3DUWLDO'HULYDWLYHV

3DUWLDO'HULYDWLYHVRI+LJKHU2UGHU
&KDQJHLQWKH2UGHURI3DUWLDO'HULYDWLRQ
6XIILFLHQW&RQGLWLRQVIRUWKH(TXDOLW\RII[\DQGI\[
<RXQJ
V7KHRUHP
6FKZDU]
V7KHRUHP so that the limit depends on the value of m, i.e. on the path of approach and is
different for the different paths followed and therefore does not exist. Hence the
function f(x, y) is not continuous at (0, 0). Again
f (0  h ,0)  f (0,0) 0
fx(0, 0) = lim lim 0
h o0 h h o0 h
PARTIAL DERIVATIVES
f (0.0  k )  f (0,0)
Let f be a function of several variables ,then the ordinary derivative of f with fy(0, 0) = lim
k o0 k
respect to one of the independent variables, keeping all other independent variables
0
= lim 0
constant is called the partial derivative . Partial derivative of f(x, y) with respect to k o0 k
x is generally denoted by wf/wx or fx or fx(x, y).Similarly those with respect to y are Definition.Let (x, y), (x + Gx, y + Gy) be two neighbouring points and let
denoted by wf/wy or fy or fy(x, y). Gf = f(x + Gx, y + Gy)  f(x, y)
wf f ( x  įx , y)  f ( x, y) The function f is said to be differentiable at (x, y) if
? lim
wx įx o0 įx
Gf = A Gx + B Gy + Gx I(Gx, Gy) + Gy \(Gx, Gy) …(1)
when these limits exist.
where A and B are constants independent of Gx, Gy and I, \ are functions of Gx, Gy
The partial derivatives at a particular point (a, b) are definined as (in case the limits
tending to zero as Gx, Gy tend to zero simultaneously.
exist)
Also, A Gx + B Gy is then called the differential of f at (x, y) and is denoted
f (a  h , b)  (a , b )
fx(a, b) = lim by df. Thus
h o0 h
f (a , b  k )  f (a , b ) df = A Gx + B Gy
fy(a, b) = lim
k o0 k From (1) when (Gx, Gy) o (0, 0), we get
f(x + Gx, y + Gy)  f(x, y) o 0
Example . For the function f(x,y),where or
­ xy f(x + Gx, y + Gy) o f(x, y)
° , ( x , y) z (0,0)
f(x, y) = ® x 2  y 2
Ÿ The function f is continuous at (x, y)
° 0
¯ , ( x , y) (0,0)
Again, from (1), when Gy = 0 (i.e., y remains constant)
both the partial derivatives exist at (0, 0) but the function is not continuous at(0, 0)
Gf = A Gx + Gx I(Gx, 0)
.
Dividing by Gx and proceeding to limits as Gx o 0, we get
Solution. Putting y = mx, we see that
wf
m A
lim f ( x , y) wx
x o0 1 m2
Similarly
 wf İ2 2 İ2
B x2 < ,y 
wy 8 8
Thus the constants A and B are respectively the partial derivatives of f with or, if
respect to x and y. İ İ
|x| < | y |
Hence a function which is differentiable at a point possesses the first order 2 2 2 2
partial derivatives thereat. Again the differential of f is given by x 3  y3 İ İ
?  0  İ, when | x | , | y |
wf wf x 2  y2 2 2 2 2
