Module 3

Application of Multivariable Calculus

• Taylor’s expansion for two variables–maxima and minima–constrained maxima and minima-
Lagrange’s multiplier method.

3.1 The Taylor’s Expansion

A useful technique in the analysis of real-valued functions is the approximation of continuous func-
tions by polynomials. Taylor’s and Maclaurin’s theorems are important tools that provide such
an approximation of the real-valued functions. These theorems are regarded as ’generalized mean
value theorems’ in the sense that mean value theorems relate the value of the function and its first
order derivative, whereas, Taylor’s and Maclaurin’s theorems generalize this relation to higher order
derivatives.

3.1.1 Taylor’s series of functions of a single variable

If a function f (x) is such that
1. f (x), f ′ (x), f ′′ (x), . . . , f (n−1) (x) are continuous in the closed interval [a, a + h]
2. f (n) (x) exists in the open interval (a, a + h), then there exists at least one number θ, between
0 and 1 , such that
′ h2 ′′ h3 ′′′ hn−1 n−1 h
n
f (a+h) = f (a)+hf (a)+ f (a)+ f (a)+. . .+ f (a) + f (n) (a+θh), 0 < θ < 1
2! 3! (n − 1)! n!
(3.1.1)

3.1.2 Taylor’s Infinite Series

h2 ′′
′ hn−1 (n−1)
Write (3.1.1) as f (a+h) = Sn +Rn , where Sn = f (a)+hf (a)+ f (a)+. . .+ f (a),
2! (n − 1)!
hn (n)
and, Rn = f (a + θh), 0 < θ < 1.
n!

58
3.1. THE TAYLOR’S EXPANSION 59

Suppose that Rn → 0, as n → ∞, then limn→∞ Sn = f (a + h), so that the series

h2 ′′ h3 hn−1 (n−1)
f (a) + hf ′ (a) + f (a) + f ′′′ (a) + . . . + f (a) + . . .
2! 3! (n − 1)!

converges and its sum is equal to f (a + h). Thus, we obtain that

(i) if f (x) possesses derivatives of every order in the interval [a, a + h], and

hn (n)
(ii) the remainder f (a + θh) tends to zero as n tends to infinity, then
n!
h2 ′′ hn
f (a + h) = f (a) + hf ′ (a) + f (a) + · · · + f (n) (a) + · · ·
2! n!
This series is known as Taylor’s infite series.

3.1.3 Maclaurin’s Theorem with Lagrange’s Form of Remainder

If a function f (x) is such that

1. f (x), f ′ (x), f ′′ (x), . . . , f (n−1) (x) are continuous in the closed interval [0, x].

2. f (n) (x) exists in the open interval (0, x), then there exists at least one number θ between 0 and
1 such

x2 ′′ xn−1 (n−1) xn
f (x) = f (0) + xf ′ (0) + f (0) + . . . + f (0) + f (n) (θx).
2! (n − 1)! n!
This is obtained from Taylor’s theorem by considering the interval [0, x] instead of [a, a + h] and
changing a to 0 and h to x in (3.1.1).

3.1.4 Maclaurin’s Infinite Series

Set 0 for a and x for a + h in Taylor’s infinite series, we obtain that

(i) if f (x) possesses derivative of every order in the interval [0, x], and

xn n
(ii) the remainder f (θx) tends to zero as n tends to infinity, then
n!

x2 ′′ xn
f (x) = f (0) + xf ′ (0) + f (0) + . . . + f (n) (0) + . . .
2! n!

This series is known as Maclaurin’s series for the expansion of function f (x) in powers of x.
60 MODULE 3. APPLICATION OF MULTIVARIABLE CALCULUS

3.1.5 Taylor’s series for functions of two variables

Taylor’s Expansion
If f (x, y) and its partial derivatives up to order (n + 1) are continuous throughout the domain D
centered at a point (x0 , y0 ), then throughout D,
   2
∂ ∂ 1 ∂ ∂
f (x0 + h, y0 + k) = f (x0 , y0 ) + h +k f (x0 , y0 ) + h +k f (x0 , y0 ) + . . .
∂x ∂y 2! ∂x ∂y
 n
1 ∂ ∂
+ h +k f (x0 , y0 ) + Rn (3.1.2)
n! ∂x ∂y
where Rn is the remainder term given by
 n+1
1 ∂ ∂
Rn = h +k f (x0 + θh, y0 + θk) , 0<θ<1 (3.1.3)
(n + 1)! ∂x ∂y

