Partial Derivatives and Differentiability: Remark
Partial Derivatives and Differentiability: Remark
Partial Derivatives and Differentiability: Remark
Definition 6.1.
Let f : U
real-valued function
f
( x)
xi
Rn R be a real-valued function.
partial derivative
is the
f ( x + he i ) f ( x )
= hlim
0
h
f (x1 , . . . , xi1 , xi + h, xi+1 , . . . , xn ) (x1 , . . . , xi1 , xi , xi+1 , . . . , xn )
= lim
h
6.2. Note that limit used in the above definition is just the limit of a function of a single variable.
Indeed, fix x = (x1 , . . . , xi1 , xi, xi+1 , . . . , xn ) and define
F (h) = (x1 , . . . , xi1 , h, xi+1 , . . . , xn )
Then
dF
f
F ( x i + h) F ( x i )
(
x) = lim
h 0
x
h
dh
Remark
= xi
This observation in fact tells us precisely how to compute the i partial derivative of f . We regard all
the coordinates except xi to be fixed (i.e., constants) and then use ordinary calculus to differentiate the
resulting function of xi (regarded as a function of a single variable).
Example 6.3. Let f (x, y) = cos(x2 y2 ). Then
i
th
f
x
f
y
sin(x2 y2 )(2xy2 )
= sin(x2 y2 )(2x2 y)
The ordinary derivative of a function f of a single variable x tells us the rate of change of a function as
x increases. Similarly, the partial derivative is interpretable as a rate of change: the partial derivative
df
dx
f
( x)
xi
is the rate at which the function f changes as one moves away from the point x in the direction e .
i
Recall that in the case of a function of a single variable, a function f (x) is differentiable only if it is
continuous; but that continuity does not guarantee differentiability. Intuitively, continuity of f (x) requires
that its graph be a continuous curve; and differentiability requires also that there is always a unique tangent
vector to the graph of f (x). In other words, a function f (x) is differentiable if and only if its graph is a
smooth continuous curve with no sharp corners (a sharp corner would be a place where there would be two
possible tangent vectors).
1
If we try to extend this graphical picture of differentiability to functions of two or more variables, it would be
natural to think of a differentiable function of several variables as one whose graph is a smooth continuous
surface, with no sharp peaks or folds. Because for such a surface it would always be possible to associate a
unique tangent plane at a given point.
However, differentiability in this sense turns out to be a much stronger condition than the mere existence
of partial derivatives. For the existence of a partial derivatives at a point x0 requires only a smooth approach
to the point f (x0 ) along the direction of the coordinate axes. We have seen examples of functions that
are discontinuous even though
f ,y
lim
0 f x,
y 0
(
0) = lim
(0
f (x, y) =
has this property, and in fact, both
f
x
and
f
y
x y )2
x2 + y 2
With this sort of phenomenon in mind we give the following definition of differentiability.
Definition .
6.4
if
1.
R2 R
differentiable
at (x0 , y0 )
f
f
Both x and y exist at the point (x0 , y0 ).
2.
(x,y )
Remark
(x0,y0)
lim
f (x , y ) (y y )
f (x, y) f (x0 , y0 ) f
(
x
,
y
)
(
x
x
)
0
0
0
0
0
0
x
y
|(x x0 )2 + (y y0 )2 | 12
=0
f
f
(
x x0 ) +
(y y0 )
x (x0 ,y0 )
y (x0 ,y0 )
is a good approximation to f (x, y) near the point (x0 , y0 ). To make contact with our graphical interpretation
of differentiability, we simply note that the graph of F (x, y) is a plane (it is linear in the variables x and y).
To make this completely obvious, recall that the solution set of any equation of form
Ax + By + Cz = D
is a plane in R3
; so taking
f
x
(x0 ,y0 )
f
y (x0 ,y0 )
f (x0 , y0 )
f
x0
x (x0 ,y0 )
f
y0
y (x0 ,y0 )
Definition 6.6.
plane in
Let f : R2
f (x0 , y0 ) +
f
f
( x x0 ) +
(y y0 )
x (x0 ,y0 )
y (x0 ,y0 )
is called the
at
x0 .
Rn
to
Rm
our notation.
Definition .
from Rn to
Rn
f ( x) =
to
R we define the
gradient
f
f
f
( x) ,
( x) , . . . ,
( x)
x1
x2
xn
of f to be the function
f (x), , . . . , f (x)
Of course, for this definition to make sense all the partial derivatives x
xn
1
Definition 6.8.
a function from U to
R.
must exist.
We say that f is
f
f (x0 + hei ) f (x0 )
(x0 ) = lim
h
0
xi
h
all exist and if
0 =
lim
x x0
f ( x ) f ( x 0 ) f ( x 0 ) ( x x0 )
x x0
Rn to Rm .
Definition 6.9.
0 =
lim
n
f (x) f (x0 ) j=0 fj (x0 ) (x x0 )ej
x x0
x x0
T (x, y) =
and so for
f1
x
f2
x
f1
y
f2
y
=
to be differentiable at (1 0) we require
f
j=0 fj x0 x x0 ej
x x0
x x0
f (x) f (x0 ) (f1 (x0 ) (x x0 ),0) (0, f2 (x0 ) (x x0 ))
x x0
x x0
lim
lim
x ) f (x 0 )
Now
f ( x)
f (x 0 )
= (xy, x + y)
= f(1, 0) = (0, 1)
xy ,
xy
= x
y
f1
f1 x0
f2 x0
(
f2
x x0
x x0
x + y ),
= (
x + y)
y, x)
x
y
(0, 1) (x 1, y 0) = y
,
(1 1)
x
(
f1 x0
(
= (1 1)
) =
f1
f2 x0
(
(1 0) = (0 1)
) =
f2
(1 0) = (1 1)
, y 0) = x 1 + y
=
=
=
Unfortunately as we have seen in the examples we discussed in the lecture on limits and continuity, it is not
so easy to see whether or not this limit exists. Luckily we have a theorem at our disposal that makes it a
lot easier to decide questions of differentiability.
Theorem 6.11. Let f : U
neighborhood of a point
in U . Then f is differentiable at
x.
fj