Multivariable calculus - Revision Notes
Vibhor Verma1
1
Assistant Professor, Shri Ram College of Commerce, University of Delhi
The Rn space
We study functions z = f (x1 , x2 , ..., xn ) of n variables, i.e., the
functions of the points x = (x1 , x2 , ..., xn ) ∈ Rn .
The distance between two points x = (x1 , x2 , ..., xn ) and
x0 = (x10 , x20 , ..., xn0 ) in Rn is given by:
v
u n
0
uX
∥x − x ∥ = t (xi − xi0 )2
i=1
x0 = (x10 , x20 , ..., xn0 ) is an interior point of a set E ⊆ Rn if there is
an open-ball with center at x0 contained entirely in E , i.e.,
{x : ∥x − x0 ∥ < ϵ} ⊆ E
for some ϵ > 0.
A set of points is open if all its points are interior points. An open
ball is an open set.
A point x0 = (x10 , x20 , ..., xn0 ) ∈ E ⊂ Rn is a boundary point of E if
any open-ball centered at x0 contains points belonging to both E
as well as Rn − E or the complement of E in Rn .
A closed set contains all its boundary points.
A set E ⊂ Rn is bounded if there exists a real number M > 0
such that ∥x∥ ≤ M ∀x ∈ E , i.e., there exists an open-ball in Rn ,
with center at the origin, containing the set E .
A set that is both closed and bounded is a compact set.
Directional Derivatives
Consider a function F : E → R where E ⊆ Rn . Given the point
x0 = (x10 , x20 , ..., xn0 ) and the direction v = (v1 , v2 , ..., vn ), define
the directional function
g (t) = F (x) = F (x0 + tv) = F (x10 + tv1 , x20 + tv2 , ..., xn0 + tvn )
where t ∈ R.
If ∥v ∥ = 1, then the directional derivative of the function F in
the direction v is given by
dx1 dx2 dxn
g ′ (t) = F1′ (x)+ F2′ (x) + · · · + Fn′ (x)
dt dt dt
′ 0 dx1 ′ 0 dx2 dxn
= F1 (x + tv) + F2 (x + tv) + · · · + Fn′ (x0 + tv)
dt dt dt
0 ′ ′ 0 ′ 0 ′ 0
Dv F(x ) = g (0) = F1 (x )v1 + F2 (x +)v2 + · · · + Fn (x )vn
is the directional derivative at x0 = (x10 , x20 , ..., xn0 ).
The gradient of a function in Rn
Along the level set F (x) = c of F : E → R (E ⊆ Rn ), the gradient
of the function F at x0 ,
∇F (x0 ) = (F1′ (x0 ), F2′ (x0 ), ..., Fn′ (x0 ))
is orthogonal to the level curve F (x) = c at x0 = (x10 , x20 , ..., xn0 ).
Dv F(x0 ) = F1′ (x0 )v1 + F2′ (x0 +)v2 + · · · + Fn′ (x0 )vn
= (F1′ (x0 ), F2′ (x0 ), ..., Fn′ (x0 )) · (v1 , v2 , ..., vn )
= ∇F (x0 ) · v
= ∥∇F (x0 )∥∥v∥ cos ϕ
Note that the change in F at x0 is maximal if ϕ = 0, i.e., when the
movement along v is in the direction of the gradient ∇F (x0 ).
Thus, the gradient vector ∇F (x0 ) points in the direction of
maximal increase of the function F at x 0 , also referred to as the
preference direction.
Elasticity of Substitution
When F (x1 , x2 ) = c, the elasticity of substitution between x2
and x1 is given by
x2
σ2,1 = ElR2,1
x1
where
F1′ (x)
R2,1 =
F2′ (x)
is the marginal rate of substitution between x2 and x1 .
Homogenous and Homothetic Functions
Suppose that F is a function of n variables defined on a domain
D ⊆ Rn . Suppose that
x = (x1 , x2 , ..., xn ) ∈ D =⇒ tx = (tx1 , tx2 , ..., txn ) ∈ D ∀t > 0.
We say that F is homogeneous of degree k ∈ R on D if
F (tx1 , tx2 , ..., txn ) = t k F (x1 , x2 , ..., xn )
Euler’s Theorem: Suppose F is a function of n variables with
continuous partial derivatives in an open domain D, where t > 0
and x = (x1 , x2 , ..., xn ) ∈ D =⇒ tx = (tx1 , tx2 , ..., txn ) ∈ D. We
say that D is a cone. Then F is homogeneous of degree k ∈ R in
D if and only if the following equation holds for all
(x1 , x2 , ..., xn ) ∈ D:
n
X
xi Fi′ (x) = kF (x)
i=1
If F (x1 , x2 , ..., xn ) is homogeneous of degree k, then:
Fi′ (x1 , x2 , ..., xn ) is homogeneous of degree k − 1, (i = 1, 2, ..., n)
F (x1 , x2 , ..., xn ) = (x1 )k F (1, xx12 , ..., xxn1 ) = · · · = (xn )k F ( xxn1 , xxn2 , ..., 1)
provided that x1 , x2 , ..., xn > 0.