df = AGx + B Gy = wx  Gy
wx wy
x 3  y3
Ÿ lim =0
Taking f = x, we get dx = Gx. ( x , y ) o( 0 , 0 ) x 2  y2
Similarly taking f = y, we obtain dy = Gy. Ÿ lim f(x, y) = f(0, 0)
( x , y ) o( 0 , 0 )
Thus the differentials dx, dy of x, y are respectively Gx and Gy, and
Hence the function is continuous at (0, 0).
wf wf
df = dx  dy f x dx + fy dy …(2) Again,
wx wy
f (h,0)  f (0,0) h 0
is the differential of f at (x, y). fx(0, 0) = lim lim 1
h o0 h h o0 h
Note 1. If we replace Gx, Gy, by h, k in equation (1) we say that the function is
f (0, k )  f (0,0) k
differentiable at a point (a, b) of the domain of definition if df can be expressed as fy(0, 0) = lim lim 1
k o0 k k o0 k
df = f(a + h, b + k)  f(a, b) Thus the function possesses partial derivatives at (0, 0).
= Ah + Bk + hI(h, k) + k\(h, k) …(3) If the function is differentiable at (0, 0), then by definition
where A = fx, B = fy and I, \ are functions of h, k tending to zero as h, k tend to df = f(h, k)  f(0, 0) = Ah + Bk + hI + k\ …(1)
zero simultaneously. when A and B are constants (A = fx(0, 0) = 1, B = fy(0, 0) = 1) and I, \ tend to
Example. Let zero as (h, k) o (0, 0).
­ x 3  y3 Putting h = U cos T, k = U sin T, and dividing by U, we get
° , ( x, y) z (0,0)
f(x, y) = ® x 2  y 2
° 0, cos3T  sin3T = cosT + I cos T + \ sin T …(2)
¯ ( x, y) (0,0)
For arbitrary T = tan1 (h/k), Uo0 implies that (h, k) o (0, 0). Thus we get
Put x = r cos T, y = r sin T,
the limit,
x 3  y3 cos3T  sin3T = cos T  sin T
? = |r(cos3T  sin3T)| d 2|r| = 2 x 2  y 2 < H,
x 2  y2
or
if cos T sin T(cos T  sin T) = 0
which is plainly impossible for arbitrary T.
 Thus the function is not differentiable at the origin. |cos T sin T|1/2 = I T + \ sin T
Example. Show that the function f, where Now for arbitrary T, Uo0 implies that (h, k) o (0, 0).
­ xy Taking the limit as Uo0, we get
° , if x 2  y 2 z 0
f(x, y) = ® x 2  y 2 |cosT sin T|1/2 = 0.
°
¯ 0, if x y 0
which is impossible for all arbitrary T.
is continuous, possesses partial derivatives but is not differentiable at the origin. Example. The function f, where
Solution. It may be easily shown that f is continuous at the origin and
­ x 2  y2
fx(0, 0) = 0 = fy(0, 0) °xy , if x 2  y 2 z 0
f(x, y) = ® x 2  y 2
If the function is differentiable at the origin then by definition ° 0,
¯ if x y 0
df = f(h, k)  f(0, 0) = Ah + Bk + hI + k\ …(1) is differentiable at the origin.
where A = fx(0, 0) = 0; B = fy(0, 0) = 0, and I, \ tend to zero as (h, k) o (0, 0). Solution. It is easy to show that
hk fx(0, 0) = 0 = fy(0, 0)
? = hI + k\ …(2)
h2  k2 Also when x + y2 z 0,
2

