Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

0.2. Multivariable Functions

Multivariable function is a function whose value depends on more than one independent variable.

Partial Derivatives:

Consider a flat plate in the shape shown in the figure below. Let's say the temperature at any point (x,y)
on this plate is given by T = f(x,y). If we move our attention along the horizontal line segment from point
(a,b) to (a+h,b), notice that we are letting x vary while keeping y constant at a value of b. The average rate
of change of the temperature with respect to the horizontal distance x on this line segment is:

Here, as we are only letting x vary while keeping y at a constant value of b, we are really considering a
function of single variable x, such as g(x) = f(x,b).

1
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

We see from these discussions that it physically means rate of change of f with respect to x when y is
fixed.

Similar arguments

can be made for the partial derivative of f with respect to y.

To sum up, if f is a function of two variables z = f(x,y), then its partial derivatives are defined as follows:

- One simply pretends that y is a constant when differentiating with respect to x and vice verse.

To give a geometric interpretation of partial derivatives, let's recall that the equation z = f(x,y) represents
a surface (S) as shown in the next figure. If f(a,b) = c, then the point P(a, b, c) lies on this surface. When
we consider the partial derivative fx(a,b), we keep the y values fixed at b. By fixing y = b, we are
restricting our attention to the curve C1 formed by the intersection of the vertical plane y = b with the
surface S. This curve f(x,b) is a function of x only (such as g(x)) and passes from the point P(a, b, c). The
partial derivative fx(a,b) (is the ordinary derivative of the curve with only x being the independent
variable g(x)) then gives the slope of the tangent line (T1) to C1 at point P (a, b, f(a,b)).

Similarly, for the partial derivative fy(a,b), we vary only the y values, while keeping x fixed at a. By doing
this we are restricting our attention to the curve on this surface formed by the intersection of the x = a
vertical plane with the surface formed by f(x,y). Let's call this curve C2. Then the partial derivative fy(a,b)
then is the derivative of the function while keeping x values fixed at a while varying y, which is the slope
of the tangent line to this curve at point P (a, b, f(a,b)).

2
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

So far, we have looked at partial derivatives of two variables. The extension of the partial differentiation
to functions of three or more variables is completely similar.

For example, if f is a function of three variables, x, y, and z then its partial derivative with respect to z is
defined as

and, it is found by regarding y and z as constants and differentiating f(x,y,z) with respect to x. This can be
interpreted as the rate of change of z with respect to x when y and z are fixed. However, we cannot
interpret it geometrically because it is not possible to draw the graph of f (it lies in four-dimensional
space).

Higher Derivatives:

If, for example f is a function of two variables (such as f(x,y)) then its partial derivatives fx and fy are also
functions of two variables. Thus, it is also possible to consider the partial derivative of the partial
derivatives. So, for functions of two variables, we can have the following second-order partial derivates:

3
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Thus, the notation fxy (or ) means that we first differentiate with respect to x and then with
respect to y, whereas in computing fyx, the order is reversed.

Does the order matter though?

Clairaut's Theorem: This theorem asserts that

- The proof of this theorem is given in the Appendix F of Stewart's book.

-We can, of course, compute third-, fourth- and even higher-order partial derivatives and the Claiaut's
Theorem can be extended to them as well. For instance,

-Also, for functions of more than 2 variables, for example for functions of 3 variables, similarly, the
Clairaut's theorem is valid, and we have:

4
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

The Gradient of a Function:

If f is a function of two variables x and y, then the gradient of f is the vector function defined by:

Note that points in the domain of a function of several variables can be written in vector notation:

Likewise, if f is a function of three variables x, y, z then the gradient of this function is defined as:

5
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Before we get to the geometric interpretation of the gradient vector, we need to look at the chain rule.

The Chain Rule:

- Suppose you know how the temperature varies in a plane. Let's say this variation is given by:

- and you are moving along a path in the plane. This path can be described with the position vector
function:

So, each of the variables x and y is a function of a variable t. That means T is indirectly a function of t:

Let's say you want to know how the temperature you experience vary with time as you move along :

The chain rule gives the formula for differentiating this function as a function of t.

