 6 Find the Jacobian determinant and the area dierential for the polar coordinate transformation. Illustrate using the
image of a "grid" of rectangles in polar coordinates.

Solution:
Since x = r cos ( ) and y = r sin ( ) ; the Jacobian
determinant is
@ (x; y)
@ (r; )

@x @y @x @y
@r @
@ @r
= cos ( ) r cos ( )
r sin ( ) sin ( )
2
2
= r cos ( ) + sin ( )
= r
=

Thus, the area dierential is dA = rdrd :


Geometrically, "rectangles" in polar coordinates are regions between
circular arcs away from the origin and rays through the origin. If
the distance changes from r to r + dr for r > 0 and some small
dr > 0; and if the polar angle changes from to + d for some
small angle d ; then the region covered is practically the same as a
small rectangle with height dr and width ds, which is the distance
from to + d along a circle of radius r.

If an arc subtends an angle d of a circle of radius r, then the length


of the arc is ds = rd . Thus,
dA = dr ds = rdrd

Check your Reading: Do "rectangles" in polar coordinates resemble rectangles if r is arbitrarily close to 0?
8

The Inverse Function Theorem


Recall taht if a coordinate transformation T maps an open region U in the uvplane to an open region V in the xy-plane, then T is 1-1 if each point in V is
the image of only one point in U:

Additionally, if every point in V is the image under T (u; v) of at least one point
in U; then T (u; v) is said to map U onto V:
If T (u; v) is a 1-1 mapping of a region U in the uv-plane onto a region V in
the xy-plane, then we dene the inverse transformation of T from V onto U by
T

(x; y) = (u; v) only if (x; y) = T (u; v)

The Jacobian determinant can be used to determine if T has an inverse transformation T 1 on at least some small region about a given point.
Inverse Function Theorem: Let T (u; v) be a coordinate transformation on an open region S in the uv-plane and let (p; q) be a
point in S: If
@ (x; y)
6= 0
@ (u; v) (u;v)=(p;q)
then there is an open region U containing (p; q) and an open region
V containing (x; y) = T (p; q) such that T 1 exists and maps V onto
U:
image
The proof of the inverse function theorem follows from the fact that the Jacobian
matrix of T 1 (x; y) ; when it exists, is given by the inverse of the Jacobian of
T,
1
@ (x; y)
yv
xv
J 1 (x; y) =
yu xu
@ (u; v)
9

which features a Jacobian determinant with a negative power. Thus, J


only if the determinant of J (u; v) is non-zero.
EXAMPLE 7

exists

Where is T (r; ) = hr cos ( ) ; r sin ( )i invertible?

Solution: The Jacobian determinant for polar coordinates is


@ (x; y)
=r
@ (r; )
which is non-zero everywhere except the origin. Thus, at any point
(r0 ; 0 ) with r0 > 0, there is an open region U in the r -plane and
an open region V containing (x; y) = (r0 cos ( 0 ) ; r0 sin ( 0 )) such
that T 1 (x; y) exists and maps V onto U:
We will explore the result in example 7 more fully in the exercises. In particular,
we will show that
!+
*
p
y
1
1
p
x2 + y 2 ; 2 tan
T (x; y) =
x + x2 + y 2

Clearly, T 1 is not dened on any open region containing (0; 0) : Also, if y = 0


and x > 0; then
2 tan

x+

0
x2 + 02

= 2 tan

But if y = 0 and x < 0; then


2 tan

x+

0
x2 + 02

= 2 tan

That is, a dierent representation of T


intersects the negative real axis.

0
x + jxj

0
x + jxj

=0

= 2 tan

0
0

must be used on any region which

Exercises
Find the velocity vector in the uv-plane to the given curve. Then nd Jacobian
matrix and the tangent vector at the corresponding point to the image of the
curve in the xy-plane.
1.
3.
5.

T (u; v) = hu + v; u vi
u = t; v = t2 at t = 1
T (u; v) = u2 v; uv 2
u = t; v = 3t at t = 2
T (u; v) = hu sec (v) ; u tan (v)i
u = t; v = at t = 1

2.
4.
6.