Taylor’s Expansion for f (x, y) at the Origin: Maclaurin’s Expansion

If (x0 , y0 ) = (0, 0), then we treat h and k as independent variables, replacing h by x and k by y
in (3.1.2), we obtain Taylor’s expansion for f (x, y) at (0, 0), also called the Maclaurin’s expansion,
given by
   2
∂ ∂ 1 ∂ ∂
f (x, y) = f (0, 0) + x +y f (0, 0) + x +y f (0, 0) + . . .
∂x ∂y 2! ∂x ∂y
 n  n+1
1 ∂ ∂ 1 ∂ ∂
+ x +y f (0, 0) + x +y f (θx, θy), 0 < θ < 1
(3.1.4)
n! ∂x ∂y (n + 1)! ∂x ∂y
When Rn → 0 as n → ∞, from (3.1.2) and (3.1.4), we get respectively, the Taylor’s series and
Maclaurin’s series, as
   2
∂ ∂ 1 ∂ ∂
f (x0 + h, y0 + k) = f (x0 , y0 ) + h +k f (x0 , y0 ) + h +k f (x0 , y0 )
∂x ∂y 2! ∂x ∂y
 3
1 ∂ ∂
+ h +k f (x0 , y0 ) + . . . (3.1.5)
3! ∂x ∂y
and,
   2  3
∂ ∂ ∂ ∂ 1 ∂ ∂
f (x, y) = f (0, 0)+ x +y f (0, 0)+ x +y f (0, 0)+ x +y f (0, 0)+. . .
∂x ∂y ∂x ∂y 3! ∂x ∂y
(3.1.6)
The results for the Taylor’s expansion can be extended to more than two variables on the similar line.

Approximation and Error Estimation

Taylor’s formula (3.1.2) provides polynomial approximation to two variables functions. The first
n + 1 terms give the polynomial of degree n and the last term gives approximation error.
3.1. THE TAYLOR’S EXPANSION 61

For example, for n = 1, the linear approximation of f (x, y) about the point (x0 , y0 ) is

f (x, y) = f (x0 + x − x0 , y0 + y − y0 ) ≈ f (x0 , y0 ) + (x − x0 ) fx (x0 , y0 ) + (y − y0 ) fy (x0 , y0 )

(3.1.7)
with error term
1 
(x − x0 )2 fxx + 2 (x − x0 ) (y − y0 ) fxy + (y − y0 )2 fyy

R1 = (3.1.8)
2!

Here, fxx , fxy and fyy are to be evaluated at the point [x0 + θ (x − x0 ) , y0 + θ (y − y0 )] , 0 < θ < 1.

Since θ is unknown we cannot evaluate R1 exactly, however, it is possible to find an upper bound
to R1 in a given rectangular region

R = {(x, y) : |x − x0 | ≤ δ1 , |y − y0 | ≤ δ2 } .

From (3.1.8)

1
|x − x0 |2 |fxx | + 2 |x − x0 | |y − y0 | |fxy | + |y − y0 |2 |fyy |

|R1 | ≤
2

If M = max {|fxx | , |fxy | , |fyy |} for all (x, y) ∈ R, then

M  M M
|R1 | ≤ |x − x0 |2 + 2 |x − x0 | |y − y0 | + |y − y0 |2 = [|x − x0 | + |y − y0 |]2 ≤ [δ1 + δ2 ]2 ,
2 2 2

for all (x, y) ∈ R. Thus we have the following result:

If M = max{|fxx |, |fxy |, |fyy }, then the maximum absolute error |R1 | in the linear approxima-
tion of f (x, y) in the rectangular region R = {(x, y) : |x − x0 | ≤ δ1 , |y − y0 | ≤ δ2 } about the point
(x0 , y0 ), is
M
[δ1 + δ2 ]2 (3.1.9)
2
Similarly, it can be shown that the maximum absolute error |R2 | in the quadratic approximation of
f (x, y) in the rectangular region R = {(x, y) : |x − x0 | ≤ δ1 , |y − y0 | ≤ δ2 } about the point (x0 , y0 ),
is
M
(δ1 + δ2 )3 , (3.1.10)
6
where M = max · {|fxxx | , |fxxy | , |fxyy | , |fyyy |}. We note that the quadratic approximation of
f (x, y) about (x0 , y0 ) is given by

1 
(x − x0 )2 fxx + 2 (x − x0 ) (y − y0 ) fxy + (y − y0 )2 fyy

f (x, y) ≈ f (x0 , y0 )+(x − x0 ) fx +(y − y0 ) fy +
2!