Theorem: Chain Rule

6
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Side note: This part can be found in Stewart's book so I will quickly go over it in the lecture.

z = f(x, y) shows a surface S and let's first start with expressing the equation of the tangent plane to this
surface at a point P(a, b, f(a, b)). C1 shows the curve obtained by intersecting the vertical plane y = b with
the surface S, and C2 is the curved obtained by intersecting the vertical plane x = a with the surface. The
point P lies on both C1 and C2. We can draw tangent lines to these points at this point (they are shown as
T1 and T2). Then the tangent plane to the surface S at the point P is the plane that contains both tangent
lines T1 and T2.

We know that any plane passing through point (a, b, f(a, b)) has an equation of the form:

Dividing this equation by C

If eqn. 1 represents the tangent plane to the surface at point P, then its intersection with y = b plane must
be the tangent line T1.

7
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Similarly, putting x = a in Eqution1, we get

Tangent plane that contains the tangent line T1 and T2.

The following figure shows the surface z = f(x,y) along with the tangent plane to this surface at point (a,
b, f(a,b)). Point (a,b) is in the domain of this function f. If the (x,y) point of interest changes by an amount
x and y from (a,b) to (a+ x, b+ y), the value of the function changes by an amount z, as shown. In
the same plot, the differential dz represents the change in the height of the tangent plane when point
changes from (a, b) to (a+ x, b+ y).

That is as we zoom into a point on a surface given by a function of two variables, the surface looks more
and more like its tangent plane and we can approximate the function by this tangent plane. This is called
linearization. (In fact the definition of differentiability is as follows: a function is differentiable at (a,b) if

8
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

the linear approximation is a good approximation when (x,y) is near (a,b).

Note that similar expressions for linear approximation can be defined for functions of three or more
variables, although not show-able graphically...

9
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

10
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Direction of the Gradient Vector of a Function of Two Variables:

A level curve of z = f(x,y) is the curve of points (x, y) (i.., points in the xy-plane or in other words points
in the domain of f(x,y)) along which f(x,y) has a constant value:

11
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Let's consider the level curve of f(x,y) = c (which is the curve formed by the set of all points in the
domain of f at which f takes on a constant given value (say c)). This curve can be parameterized and
given as a continuously differentiable function:

Since f is constant c on the curve, we have

If we take any time, say to, the corresponding position vector would be

We see here that, at each point on the level curve of f, the gradient vector, if not , is perpendicular to the
tangent vector (thereby perpendicular to the level curve). This is shown in the figure below:

12
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

13
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

14
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

15
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Generalization:

Now, we can generalize this finding to any curve in the xy-plane. Let's consider a planar curve given by:

The equation of this planar curve can also be written in closed form as:

We can view this curve as the c-level curve of f. Let's say, the vector equation of this curve will be .
Following the same procedure as before, we can conclude that at any given point on this curve, the
gradient vector is perpendicular to the tangent vector passing from that point.

16
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

17
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Gradient of a Function of Three Variables:

For functions of 3 variables, we have level surfaces instead of level curves, but the results are similar.

By extension, if f = f(x, y, z) is a non-constant and continuously differentiable function, then at each point
of the domain of this function, the gradient vector, if not , is perpendicular to the level surface that
passes through that point.

(Remember that level surfaces are surfaces with equations f(x, y, z) = c, where c is a constant. Figure
below shows an example level surface plot (a family of concentric spheres). If the point (x, y, z) moves
along the level surface, the value of f(x, y, z) remains fixed. )

To show this, let's consider any point (xo, yo, zo) in the domain of f (x, y, z) (which is a non-constant
function of three variables). The level surface passing through this point has equation:

18
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

This same argument applies to every differentiable curve that lies on this level surface and passes through
the point (xo, yo, zo) with non-zero tangent vector.

19
Last updated Sept 10, 2019, Prepared by: Alis Ekmekci

Example:

20