10

T (u; v) = h2u + v; 3u vi
u = t; v = t2 at t = 1
T (u; v) = u2 v 2 ; 2uv
u = cos (t) ; v = sin (t) at t = 0
T (u; v) = hu cosh (v) ; u sinh (v)i
u = t; v = t2 at t = 1

Find the Jacobian determinant and area di erential of each of the following
transformations.
7.
9.
11.
13.
15.
17.

T
T
T
T
T
T

(u; v) = hu + v; u vi
(u; v) = u2 v 2 ; 2uv
(u; v) = huev ; ue v i
(u; v) = h2u cos (v) ; 3u sin (v)i
(u; v) = heu cos (v) ; e u sin (v)i
(u; v) = hsin (u) sinh (v) ; cos (u) cosh (v)i

8.
10.
12.
14.
16.
18.

T
T
T
T
T
T

(u; v) = huv; u vi
(u; v) = u3 3uv 2 ; 3u2 v v 3
(u; v) = heu cos (v) ; eu sin (v)i
(u; v) = u2 cos (v) ; u2 sin (v)
(u; v) = heu cosh (v) ; e u sinh (v)i
(u; v) = hsin (uv) ; cos (uv)i

In each of the following, sketch several coordinate curves of the given coordinate
system to form a grid of "rectangles" (i.e., make sure the u-curves are close
enough to appear straight between the v-curves and vice-versa. Find the area
di erential and discuss its relationship to the "coordinate curve grid". (19 - 22
are linear transformations and have a constant Jacobian determinant)
19.

T (u; v) = h2u;
D vi

up v u+v
; p2
2

20.

21.
23.

T (u; v) =
parabolic coordinates

25.

T (u; v) = u2 v 2 ; 2uv
elliptic coordinates

22.
22.
24.

T (u; v) = hcosh (u) cos (v) ; sinh (u) sin (v)i

T (u; v) = D
hu + 1; vi
p

T (u; v) = u 2 3v ;
tangent coordinates
D

T (u; v) =

vi

by letting x = u + v; y = u v and solving for u and v: Then nd J


both (a) directly from T 1 (x; y) and (b) from the formula (4).
28. Find T 1 (x; y) for the transformation
T (u; v) = hu + 4; u

(x; y)

vi

Then nd J 1 (x; y) both (a) directly from T 1 (x; y) and (b) from the formula
(4).
29. At what points (u; v) does the coordinate transformation
T (u; v) = heu cos (v) ; eu sin (v)i
have an inverse? Can the same inverse be used over the entire uv-plane?
11

sin(u)
sinh(v)
cosh(v) cos(u) ; cosh(v) cos(u)

(x; y) for the transformation


T (u; v) = hu + v; u

3u+v
2

u
v
T (u; v) = u2 +v
2 ; u2 +v 2
bipolar coordinates
D

Some of the exercises below refer to the following formula for the inverse of the
Jacobian:
1
@ (x; y)
yv
xv
J 1 (x; y) =
(4)
yu xu
@ (u; v)
27. Find T

30. At what points (u; v) does the coordinate transformation


T (u; v) = hu cosh (v) ; u sinh (v)i
have an inverse.
31. Show that if T (u; v) = hau + bv; cu + dvi where a; b; c; d are constants
(i.e., T (u; v) is a linear transformation ), then J (u; v) is the matrix of the linear
transformation T (u; v) :
32. Show that if T (u; v) = hau + bv; cu + dvi where a; b; c; d are constants
(i.e., T (u; v) is a linear transformation ), then
@ (x; y)
= ad
@ (u; v)

bc

33. Show that if f (u; v) is dierentiable, then


@ (f; f )
=0
@ (u; v)
34. Show that if f (u; v) and g (u; v) are dierentiable and if k is constant,
then
@ (f; g)
@ (kf; g)
=k
@ (u; v)
@ (u; v)
35. Explain why if x > 0; then the inverse of the polar coordinate transformation is
Dp
y E
T 1 (x; y) =
x2 + y 2 ; tan 1
x
1
36. The Jacobian Matrix of (r; ) = T (x; y) is
K (x; y) =

rx

ry

Find K (x; y) for T 1 (x; y) in exercise 35, and then use polar coordinates to
explain its relationship to
J

