Indian Institute of Technology Kanpur
Department of Mathematics and Statistics
Lecture 4: Notes on several variables calculus
Semester: 2024-2025 (II)
By
Dr. Bidyut Sanki
Contents
1. Chain rule: The fundamental theorem 2
2. Properties of derivatives 3
3. Derivative of function in terms of component functions 4
4. Derivative of binary operation addition 4
5. Derivative of binary operation multiplication 5
1
2
1. Chain rule: The fundamental theorem
Theorem 1 (Chain rule). If f : Rn → Rm is differentiable at a ∈ Rn and g : Rm →
Rk is differentiable at b = f (a) ∈ Rm , then the composition g ◦ f : Rn → Rk is
differentiable at a. Furthermore, the derivative of g ◦ f at a is given by
d(g ◦ f )a = dgb ◦ dfa .
Exercise 1. Suppose f : Rn → Rn be a differentiable function with a differentiable
inverse. Show that (f −1 )0 (a) = [f 0 (b)]−1 where f (b) = a.
Proof of Theorem 1. Let dfa = Tf and dgb = Tg . We define
(1) φ(x) = f (x) − f (a) − Tf (x − a),
(2) ψ(y) = g(y) − g(b) − Tg (y − b),
(3) ρ(x) = g ◦ f (x) − g ◦ f (a) − Tg ◦ Tf (x − a).
Then
kφ(x)k
(4) lim = 0 and
x→a kx − ak
kψ(y)k
(5) lim = 0.
y→b ky − bk
kρ(x)k
To show: lim = 0. We have
x→a kx−ak
ρ(x) =g ◦ f (x) − g ◦ f (a) − Tg ◦ Tf (x − a)
=g(f (x)) − g(b) − Tg (f (x) − f (a) − φ(x)) by (1)
=g(f (x)) − g(b) − Tg (f (x) − b) + Tg (φ(x))
=ψ(f (x)) + Tg (φ(x))
kρ(x)k kψ(f (x))k kTg (φ(x))k
=⇒ lim ≤ lim + lim .
x→a kx − ak x→a kx − ak x→a kx − ak
Now, it suffices to prove
kψ(f (x))k
(6) lim = 0 and
x→a kx − ak
kTg (φ(x))k
(7) lim = 0.
x→a kx − ak
For equation (6), if > 0, it follows from (5) that for some δ > 0
kψ(f (x))k < kf (x) − bk, if kf (x) − bk < δ.
3
The inequality kf (x) − bk < δ holds true if kx − ak < δ 0 , for some suitable δ 0 > 0.
then
kψ(f (x))k <kf (x) − f (a)k
=kφ(x) + Tf (x − a)k
=⇒ kψ(f (x))k <kφ(x)k + M kx − ak.
Note that is an arbitrary positive real. Now, from equation (4) it follows that
kψ(f (x))k
lim = 0.
x→a kx − ak
kTg (φ(x))k kφ(x)k
Exercise: Prove inequality (7). Hint. lim ≤ M lim = 0 by (4).
x→a kx−ak x→a kx−ak
Exercise 2. Let g : R → R be a continuous function. Find f 0 for the following:
R x+y
(1) f (x, y) = a g.
R xy
(2) f (x, y) = a g.
R sin(x sin(y sin z))
(3) f (x, y, z) = xy g.
2. Properties of derivatives
2.0.1. Derivative of Constant function.
Theorem 2. If f : Rn → Rm is a constant function, then
Df (a) = dfa = 0, ∀a ∈ Rn .
Proof. Suppose f : Rn → Rm is constant function, i.e., for some b ∈ Rm , the function
is defined by following:
f (x) = b, for all x ∈ Rn .
Then we have
kf (a + h) − f (a) − 0(h)k kb − b − 0k
lim = lim = 0.
h→0 khk h→0 khk
Thus, we conclude that dfa = 0, ∀a ∈ Rn .
Derivative of linear transformation
Theorem 3. If f : Rn → Rm is a linear transformation, then
Df (a) = dfa = f, ∀a ∈ Rn .
4
Proof. Suppose f : Rn → Rm is a linear transformation.
To show:
dfa = f, ∀a ∈ Rn .
kf (a + h) − f (a) − f (h)k kf (a) + f (h) − f (a) − f (h)k
lim = lim = 0.
h→0 khk h→0 khk
Thus, we conclude that dfa = f, ∀a ∈ Rn .
3. Derivative of function in terms of component functions
Theorem 4. If f : Rn → Rm , f (x) = (f1 (x), . . . , fm (x)), then f is differentiable at
a point a ∈ Rn if and only if each fi : Rn → R, i = 1, . . . , m, is differentiable at a
and
dfa = ((df1 )a , . . . , (dfm )a ).
Thus f 0 (a) is the m × n matrix whose ith row is fi0 (a).
Proof. Suppose fi ’s are differentiable, for i = 1, . . . , m, and T = ((df1 )a , . . . , (dfm )a ).
Then
f (a + h) − f (a) − T (h)
=(f1 (a + h) − f1 (a) − (df1 )a (h), . . . , fm (a + h) − fm (a) − (dfm )a (h))
m
kf (a + h) − f (a) − T (h)k X |fi (a + h) − fi (a) − (dfi )a (h)|
=⇒ lim ≤ lim = 0.
h→0 khk h→0 khk
i=1
Conversely, suppose f is differentiable. Then by chain rule πi ◦f = fi is differentiable.
