The Chain Rule
In an introductory calculus course, it is shown that the chain rule is
instrumental in both the theory and the applications of derivatives. In this
section, we generalize the all-important chain rule to functions of
two or more variables.
Let's suppose that w = f( x,y) is differentiable at (p,q) , and then let's suppose that x = h( t) , y = k(t) where h and k are differentiable at t0 and (h( t0) ,k( t0) ) = ( p,q) .
Then definition 3.2 in section 3 says that
|
Dw = fx( p,q) Dx+fy( p,q) Dy + o1(Dx,Dy)Dx + o2(Dx,Dy)Dx |
|
so that if we divide by Dt and let Dt approach 0, then we
obtain
|
|
lim
Dt® 0
|
|
Dw
Dt
|
= fx( p,q) |
lim
Dt® 0
|
|
Dx
Dt
|
+fy( p,q) |
lim
Dt® 0
|
|
Dy
Dt
|
+ |
lim
Dt® 0
|
|
o1(Dx,Dy)
|
Dx
Dt
|
+o2(Dx,Dy)
|
Dy
Dt
|
|
| (1) |
In order for dx/dt to exist, it is necessary that Dx approaches 0 as Dt approaches 0. Likewise, Dy approaches 0 as Dt approaches 0. Since o1(Dx,Dy) and o2(Dx,Dy) approach 0 as Dx and Dy approach 0, the last limit in (1) becomes
lim
Dt® 0
|
|
o1(Dx,Dy)
|
Dx
Dt
|
+o2(Dx,Dy)
|
Dy
Dt
|
|
=
|
 |
lim
Dt® 0
|
o1(Dx,Dy)
|
|
dx
dt
|
+
|
|
lim
Dt® 0
|
o2(Dx,Dy)
|
|
dy
dt
|
= 0
|
|
Consequently, (1) reduces to
|
|
dw
dt
|
= fx( p,q) |
dx
dt
|
+fy( p,q) |
dy
dt
|
|
| (2) |
which is known as the chain rule for functions of 2 variables.
Moreover, the chain rule can also be written in the form
|
|
dw
dt
|
= |
¶f
¶x
|
|
dx
dt
|
+ |
¶f
¶y
|
|
dy
dt
|
|
|
EXAMPLE 1 Find dw/dt given that w = x2+y3 and that x = cot( t) , y = sin( t) .
Solution: The first partial derivatives of w = x2+y3 are
As a result, the chain rule says that
|
|
dw
dt
|
= 2x |
dx
dt
|
+3y2 |
dy
dt
|
|
|
and since dx/dt = -csc2( t) and dy/dt = cos( t), we have
|
|
|
|
2cot( t) · [ -csc2(t) ] +3[ sin( t) ]2 cos(t) |
| |
|
| -2cot( t) csc2( t) +3sin2(t) cos( t) |
|
|
EXAMPLE 2 Find dw/dt given that w = cos( xy)
and that x = pet and y = e-t.
Solution: The first partial derivatives are
|
|
¶w
¶x
|
= -ysin( xy) , |
¶w
¶y
|
= -xsin( xy) |
|
As a result, the chain rule says that
|
|
dw
dt
|
= -ysin( xy) |
dx
dt
|
-xsin( xy) |
dy
dt
|
|
|
and since dx/dt = pet and dy/dt = -e-t, we have
|
|
|
|
-e-tsin( pete-t) ( pet) -petsin( pete-t) (-e-t) |
| |
|
|
-pete-tsin( pete-t) +pete-tsin( pete-t) |
| |
|
| |
|
|
|
Check your Reading: Substitute x = pet and y = e-t into w = cos( xy). Why would this also
imply that dw/dt = 0?