Example 3.1.1. Find the quadratic Taylor series polynomial approximation to the function f (x, y)
= sin x sin y about the origin. Obtain the maximum absolute error if |x| ≤ 0.1 and |y| ≤ 0.1.
62 MODULE 3. APPLICATION OF MULTIVARIABLE CALCULUS

Solution: We have
f (x, y) = sin x sin y, f (0, 0) = 0
fx (x, y) = cos x sin y, fx (0, 0) = 0
fy (x, y) = sin x cos y, fy (0, 0) = 0
fxx (x, y) = − sin x sin y, fxx (0, 0) = 0
fxy (x, y) = cos x cos y, fxy (0, 0) = 1
fyy (x, y) = − sin x sin y, fyy (0, 0) = 0

fxxx (x, y) = − cos x sin y

fxxy (x, y) = − sin x cos y
fxyy (x, y) = − cos x sin y
fyyy (x, y) = − sin x cos y

The quadratic approximation about (0, 0) is given by

1  2
x fxx (0, 0) + 2xyfxy (0, 0) + y 2 fyy (0, 0)

f (x, y) ≈ f (0, 0) + {xfx (0, 0) + yfy (0, 0)} +
2!

Substituting for f (0, 0), fx (0, 0), fy (0, 0), etc., we obtain

sin x sin y ≈ xy.

The maximum absolute error in the quadratic approximation, refer (3.1.10), is given by

M M
|R2 | ≤ [|x| + |y|]3 ≤ [0.1 + 0.1]3 = 0.00133M
6 6

where M = max . {|fxxx | , |fxxy | , |fxyy | |fyyy |} for all (x, y) ∈ R : |x| ≤ 0.1 and |y| ≤ 0.1.

Here, the derivatives are to be evaluated at point (θx, θy), 0 < θ < 1.

Since, the third order derivatives in this case, as derived above, are products of sines and cosines
thus they can never exceed unity and hence M can at the most be equal to 1 ; and thus |R2 | ≤
0.00133.

Example 3.1.2. Find the linear and quadratic Taylor series polynomial approximation to the function
f (x, y) = x2 y + 3y − 2 about the point (−1, 2) and obtain the maximum absolute error in the region
|x + 1| ≤ 0.1 and |y − 2| ≤ 0.1.
3.1. THE TAYLOR’S EXPANSION 63

Solution. We have
f (x, y) = x2 y + 3y − 2, f (−1, 2) = 6
fx = 2xy fx (−1, 2) = −4
f y = x2 + 3 fy (−1, 2) = 4
fxx = 2y fxx (−1, 2) = 4
fxy = 2x fxy (−1, 2) = −2
fyy = 0 fyyy (−1, 2) = 0
fxxx = 0 fxxx (−1, 2) = 0
fxxy = 2 fxxy (−1, 2) = 2
fxyy = 0 fxyy (−1, 2) = 0
fyyx = 0 fyyx (−1, 2) = 0
fyyy = 0 fyyy (−1, 2) = 0
The linear approximation of f (x, y) about (−1, 2) is

f (x, y) ≈ f (−1, 2) + [(x + 1)fx (−1, 2) + (y − 2)fy (−1, 2)]

= 6 + (x + 1)(−4) + (y − 2)(4)
= 6 − 4(x + 1) + 4(y − 2).

The maximum absolute error in the linear approximation, refer (3.1.9), is

M M
|R1 | ≤ [|x + 1| + |y − 2|]2 ≤ [0.1 + 0.1]2 = 0.02M
2 2
where M = max {|fxx | , |fxy | , |fyy |} for all (x, y) ∈ R : |x + 1| ≤ 0.1 and |y − 2| ≤ 0.1.

Now,

|fxx | = |2y| = 2|y| = 2|(y − 2) + 2|

≤ 2[|y − 2| + 2] ≤ 2(0.1 + 2) = 4.2
|fxy | = |2x| = 2|x| = 2|(x + 1) − 1|
≤ 2[|x + 1| + 1] ≤ 2[0.1 + 1] = 2.2
|fyy | = 0.
Hence, M = max{4.2, 2.2, 0} = 4.2 , and thus |R1 | ≤ 0.02(4.2) = .084

Next, the quadratic approximation about (−1, 2) is

1 
f (x, y) ≈f (−1, 2) + [(x + 1)fx (−1, 2) + (y − 2)fy (−1, 2)] + (x + 1)2 fxx (−1, 2)
 2!
+2(x + 1)(y − 2)fxy (−1, 2) + (y − 2)2 fyy (−1, 2)
1
=6 + (x + 1)(−4) + (y − 2)4 + 4(x + 1)2 + 2(x + 1)(y − 2)(−2)

2
=6 − 4(x + 1) + 4(y − 2) + 2(x + 1)2 − 2(x + 1)(y − 2).
64 MODULE 3. APPLICATION OF MULTIVARIABLE CALCULUS

The maximum absolute error in the quadratic approximation, refer (3.1.10), is

M M
|R2 | ≤ [|x + 1| + |y − 2|]3 ≤ [0.1 + 0.1]3
6 6
.008M
= ≈ .00133M,
6
where M = max {|fxxx | , |fxxy | , |fxyy | , |fyyy |} for all (x, y) ∈ R : |x + 1| ≤ 0.1 and |y − 2| ≤ 0.1.

