Jacobians and their applications

by Joseph Ruan, proud Member of the Math Squad.

Basic Definition

The Jacobian Matrix is just a matrix that takes the partial derivatives of each element of a transformation. In general, the Jacobian Matrix of a transformation F, looks like this:

F₁,F₂, F₃... are each of the elements of the output vector and x₁,x₂, x₃ ... are each of the elements of the input vector.

So for example, in a 2 dimensional case, let T be a transformation such that T(u,v)=<x,y> then the Jacobian matrix of this function would look like this:

$ J(u,v)=\begin{bmatrix} \frac{\partial x}{\partial u} & \frac{\partial x}{\partial v} \\ \frac{\partial y}{\partial u} & \frac{\partial y}{\partial v} \end{bmatrix} $

This Jacobian matrix noticably holds all of the partial derivatives of the transformation with respect to each of the variables. Therefore each row contains how a particular output element changes with respect to each of the input elements. This means that the Jacobian matrix contains vectors that help describe how a change in any of the input elements affects the output elements.

To help illustrate making Jacobian matrices, let's do some examples:

Example #1:

What would be the Jacobian Matrix of this Transformation?

$ T(u,v) = <u\times \cos v, u\times \sin v> $ .

Solution:

$ x=u \times \cos v \longrightarrow \frac{\partial x}{\partial u}= \cos v , \; \frac{\partial x}{\partial v} = -u\times \sin v $

$ y=u\times\sin v \longrightarrow \frac{\partial y}{\partial u}= \sin v , \; \frac{\partial y}{\partial v} = u\times \cos v $

$ J(u,v)=\begin{bmatrix} \frac{\partial x}{\partial u} & \frac{\partial x}{\partial v} \\ \frac{\partial y}{\partial u} & \frac{\partial y}{\partial v} \end{bmatrix}= \begin{bmatrix} \cos v & -u\times \sin v \\ \sin v & u\times \cos v \end{bmatrix} $

Example #2:

What would be the Jacobian Matrix of this Transformation?

$ T(u,v) = <u, v, u^v>,u>0 $ .

Solution:

Notice, that this matrix will not be square because there is a difference in dimensions of the input and output, i.e. the transformation is not injective.

$ x=u \longrightarrow \frac{\partial x}{\partial u}= 1 , \; \frac{\partial x}{\partial v} = 0 $

$ y=v \longrightarrow \frac{\partial y}{\partial u}=0 , \; \frac{\partial y}{\partial v} = 1 $

$ z=u^v \longrightarrow \frac{\partial y}{\partial u}= u^{v-1}\times v, \; \frac{\partial y}{\partial v} = u^v\times ln(u) $

$ J(u,v)=\begin{bmatrix} \frac{\partial x}{\partial u} & \frac{\partial x}{\partial v} \\ \frac{\partial y}{\partial u} & \frac{\partial y}{\partial v} \\ \frac{\partial z}{\partial u} & \frac{\partial z}{\partial v} \end{bmatrix}= \begin{bmatrix} 1 & 0 \\ 0 & 1 \\ u^{v-1}\times v & u^v\times ln(u)\end{bmatrix} $

Example #3:

What would be the Jacobian Matrix of this Transformation?

$ T(u,v) = <\tan (uv)> $ .

Solution:

Notice, that this matrix will not be square because there is a difference in dimensions of the input and output, i.e. the transformation is not injective.

$ x=\tan(uv) \longrightarrow \frac{\partial x}{\partial u}= \sec^2(uv)\times v, \; \frac{\partial x}{\partial v} = sec^2(uv)\times u $

$ J(u,v)= \begin{bmatrix}\frac{\partial x}{\partial u} & \frac{\partial x}{\partial v} \end{bmatrix}=\begin{bmatrix}\sec^2(uv)\times v & sec^2(uv)\times u \end{bmatrix} $

Application #1: Jacobian Determinants

The determinant of Example #1 gives:

$ \left|\begin{matrix} \cos v & -u \times \sin v \\ \sin v & u \times \cos v \end{matrix}\right|=~~ u \cos^2 v + u \sin^2 v =~~ u $

Notice that, in an integral when changing from cartesian coordinates (dxdy) to polar coordinates $ (drd\theta) $, the equation is as such:

$ dxdy=r\times drd\theta=u\times dudv $

It is easy to extrapolate, then, that the transformation from one set of coordinates to another set is merely

$ dC1=det(J(T))dC2 $

where C1 is the first set of coordinates, det(J(C1)) is the determinant of the Jacobian matrix made from the Transformation T, T is the Transformation from C1 to C2 and C2 is the second set of coordinates.

It is important to notice several aspects: first, the determinant is assumed to exist and be non-zero, and therefore the Jacobian matrix must be square and invertible. This makes sense because when changing coordinates, it should be possible to change back.