Pn Pn ′′
i=1 j=1 xi xj Fij (x1 , x2 , ..., xn ) = k(k − 1)F (x1 , x2 , ..., xn )
El1 (x1 , x2 , ..., xn ) + El2 (x1 , x2 , ..., xn ) + Eln (x1 , x2 , ..., xn ) = k where
Eli is the elasticity of F with respect to xi , (i = 1, 2, ..., n).
Let F be a function of n variables defined in a cone K . Then F is
called homothetic, provided that
x, y ∈ K , F (x) = F (y) =⇒ F (tx) = F (ty) ∀ t > 0
A homogeneous function F of any degree k is homothetic:
F (tx) = t k F (x) = t k F (y) = F (ty) ∀ t > 0.
The function G (x) = H(F (x)) where H is stirctly increasing and F
is homoegeneous of degree k is homothetic:
G (tx) = H(F (tx)) = H(t k F (x)) = H(t k F (y)) = H(F (ty)) =
G (ty) ∀ t > 0
Any strictly increasing function of a homothetic function is
homothetic.
The Extreme Value Theorem
Theorem: Let F (x) = F (x1 , x2 , ..., xn ) be defined over a set
S ⊂ Rn and let H be a function defined over the range of F .
Define G over S by
G (x1 , x2 , ..., xn ) = H(F (x1 , x2 , ..., xn ))
Then:
If F is increasing and c = (c1 , c2 , ..., cn ) maximizes
(minimizes) F over S, then c also maximizes(minimizes) G
over S.
If F is strictly increasing, then c = (c1 , c2 , ..., cn ) maximizes
(minimizes) F over S iff c also maximizes(minimizes) G over
S.
Extreme Value Theorem: If F is a continuous function over a
closed, bounded set S ⊆ Rn , then there exists both a maximum
c = (c1 , c2 , ..., cn ) and a minimum d = (d1 , d2 , ..., dn ) in S-that is,
there exist c and d such that
F (d) ≤ F (x) ≤ F (c) ∀ x ∈ S
Concave/ Convex Functions
A set S ⊆ Rn is a convex set if
x, y ∈ S =⇒ (1 − λ)x + λy ∈ S ∀ λ ∈ [0, 1]
A function F (x) = F (x1 , x2 , ..., xn ) defined on a convex set S is
concave in S if
x, x0 ∈ S =⇒ F ((1−λ)x0 +λx) ≥ (1−λ)F (x0 )+λF (x0 ) ∀ λ ∈ (0, 1)
A function F (x) = F (x1 , x2 , ..., xn ) defined on a convex set S is
strictly concave in S if
x0 , x ∈ S, x0 ̸= x =⇒ F ((1 − λ)x0 + λx) > (1 − λ)F (x0 ) + λF (x)
∀ λ ∈ (0, 1)
A function F (x) = F (x1 , x2 , ..., xn ) defined on a convex set S is
convex in S if
x, x0 ∈ S =⇒ F ((1−λ)x0 +λx) ≤ (1−λ)F (x0 )+λF (x0 ) ∀ λ ∈ (0, 1)
A function F (x) = F (x1 , x2 , ..., xn ) defined on a convex set S is
strictly convex in S if
x0 , x ∈ S, x0 ̸= x =⇒ F ((1 − λ)x0 + λx) < (1 − λ)F (x0 ) + λF (x)
∀ λ ∈ (0, 1)
F is convex in S iff −F is concave in S.
F is strictly convex in S iff −F is strictly concave in S.
Jensen’s inequaltiy: A function F of n variables is concave on a
convex set S ⊆ Rn iff the following inequality is satisfied for all
x1 , x2 , ..., xm ∈ S and all λ1 , λ2 , ..., λm with
λ1 + λ2 + · · · + λm = 1:
F (λ1 x1 + λ2 x2 + · · · + xm ) ≥ λ1 F (x1 ) + λ2 F (x2 ) + · · · + λm F (xm )
Multivariable Optimization
Local Maximum: The instrument vector x∗ is a local maximum if
it is in the oppurtunity set X ⊆ Rn and yeilds a value of the
objective function larger than or equal to that obtained by any
feasible vector sufficiently close to it:
F (x∗ ) ≥ F (x) ∀ x ∈ X ∩ Nϵ (x∗ )
where Nϵ (x∗ ) is the ϵ-neighbourhood of x∗ for some ϵ > 0,
however small, i.e., the set of all x∗ such that
∗ n ∗ 2
pP
∥x − x ∥ = i=1 (xi − xi ) < ϵ
x∗ is a strict local maximum if
F (x∗ ) > F (x) ∀ x ∈ X ∩ Nϵ (x∗ ), x ̸= x∗
Suppose that F (x) = F (x1 , x2 , ..., xn ) is a C 2 function in a domain
S ⊆ Rn . The matrix
H(x) = [Fij′′ (x)]n x n
is called the Hessian matrix of F at x. Denote the k th leading
principal minor of H(x) by Dk (x) (k = 1, 2, ..., n) and an arbitrary
principal minor of order r by ∆r (x) (r = 1, 2, ..., n).