4. Derivative of binary operation addition
Theorem 5. If S : R2 → R is defined by S(x, y) = x + y, then dS(a,b) = S, ∀(a, b) ∈
R2 .
Proof. Let (x1 , y1 ), (x2 , y2 ) ∈ R2 and r ∈ R. Then
S((x1 , y1 ) + r(x2 , y2 )) =S(x1 + rx2 , y1 + ry2 )
=x1 + rx2 + y1 + ry2
=(x1 + y1 ) + r(x2 + y2 )
=S(x1 , y1 ) + rS(x2 , y2 ).
Therefore, S is linear. Hence, by Theorem 3, we conclude that dS(a,b) = S.
5
5. Derivative of binary operation multiplication
Theorem 6. If P : R2 → R is defined by P (x, y) = xy, then
dP(a,b) (x, y) = bx + ay, i.e., P 0 (a, b) = (b, a).
Before, going to the proof, we solve an exercise problem.
hk
Exercise 3. Show that lim = 0.
(h,k)→0 k(h,k)k
Solution. For (h, k) ∈ R2 , we have
(
h2 if k ≤ h
|hk| ≤
k2 if h ≤ k.
Therefore, |hk| ≤ h2 + k 2 and , then
|hk| h2 + k 2
lim ≤ lim = 0.
(h,k)→0 k(h, k)k (h,k)→0 k(h, k)k
hk
Thus we conclude that lim = 0.
(h,k)→0 k(h,k)k
Proof of Theorem 6. Consider the linear transformation T : R2 → R defined by
T (x, y) = bx + ay, ∀(x, y) ∈ R2 .
Then
P (a + h, b + k) − P (a, b) − T (h, k) (a + h)(b + k) − ab − bh − ak
lim = lim
(h,k)→0 k(h, k)k (h,k)→0 khk
hk
= lim
(h,k)→0 k(h, k)k
=0 (by Exercise 3).
Threfore, dP(a,b) = T.
Exercise 4. Let Ei , i = 1, . . . , k, be Euclidean spaces of various dimensions. A
function
f : E1 × · · · × Ek → Rp
is called multilinear if for each choice of xj ∈ Ej , i 6= j, the function g : Ei → Rp ,
defined by
g(x) = f (x1 , . . . , xi−1 , x, xi+1 , . . . , xk )
is a linear transformation.
6
(1) If f is multilinear and i 6= j, show that for h = (h1 , . . . , hk ) with hl ∈ El , we
have
kf (a1 , . . . , hi , . . . , hj , . . . , ak )k
lim = 0.
h→0 khk
k
P
(2) Show that df(a1 ,...,ak ) (x1 , . . . , xk ) = f (a1 , . . . , ai−1 , xi , ai+1 , . . . , ak ).
i=1
Corollary 5.1. If f, g : Rn → R are differentiable at a ∈ Rn
(1) D(f + g)(a) = d(f + g)a = dfa + dga and
(2) D(f g)(a) = d(f · g)a = f (a)dga + g(a)dfa
(3) Moreover, if g(a) 6= 0, then
f f g(a)dfa − f (a)dga
D (a) = d = .
g g a [g(a)]2
Proof. (1) We have the sum function f + g : Rn → R defined by (f + g)(x) =
f (x) + g(x), for all x ∈ Rn . Then S ◦ (f, g) = f + g, where (f, g) : R →
R2 , x 7→ (f (x), g(x)) and S : R2 → R, S(x, y) = x + y.
Now, by Chain Rule, we have
d(S ◦ (f, g))a (x) = dS(f (a),g(a)) ◦ d(f, g)a (x)
= S(dfa , dga )(x)
= dfa (x) + dga (x)
= (dfa + dga )(x)
=⇒ d(f + g)a = dfa + dga .
(2) Similarly,
f · g = P (f, g)
=⇒ d(f · g)a = dP(f (a),g(a)) d(f, g)a
= dP(f (a),g(a)) (dfa , dga )
= g(a)dfa + f (a)dga .
(3) Let h : R \ {0} → R be defined by h(x) = x1 . The function h is differentiable
at every point a ∈ R such that a 6= 0. Then fg = P (f, h ◦ g). Therefore, by
7
Chain Rule, we have
f
d = dPf (a), 1 (dfa , dhg(a) dga )
g a g(a)
dga
= dP f (a), 1
dfa , −
g(a) [g(a)]2
dfa f (a)dga
= −
g(a) [g(a)]2
g(a)dfa − f (a)dga
= .
[g(a)]2
Exercise 5 (Directional derivative). Let f : Rn → R. For x ∈ Rn the limit
f (a + tx) − f (a)
lim
t→0 t
if exists, denoted by (Dx f )a , is called the directional derivative of a in the direction
x.
(1) show that (Dei f )a = (Di f )a .
(2) Show that (Dtx f )a = t(Dx f )a .
(3) If f is differentiable at a, show that
(Dx f )a = dfa (x)
and therefore
(Dx+y f )a = (Dx f )a + (Dy f )a .