Moreover, recall from linear algebra that, in a two dimensional case, the determinant of a matrix of two vectors describes the area of the parallelogram drawn by it, or more accurately, it describes the scale factor between the unit square and the area of the parallelogram. If we extend the analogy, the determinant of the Jacobian would describe some sort of scale factor change from one set of coordinates to the other. Here is a picture that should help:

This is the general idea behind change of variables. It is easy to that the jacobian method matches up with one-dimensional change of variables:

$ T(u)=\begin{matrix}u^2\end{matrix}=\begin{matrix}x\end{matrix} $ $ ~~~,~~~~~~~~ J(u)=\begin{bmatrix}\frac{\partial x}{\partial u}\end{bmatrix}=\begin{bmatrix}2u\end{bmatrix} $$ ,~~~~~~~ dx=\left|J(u)\right|\times du=2u\times du $

Example #4:

Compute the following expression:

$ \iint(x^2+y^2)dA $

where dA is the region bounded by y²-x²=9, y²-x²=16, 2xy=14, 2xy=6.

Solution:

First, let's graph the region.

$ Let ~~~~ v=2xy, ~~~~ u=y^2-x^2 $.

Why do we do this? Let's look at the bounds: they're from y²-x² =9 to y² -x²=16 and 2xy=6 and 2xy=14.It'd be quite a simple task if the integral looked something like this:

$ \int^{16}_9 \int^{14}_6 k(u,v)\mathrm{d}u\mathrm{d}v $

$ u=2xy\longrightarrow \frac{\partial u}{\partial x}= 2y , \; \frac{\partial u}{\partial y} = 2x $

$ v=y^2-x^2\longrightarrow \frac{\partial v}{\partial x}= -2x , \; \frac{\partial v}{\partial y} =2y $

$ J(u,v)=\begin{bmatrix} \frac{\partial u}{\partial x} & \frac{\partial u}{\partial y} \\ \frac{\partial v}{\partial x} & \frac{\partial v}{\partial y} \end{bmatrix}= \begin{bmatrix} 2y & 2x \\ -2x & 2y\end{bmatrix} $

$ dxdy=\left|J(u,v)\right| \times dudv =\left|\begin{matrix} 2y & 2x \\ 2x & -2y\end{matrix}\right|\times dudv=(4y^2+4x^2)dudv $

The new region looks like this:

Therefore,

$ \iint(x^2+y^2)dA=\int^{16}_9 \int^{14}_6 (x^2+y^2)/(4y^2+4x^2)\mathrm{d}u\mathrm{d}v=\int^{16}_9 \int^{14}_6 1/4\mathrm{d}u\mathrm{d}v=(16-9)\times (14-6)\times 1/4=7/2. $

Example #5:

Compute the following expression:

$ \iint(1/\sqrt{4x^2+9y^2})dA $

where dA is the region bounded by (2x)²+(3y)²=36.

Solution:

Again, let's graph the region.

There are many ways to approach this, but to start, let's choose the change of variables to be u=2x, v =3y. Why?

$ (2x)^2+(3y)^2=36\longrightarrow u^2+v^2=36 $

The region now looks like this:

We've turned the region from an ellipse to a circle!

$ u=2x\longrightarrow \frac{\partial u}{\partial x}= 2 , \; \frac{\partial u}{\partial y} = 0 $

$ v=3y\longrightarrow \frac{\partial v}{\partial x}=0 , \; \frac{\partial v}{\partial y} =3 $

$ J(u,v)=\begin{bmatrix} \frac{\partial u}{\partial x} & \frac{\partial u}{\partial y} \\ \frac{\partial v}{\partial x} & \frac{\partial v}{\partial y} \end{bmatrix}= \begin{bmatrix} 2 & 0 \\ 0 & 3\end{bmatrix} $

$ dxdy=\left|J(u,v)\right| \times dudv =\left|\begin{matrix} 2 & 0 \\ 0 & 3\end{matrix}\right|\times dudv=(6)dudv $

$ \iint(1/\sqrt{4x^2+9y^2})dA =\iint 6/\sqrt{u^2+v^2}dudv $

Now, let's change the coordinates to polar:

$ r\sin \theta = u, r\cos\theta = v, \iint 6/\sqrt{u^2+v^2}dA=\iint 6/\sqrt{r^2}\times r\times drd\theta=\int^6_0 \int^{2\pi}_0 6 dr d\theta=6\times 36\times \pi=216 pi $

Application #2: Transforming Tangent vectors

The second major application of Jacobian Matrices comes from the fact that it is made of the partial derivatives of the transformation with respect to each original element. Therefore, when a function F is transformed by a transformation T, the tangent vector of F at a point is likewise transformed by the Jacobian Matrix.

Here is a picture to illustrate this. Note: this picture was taken from etsu.edu.

.