Necessary and Sufficient Conditions for Local Extreme
Points: Suppose F (x1 , x2 , ..., xn ) is a C 2 function in a set S ⊆ Rn
and let x0 be an interior stationary point in S- that is,
Fi′ (x0 ) = 0 (i = 1, 2, ..., n)
x0 local maximum point =⇒ (−1)r ∆r (x0 ) ≥ 0 for all
principal minors ∆r (x0 ) ≥ 0 of order r = 1, 2, ..., n.
(−1)k Dk (x0 ) > 0, k = 1, 2, ..., n =⇒ x 0 local maximum
point.
x0 local minimum point =⇒ ∆r (x0 ) ≥ 0 for all principal
minors ∆r (x0 ) ≥ 0 of order r = 1, 2, ..., n.
Dk (x0 ) > 0, k = 1, 2, ..., n =⇒ x 0 local minimum point.
A stationary point x0 for which neither
(−1)k Dk (x0 ) > 0, k = 1, 2, ..., n nor
Dk (x0 ) > 0, k = 1, 2, ..., n is a saddle point.
F is concave in S ⇐⇒ H(x) = [Fij′′ (x)] is negative
semidefinite for all x ∈ S ⇐⇒ (−1)r ∆r (x) ≥ 0 for all
principal minors ∆r (x) ≥ 0 of order r = 1, 2, ..., n and all x ∈ S
F is convex in S ⇐⇒ H(x) = [Fij′′ (x)] is positive
semidefinite for all x ∈ S ⇐⇒ ∆r (x) ≥ 0 for all
principal minors ∆r (x) ≥ 0 of order r = 1, 2, ..., n and all x ∈ S
H(x) is negative definite for all
x ∈ S ⇐⇒ (−1)k Dk (x) > 0, k = 1, 2, ..., n ∀ x ∈ S =⇒ F is
strictly concave in S ⊆ Rn
H(x) is positive definite for all
x ∈ S ⇐⇒ Dk (x) > 0, k = 1, 2, ..., n ∀ x ∈ S =⇒ F is strictly
convex in S
Sufficient Conditions for Global Extreme Points:
Let x0 be an interior stationary point of a C 2 function F in a
convex set S ⊆ Rn .
(−1)r ∆r (x) ≥ 0 for all principal minors ∆r (x) ≥ 0 of order
r = 1, 2, ..., n and all x ∈ S =⇒ x0 is a maximum point for F in
S.
∆r (x) ≥ 0 for all principal minors ∆r (x) ≥ 0 of order r = 1, 2, ..., n
and all x ∈ S =⇒ x0 is a minimum point for F in S.
Quasi-concave/convex functions
The set of points in Rn for which the value of the objective
function is constant
{x ∈ Rn |F (x) = c}
is called a contour of the function F .
The set
τc = {x ∈ Rn |F (x) ≥ c}
is an upper contour set of F .
The set
{x ∈ Rn |F (x) ≤ c}
is a lower contour set of F .
The function F defined over a convex set S ⊂ Rn , is
quasi-concave if the upper level set τc = {x ∈ Rn |F (x) ≥ c} is
convex ∀ c ∈ R.
The function F defined over a convex set S ⊂ Rn , is
quasi-concave if the upper level set τc = {x ∈ Rn |F (x) ≥ c} is
convex ∀ c ∈ R.
The function F defined over a convex set S ⊂ Rn , is quasi-convex
if the lower level set {x ∈ Rn |F (x) ≤ c} is convex ∀ c ∈ R.
F is quasi-concave iff −F is quasi-convex.
If F (x) is concave, F (x) is quasiconcave.
If F (x) is convex, F (x) is quasiconvex.
Theorem: A function F defined over a conves set S ⊆ Rn is
quasi-concave iff, for all x, x0 ∈ S and all λ ∈ [0, 1],
F (x) ≥ F (x0 ) =⇒ F ((1 − λ)x + λx0 ) ≥ F (x0 )