(r; ) =

1
r

r cos ( ) r sin ( )
sin ( ) cos ( )

37. Show that if x < 0; then the inverse of the polar coordinate transformation is
Dp
y E
T 1 (x; y) =
x2 + y 2 ;
+ tan 1
x
38. Use the following steps to show that if (x; y) is not at the origin or on
the negative real axis, then
*
!+
p
y
1
1
p
T (x; y) =
x2 + y 2 ; 2 tan
x + x2 + y 2
is the inverse of the polar coordinate transformation.
12

a. Verify the identity


tan ( ) =

sin (2 )
1 + cos (2 )

b. Let = =2 in a. Multiply numerator and denominator by r:


c. Simplify to an equation in x; y; and :
39. The coordinate transformation of rotation about the origin is given by
T (u; v) = hcos ( ) u + sin ( ) v;

sin ( ) v + cos ( ) ui

where is the angle of rotation. What is the Jacobian determinant and area
dierential for rotation through an angle ? Explain the result geometrically.
40. The coordinate transformation of scaling horizontally by a > 0 and
scaling vertically by b > 0 is given by
T (u; v) = hau; bvi
What is its area dierential? Explain the result geometrically.
41. A transformation T (u; v) is said to be a conformal transformation if its
Jacobian matrix preserves angles between tangent vectors. Consider that the
vector h1; 0i is parallel to the line r = and that the vector h1; 1i is parallel to
the line r = : Also, notice that r = and r = intersect at (r; ) = ( ; ) at a
45 angle.

For J (r; ) for polar coordinates, calculate


v=J( ; )

1
0

and w = J ( ; )

1
1

Is the angle between v and w a 45 angle? Is the polar coordinate transformation


conformal?
42. Find the Jacobian and repeat exercise 41 for the transformation
T ( ; ) = he cos ( ) ; e sin ( )i
13

43. Write to Learn: Write a short essay in which you calculate the area
dierential of the transformation T ( ; ) = he cos ( ) ; e sin ( )i both computationally and geometrically.
44. Write to Learn: A coordinate transformation T (u; v) = hf (u; v) ; g (u; )i
is said to be area preserving if the area of the image of any region R in the uvplane is the same as the area of R: Write a short essay which uses the area
dierential to explain why a rotation through an angle is area preserving.
45. Proof of a Simplied Inverse Function Theorem: Suppose that
the Jacobian determinant of T (u; v) = hf (u; v) ; g (u; v)i is non-zero at a point
(p; q) and suppose that r (t) = hp + mt; q + nti ; t in [ "; "] ; is a line segment
in the uv-plane (m and n are numbers). Explain why if " is su ciently close
to 0, then there is a 1-1 correspondence between the segment r (t) and its image T (r (t)) ; t in [ "; "] : (Hint: rst show that x (t) = f (p + mt; q + nt) is
monotone in t for t in [ "; "] ).
46. Write to Learn: Let T (u; v) = hx (u; v) ; y (u; v)i be dierentiable at
p = (p; q) and assume that its Jacobian matrix is of the form
a b
c d

J=

By letting u = hp + h; qi in denition 5.1, (so that u


notation ), show that
lim

u!p

jT (u)

T (p)
jju

J (p) (u
pjj

p)j

p = [h 0] in matrix

=0

is transformed into
lim

h!0+

jjhx (p + h; q)

x (p; q) ; y (p + h; q)
h

y (p; q)i

hah; chijj

=0

Use this to show that a = xu and c = yu : How would you nd b and d? Explain
your derivations and results in a short essay.

14

You might also like