In other words, if we transform a function, we can find the new tangent vector at a transformed point.

Example #6:

$ <u,v>=<t^2,t^4>,~~~T(u,v)=<u^3-v^3,3uv>=<x,y> $

Find the velocity vector of T(u,v) at t=1.

Solution:

$ x=u^3-v^3\longrightarrow \frac{\partial x}{\partial u}= 3u^2 , \; \frac{\partial x}{\partial v} = -3v^2 $

$ y=3uv\longrightarrow \frac{\partial y}{\partial u}=3v , \; \frac{\partial y}{\partial v} =3u $

$ J(u,v)=\begin{bmatrix} \frac{\partial x}{\partial u} & \frac{\partial x}{\partial v} \\ \frac{\partial y}{\partial u} & \frac{\partial y}{\partial v} \end{bmatrix}= \begin{bmatrix} 3u^2 & -3v^2 \\ 3v & 3u\end{bmatrix} $

$ u'(t)=\begin{bmatrix}2t\\4t^3\end{bmatrix} $

$ J(u,v)\times u'(t)=\begin{bmatrix}3u^2 & -3v^2 \\3v & 3u\end{bmatrix} \times \begin{bmatrix}2t\\4t^3\end{bmatrix}= \begin{bmatrix}3u^2\times 2t-3v^2\times 4t^3\\3v\times 2t+3u\times 4t^3\end{bmatrix}= \begin{bmatrix}3(t^2)^2\times 2t-3(t^4)^2\times 4t^3\\3(t^4)\times 2t+3(t^2)\times 4t^3\end{bmatrix}= \begin{bmatrix}6t^5-12t^{11}\\18t^5\end{bmatrix} $

Therefore at t=1, the transformed tangent vector is:

$ \begin{bmatrix}-6\\18\end{bmatrix} $

Application #3: Linearization of Systems of Differential Equations

There is one last major application of Jacobians: approximating systems of Differential equations. Let's take the general expression:

$ \vec x'=f(\vec x,t) $

Around a point, if f(x,t) is differentiable, the system can be approximated to:

$ \vec x' \approx A\vec x ,~~~~A=J(\vec x ,t_0) $

The reason this works is similar to the idea behind one dimensional differentials:

$ dy\approx f'(x)\times dx $

Why is this important? This can help us check stability for certain systems of differential equations.

Example #7:

Find the stability of the all the critical points of

$ \begin{bmatrix} x'\\ y'\\ z'\end{bmatrix}=\begin{bmatrix}z-x\\ 3/(2+x)-3y\\3y-z\end{bmatrix} $

Solution:

To find the critical points, we set x'=y'=z'=0.

$ z=x,~~~~ y=1/3x,~~~~ 3/(2+x)-x=0 \longrightarrow (x+3)(x-1)=0,~~~~ (1,1/3,1)~~~~ (-3,-1,-3) $

Now let's find the Jacobian at (1,1/3,1)

$ J(x',y',z')=\begin{bmatrix}-1 & 0 & 1\\ -3/(2+x)^2 & -3 & 0\\ 0 & 3 & -1 \end{bmatrix}=\begin{bmatrix}-1 & 0 & 1\\ -1/3 & -3 & 0\\ 0 & 3 & -1 \end{bmatrix}. $

The eigenvalues are -3.2055694304005904, -0.8972152847997048 ± 0.6654569511528134i. Since the real parts are all negative, this critical point is locally stable.

Now let's find the Jacobian at (-3,-1,-3)

$ J(x',y',z')=\begin{bmatrix}-1 & 0 & 1\\ -3/(2+x)^2 & -3 & 0\\ 0 & 3 & -1 \end{bmatrix}=\begin{bmatrix}-1 & 0 & 1\\ -3 & -3 & 0\\ 0 & 3 & -1 \end{bmatrix}. $

The eigen values are -4, -2.5 ± 1.65831i. Because all of the real parts are negative, this critical point is locally stable.

Example #8:

Find the stability of the all the critical points of

$ \begin{bmatrix} x'\\ y'\\ z'\end{bmatrix}=\begin{bmatrix}z+2y^2+x\\ y-z^{3/2}\\ \sin(z) \end{bmatrix} $

Solution:

The only critical point in this case is (0,0,0)

Now, let's do the Jacobian.

$ J(x',y',z')=\begin{bmatrix}1 & 2y & 1\\ 0 & 1 & 3/2*z^{1/2}\\ 0 & 0 & \cos(z) \end{bmatrix}=\begin{bmatrix}1 & 0 & 1\\ 0 & 1 & 0\\ 0 & 0 & 1 \end{bmatrix}. $

The eigenvalue is just 1. Since this is positive, the critical point is unstable.

Sources:

1. http://en.wikipedia.org/wiki/Jacobian_matrix_and_determinant

Back to Math Squad page