Putting k = mh and letting ho0, we get | x 4 y  4x 2 y 3  y 5 | 6( x 2  y 2 ) 5 / 2

|fx| = d = 6(x2 + y2)1/2
m (x 2  y 2 ) 2 (x 2  y 2 ) 2
lim (I + m\) = 0
h o0
1 m2 Evidently
which is impossible for arbitrary m. lim fx(x, y) = 0 = fx(0, 0)
( x , y ) o( 0 , 0 )
Hence the function is not differentiable at (0, 0).
Thus fx is continuous at (0, 0) and fy(0, 0) exists,
Example. Prove that the function
Ÿ f is differentiable at (0, 0)
f(x, y) = | xy |
PARTIAL DERIVATIVES OF HIGHER ORDER
is not differentiable at the point (0, 0), but fx and fy both exist at the origin .
If a function f has partial derivatives of the first order at each point (x, y) of
f (h,0)  f (0,0) 0
Solution. fx(0, 0) = lim lim 0 a certain region, then fx, fy are themselves functions of x, y and may also possess
h o0 h h o0 h
partial derivatives. These are called second order partial derivatives of f and are
f (0, k )  f (0,0) 0
fy(0, 0) = lim lim 0 denoted by
k o0 k k o0 k

If the function is differentiable at (0, 0), then by definition w § wf · w 2f

¨ ¸ = fxx = f 2
wx © wx ¹ wx 2 x
f(h, k)  f(0, 0) = 0h + 0k + hI + k\
w § wf · w 2f
where I and \ are functions of h, k and tend to zero as (h, k) o (0, 0). ¨ ¸ = fyy = f 2
wy ¨© wy ¸¹ wy 2 y
Putting h = U cos T, k = U sin T and dividing by U, we get
 w § wf · w 2f f x (0, k )  f x (0,0)
¨ ¸ = fxy fyx(0, 0) = lim
wx ¨© wy ¸¹ wxwy k o0 k
But
w § wf · w 2f
¨ ¸ = fyx
f (h,0)  f (0,0)
wy © wx ¹ wywx fx(0, 0) = lim 0
h o0 h
Thus(in case the limits exist)
f (h, k )  f (0, k ) hk (h 2  k 2 )
f (a  h , b )  f x (a , b) fx(0, k) = lim lim k
fxx(a, b) = lim x h o0 h h o0 h ( h 2  k 2 )
h o0 h
f y (a  h , b )  f y (a , b) k 0
? fyx(0, 0) = lim = 1
fxy(a, b) = lim k o0 k
h o0 h
f x (a , b  k )  f x (a , b) ? fxy(0, 0) z fyx(0, 0)
fyx(a, b) = lim
k o0 k Sufficient Conditions for the Equality of fxy and fyx
f y (a , b  k )  f y (a , b) We have two theorems to show that fxy = fyx at a point.
fyy(a, b) = lim
k o0 k Theorem ( Young’s theorem). If fx and fy are both differentiable at a point (a, b)
of the domain of definition of a function f, then
Change in the Order of Partial Derivation fxy(a, b) = fyx(a, b)
Consider an example to show that fxy may be different from fyx. Proof. We prove the theorem by taking equal increment h both for x and y and
Example . Let calculating I(h, h) in two different ways.
xy( x 2  y 2 ) Let (a + h, b + h) be a point of this neighbourhood. Consider
f(x, y) = , (x, y) z (0, 0),
x 2  y2 I(h, h) = f(a + h, b + h) f(a + h, b)  f(a, b + h) + f(a, b)
f(0, 0) = 0, then at the origin fxy z fyx. G(x) = f(x, b + h)  f(x, b)
Solution. Now so that
f y (h,0)  f y (0,0) I(h, h) = G(a + h)  G(a) …(1)
fxy(0, 0) = lim
h o0 h Since fx exists in a neighbourhood of (a, b), the function G(x) is derivable in
f (0, k )  f (),0) 0 (a, a + h) and therefore by Lagrange’s mean value theorem, we get from (1),
fy(0, 0) = lim lim =0
k o0 k k o0 k
I(h, h) = hGc(a + Th), 0 < T < 1
f (h, k )  f (h,0) hk (h 2  k 2 )
fy(h, 0) = lim lim h = h{fx(a + Th, b + h)  fx(a + Th, b)} …(2)
k o0 k k o0 k ˜ ( h 2  k 2 )
Again, since fx is differentiable at (a, b), we have
h 0
? fxy = lim =1 fx(a + Th, b + h) fx(a, b) = ThIxx(a, b) + hfyx(a, b)
h o0 h
+ ThI1(h, h) + h\1(h, h) …(3)
Again
and
 fx(a + Th, b)  fx(a, b) = Thfxx(a, b) + ThI2(h, h) …(4) or
where I1, \1, I2 all tend to zero as ho0. 1 ­ f (a  h , b  k )  f (a  h , b ) f (a , b  k )  f (a , b) ½
®  ¾
From (2), (3), (4), we get h¯ k k ¿

I(h, h)/h2 = fyx(a, b) + TI1(h, h) + \1(h, h)  TI2(h, h) …(5) = fyx(a + Th, b + Tck)

Similarly, taking Taking limits when ko0, since fy and fyx exist in a neighbourhood of (a, b),
H(y) = f(a + h, y)  f(a, y) we get
we can show that f y (a  h , b)  f y (a , b)
lim f yx (a + Th, b + Tck)
h k o0
I(h, h)/h2 = fxy(a, b) + I3(h, h) + Tc\2(h, h)  Tc\2(h, h) …(6)
where I3, \2, \3 all tend to zero as ho0. Again, taking limits as ho0, since fyx is continuous at (a, b), we get

