MATH298
Set 12
2008/02/13
Total derivative
Let f be a function of two variables, f (x, y), where x and
y are not independent, but depend on a third variable,
p. Then f becomes a function of p,
f (p) = f (x(p), y(p))
How to nd the derivative of thus dened function f (p)?
Suppose that p increases from p to p + p. The change
of x and y will be x dx
p and y dx
p. The indp
dp
crease of f will be, as we alredy know, f
Summarising,
f
f
f
x
+
y.
x
y
f dx
f dy
p +
p.
x dp
y dp
Dividing by p and letting p 0 (then becomes =),
we get the formula for the the total derivative of f :
f dx
f dy
df
=
+
dp
x dp
y dp
(as opposed to the partial derivatives from which it is
composed).
1
Example 1
Elevation of land above sea level, H, depends on two
map coordinates x, y in the following way:
H(x, y) = e0.01(x +y ) .
2
A car travels thorough this terrain, so its coordinates
depend on time in the following way:
x(t) = 7 + 10 cos(10t),
y(t) = 4 + 10 sin(10t).
Find the speed with which the altitude of the car increases or decreases at t = 0.
Solution
dH
H dx
H dy
=
+
dt
x dt
y dt
= 0.02xe0.01(x +y ) (100 sin(10t))
2
0.02ye0.01(x +y ) (100 cos(10t))
2
At t = 0, x = 3, y = 4, this becomes
dH
= (230241)e0.0125
dt
= 8e0.25 6.230
Directional derivative
Let us calculate the slope of z = f (x, y) at
some point P = (a, b) in direction , measured
anticlockwise from the positive x axis. This
can be done if we consider x and y as functions
of a parameter s,
x = a + s cos ,
y = b + s sin ,
and point P corresponds to s = 0.
Then the full derivative of z by s is its directional derivative in the direction :
f dx
f dy
dz
=
+
ds
x ds
x dt
= cos
f
f
+ sin
x P
y P
3
Example 2
Find the slope of the surface z = xy + x2 at
P : (2, 3) in the direction = 1200.
Solution
f
= (y + 2x)(2,3) = 7,
x (2,3)
f
= (x)(2,3) = 2,
y (2,3)
and
cos(1200) = sin(300) = 1/2,
0
0
sin(120 ) = cos(30 ) = 3/2,
so
dz
= 7(1/2) + 2( 3/2)
ds
= (7 + 2 3)/2 5.232
4
Dierentiating implicit functions
Sometimes function y(x) can be dened implicitly,
through an equation like
u(x, y) = 0
with some known (explicit) function of two variables,
u(x, y).
How to dierentiate function y(x) dened in this way?
Function u can be considered as a function of a single
parameter, coinciding with x:
u = u(x, y),
x = p,
y = y(p)
Since u should be identically zero for all x, its total
derivative by p should also be zero. Thus,
du
u dx
u dy
=
+
= 0,
dp
x dp
y dp
and therefore
dy
=
dx
dy
dp
p=x
u
=
x
u
.
y
Example 3
Dierentiate function y(x) dened via equation
x ln(x) + y ln(y) = 2e
at the point x = e, y = e.
Solution First of all,
e ln(e) + e ln(e) = e + e = 2e
so y(e) = e indeed satises the equation.
Function u can be chosen as
u = x ln(x) + y ln(y) 2e,
its partial derivatives are
ux = ln(x) + 1,
uy = ln(y) + 1,
and
ux
ln(x) + 1
dy
=
=
dx
uy
ln(y) + 1
which at x = e, y = e gives
2
dy
= =
dx
2
.
6
Total derivative for a function of
several variables
You know already the incremental approximation for a
function of two variables. Similarly, the incremental approximation for a smooth function of n independent variables x1 , x2 , x3 , . . . , xn can be written as
f
f
f
f
x1 +
x2 + . . .
xn .
x1
x2
xn
The k-th term of this formula represents the change in
f (x1 , . . . , xn ) due to change xk in xk , keeping other n1
of these variables constant.
The total dierential of a function of n independent
variables x1 , x2 , x3 , . . . , xn is
f
f
f
dx1 +
dx2 + . . .
dxn .
x1
x2
xn
This is a symbolic formula, meaning that the above approximate expression becomes more and more accurate
as all xn 0. If all xk depend on a parameter p, then
the formula for the total derivative df /dp can be obtained by symbolically dividing this by dp etc.
df =
Example 4
Equation 2x3 3x 45 = 0 has a solution
x = 3. Find an approximate solution to the equation
2.1x3 2.9x 47 = 0.
Solution. Let ax3 bx c = f (a, b, c, x). We know
that f (a, b, c, x) = 0 for (a, b, c, x) = (2, 3, 45, 3). We put
a = 0.1, b = 0.1, and c = 2 and seek x such that
f (a + a, b + b, c + c, x + x) = 0, i.e., f = 0. Thus
f
f
f
f
a +
b +
c +
x 0,
a
b
c
x
where, at a = 2, b = 3, c = 45 and x = 3,
f
f
= x3 = 27,
a
f
= x = 3,
b
f
f
= 1,
= 3ax2 b = 3 2 32 3 = 51.
c
x
Therefore, approximately,
f (27) 0.1 + (3) (0.1) + (1) 2 + 51 x = 0
Hence, we obtain
x = (2.7 0.3 + 2)/51 = 1/51 0.02,
x 2.98
or
.
8