Hence, M = max{0, 2, 0, 0} = 2, hence |R2 | ≤ (0.00133)(2) = 0.00266.

Example 3.1.3. Expand f (x, y) = tan−1 (xy) in powers of (x − 1) and (y − 1) up to second degree
terms. Hence compute f (1.1, 0.8).

Solution: We have
f (x, y) = tan−1 xy f (1, 1) = π/4 ≈ 0.7854
y
fx = fx (1, 1) = 12
1 + x2 y 2
x 1
fy = fy (1, 1) = 2
1 + x2 y 2
−2xy 3
fxx = fxx (1, 1) = − 21
(1 + x2 y 2 )2
1 − x2 y 2
fxy = fxy (1, 1) =0
(1 + x2 y 2 )2
−2x3 y
fyy = fyy (1, 1) = − 21 .
(1 + x2 y 2 )2

Taylor’s series expansion of f (x, y) about (1, 1), up to terms of degree 2 , is

f (x, y) ≈ f (1, 1) + [(x − 1)fx (1, 1) +(y − 1)fy (1, 1)]

1
+ (x − 1)2 fxx (1, 1) + 2(x − 1)(y − 1)fxy (1, 1) + (y − 1)2 fyy (1, 1) .

2
Substituting the values for f (1, 1), fx (1, 1), fy (1, 1) etc. we get
    
−1 1 1
tan xy ≈ 0.7854 + (x − 1) + (y − 1)
2 2
    
1 2 1 2 1
+ (x − 1) − + 2(x − 1)(y − 1)(0) + (y − 1) −
2 2 2

or, tan−1 xy ≈ 0.7854 + 21 (x − 1) + 21 (y − 1) − 14 (x − 1)2 − 14 (y − 1)2 .

To compute the value of tan−1 xy at (1.1, 0.8), put x = 1.1 and y = 0.8, we obtain

1 1 1 1
f (1.1, 0.8) = 0.7854 + (0.1) + (−0.2) − (0.1)2 − (−0.2)2
2 2 4 4
= 0.7229.
3.1. THE TAYLOR’S EXPANSION 65
 
Example 3.1.4. Evaluate ln (1.03)1/3 + (0.98)1/4 − 1 approximately using linear Taylor’s series
approximation.
 
Solution: Let f (x, y) = ln x1/3 + y 1/4 − 1 . Take x0 = 1, y0 = 1, h = 0.03 and k = −0.02.
1 −2/3 1 −3/4
3
x 4
y
We have, fx = and fy = .
x1/3 + y 1/4 −1 x1/3 + y 1/4 −1
Thus, f (1, 1) = 0, fx (1, 1) = 1/3 and fy (1, 1) = 1/4.

Linear Taylor’s series approximation is

f (x0 + h, y0 + k) ≈ f (x0 , y0 ) + [hfx (x0 , y0 ) + kfy (x0 , y0 )] .

Substituting the values, we get

ln (1.03)1/3 + (0.98)1/4 − 1 ≈ 0 + (.03)(1/3) − (.02)(1/4)

 

= .01 − .005 = .005 (approx.)

Homework

1. Obtain the quadratic Taylor’s series polynomial approximation to the function f (x, y) = 2x3
+3y 3 − 4x2 y about the point (1, 2) and obtain the maximum absolute error in the region
|x − 1| ≤ 0.01, |y − 2| ≤ 0.1.

2. Using Taylor’s series find a quadratic approximation of cos x cos y at the origin and also esti-
mate the error in the approximation, if |x| ≤ 0.1 and |y| ≤ 0.1.

3. Find the quadratic Taylor’s series approximation of ex cos y about the point (1, π/4).
(x + h)(y + k)
4. Expand in powers of h and k up to the second degree terms.
(x + h) + (y + k)
5. Find the cubic Taylor’s series polynomial approximation of f (x, y) = tan−1 (y/x) and hence
compute f (1.1, 0.9) approximately.

6. If x2 − xy + y 2 is to be approximated by a linear Taylor’s series polynomial about the point

(2, 3), then find a square with center at (2, 3) such that the error of approximation is less than
or equal to 0.1 in magnitude for all points within this square.