On taking the limit as ho0, we obtain from (5) and (6) fxy(a, b) = lim lim fyx(a + Th, b + Tck) = fyx(a, b)
h o0 k o0

ij( h , h )  If the conditions of Young’s or Schwarz’s theorem are satisfied then fxy = fyx at a
lim = fxy(a, b) = fyx(a, b)
h o0 h2 point (a, b). But if the conditions are not satisfied, we cannot draw any conclusion
Theorem. (Schwarz’s theorem). If fy exists in a certain neighbourhood of a point regarding the equality of fxy and fyx. Thus the conditions are sufficient but not
(a, b) of the domain of definition of a function f, and fyx is continuous at (a, b), then necessary.
fxy(a, b) exists, and is equal to fyx(a, b). Example . Show that for the function
Proof. Let (a + h, b + k) be a point of this neighbourhood of (a, b).
­ x 2 y2
Take ° , ( x , y) z (0,0)
f(x, y) = ® x 2  y 2
I(h, k) = f(a + h, b + k) f(a + h, b) f(a, b + k) + f(a, b) ° 0,
¯ ( x , y) (0,0)
G(x) = f(x, b + k)  f(x, b) Solution. Here fxy(0, 0) = fyx(0, 0) since
so that f ( x ,0)  f (0,0)
fx(0, 0) = lim 0
I(h, k) = G(a + h)  g(a) …(1) x o0 x
Since fx exists in a neighbourhood of (a, b), the function g(x) is derivable in Similarly, fy(0, 0) = 0.
(a, a + h), and therefore by Lagrange’s mean value theorem, we get from (1) Also, for (x, y) z (0, 0).
I(h, k) = hGc(a + Th), 0<T<1 ( x 2  y 2 ).2 xy 2  x 2 y 2 .2 x 2 xy 4
fx(x, y) =
= h{fx(a + Th, b + k) fx(a + Th, b)} …(2) (x 2  y 2 ) 2 (x  y 2 ) 2
2

Again, since fyx exists in a neighbourhood of (a, b), the function fx is 2x 4 y

fy(x, y) =
derivable with respect to y in (b, b + k), and therefore by Lagrange’s mean value (x  y 2 ) 2
2

theorem, we get from (2) Again

I(h, k) = hkfyx(a + Th, b + Tck), 0 < Tc < 1
 f x (0, y)  f x (0,0)
fyx(0, 0) = lim =0
y o0 y
and
fxy(0, 0) = 0, so that fxy(0, 0) = fyx(0, 0)
For (x, y) z (0, 0), we have
8xy 3 ( x 2  y 2 ) 2  2 xy 4 .4 y( x 2  y 2 )
fyx(x, y) =
(x 2  y 2 ) 4

8x 3 y 3
=
(x 2  y 2 )3
and it may be easily shown (by putting y = mx) that
lim fyx(x, y) z 0 = fyx(0, 0)
( x , y ) o( 0 , 0 )

so that fyx is not continuous at (0, 0), i.e., the conditions of Schwarz’s theorem are
not satisfied.
We now show that the conditions of Young’s theorem are also not satisfied.
f x ( x ,0)  f x (0,0)
fxx(0, 0) = lim =0
x o0 x
Also fx is differentiable at (0, 0) if
fx(h, k)  fx(0, 0) = fxx(0, 0). H + fyx(0, 0). K + hI + k\
or
2hk 4
= hI + k \
(h 2  k 2 ) 2

where I, \ tend to zero as (h, k) o (0, 0).

Putting h = U cos T and k = U sin T, and dividing by U, we get
2 cos T sin4 T = cos T.I + sin T\
and (h, k)o(0, 0) is same thing as Uo0 and T is arbitrary. Thus proceeding to
limits, we get
2 cos T sin4T = 0
which is impossible for arbitrary T,
Ÿ fx is not differentiable at (0, 0)
Similarly, it may be shown that fy is not differentiable at (0, 0).
Thus the conditions of Young’s theorem are also not satisfied but, as shown above,
fxy(0, 0) = fyx(0, 0)