Properties of

Differentiable Functions

CALCULUS I
HELENA ALMEIDA, JOÃO FARINHA
PATRÍCIA XUFRE, PEDRO CHAVES
CALCULUS I

Agenda
1. Cauchy’s Rule
2. Differentiability in Vector Functions
3. Operations with Differentiable Functions
4. Differentiability and Continuity
5. Differentiability and Partial Derivatives
6. Differentiability and Directional Derivatives
7. 𝐶 𝑘 Function

2
CALCULUS I

1. Cauchy’s Rule
Let 𝑓: 𝐷𝑓 ⊂ ℝ → ℝ and 𝑔: 𝐷𝑔 ⊂ ℝ → ℝ be differentiable at an open interval 𝐼. If 𝑎 is one of the endpoints
of 𝐼 (finite or infinite), 𝑔′(𝑥) ≠ 0 in a neihgborhood of 𝑎 and lim 𝑓 𝑥 = lim 𝑔 𝑥 = 0 or ∞ , then:
𝑥→𝑎 𝑥→𝑎

𝑓 𝑥 𝑓′ 𝑥
lim = lim (if the latter limit exists)
𝑥→𝑎 𝑔 𝑥 𝑥→𝑎 𝑔′ 𝑥

Types of indeterminate forms:

0 ∞
0 ∞
1789 - 1857

3
CALCULUS I

Examples:


RC  
1
ln x  1
lim = lim x = lim =0
x → + x 5 x → + 5 x 4 x → + 5 x 5


RC  
1
0 
= lim+ x = lim+ (− x ) = 0
ln x
lim+ x. ln x = lim+
x →0 x →0 1 x →0 1 x →0
− 2
x x

4
CALCULUS I

∞
𝑪𝑹 1
00 0. ∞ lim ln 2𝑥 ∞ lim 𝑥1
𝑥 lim ln 2𝑥 𝑥 lim 𝑥ln 2𝑥 𝑥→0+ 1 +
𝑥→0 − 2 − lim 𝑥
lim+ 2𝑥 𝑥 = lim+ 𝑒 ln 2𝑥 =𝑒 𝑥→0+ =𝑒 𝑥→0+ =𝑒 𝑥 =𝑒 𝑥 =𝑒 𝑥→0+
𝑥→0 𝑥→0
0
=𝑒 =1

5
CALCULUS I

2. Differentiability of a Vector Function

𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ𝑚 is differentiable at 𝑎 if and only if each of its 𝑚 component functions is
differentiable at 𝑎.

Proof:
𝜕𝑓1 𝜕𝑓1 𝜕𝑓1
𝑎 𝑎 … (𝑎)
𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛
𝜕𝑓2 𝜕𝑓2 𝜕𝑓2
𝑓1 (𝑎 + 𝑢) 𝑓1 (𝑎) 𝑎 𝑎 … (𝑎) 𝑢1 𝑅1
𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛
𝑓2 (𝑎 + 𝑢) 𝑓 (𝑎) 𝑢2 𝑅2
= 2 + … +
… … …
… 𝑢𝑛 𝑅𝑚
𝑓𝑚 (𝑎 + 𝑢) 𝑓𝑚 (𝑎)
𝜕𝑓𝑚 𝜕𝑓𝑚 𝜕𝑓𝑚
𝑎 𝑎 … (𝑎)
𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛

Simplifying the right – hand side of the equation, we obtain a 𝑚 × 1 matrix which is the same as the matrix in
the left – hand side if and only if their entries are the same. And that is the same as saying that 𝑓𝑖 is
differentiable at 𝑎.

6
CALCULUS I

3. Operations with Differentiable Functions

Let 𝑓: 𝐷 ⊂ ℝ𝑛 → ℝ𝑚 , 𝑔: 𝐷 ⊂ ℝ𝑛 → ℝ𝑚 , ℎ: 𝐷ℎ ⊂ ℝ𝑚 → ℝ𝑝 , with 𝐶𝐷𝑓 ∩ 𝐷ℎ ≠ ∅ and 𝑎 ∈ 𝑖𝑛𝑡 𝐷 .

Sum:
If 𝑓 and 𝑔 are differentiable at 𝑎 , then 𝒇 + 𝒈 is differentiable at 𝑎.

Product:
Se 𝑓 and 𝑔 are differentiable at 𝑎 , then 𝒇. 𝒈 is differentiable at 𝑎.

Composition:
If 𝑓 is differentiable at 𝑎 and ℎ is differentiable at 𝑓(𝑎), then 𝒉 ∘ 𝒇 is differentiable at 𝑎.

7
CALCULUS I

4. Differentiability and Continuity

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ𝑚 and 𝑎 ∈ 𝑖𝑛𝑡 𝐷𝑓 .
If 𝒇 is differentiable at 𝒂 , then 𝒇 is continuous at 𝒂.

Proof:
If 𝑓 is differentiable at 𝑎, then each 𝑓𝑖 is differentiable at 𝑎, therefore, for each 𝑓𝑖 , with 𝑖 = 1, … , 𝑚, we have:

𝜕𝑓 𝜕𝑓
𝑅(𝑥−𝑎) 𝑓𝑖 𝑥 − 𝑓𝑖 𝑎 +𝜕𝑥 𝑖 𝑎 𝑥1 −𝑎1 +⋯+𝜕𝑥 𝑖 𝑎 𝑥𝑛 −𝑎𝑛 𝜕𝑓
lim (𝑥−𝑎) = 0 ⇔ lim 1 𝑛
= 0 ⇒ lim 𝑓𝑖 𝑥 = lim ቂ𝑓𝑖 𝑎 + 𝜕𝑥 𝑖 𝑎 𝑥1 − 𝑎1 +
𝑥→𝑎 𝑥→𝑎 (𝑥−𝑎) 𝑥→𝑎 𝑥→𝑎 1
𝜕𝑓
⋯ + 𝜕𝑥 𝑖 𝑎 𝑥𝑛 − 𝑎𝑛 ቃ ⇔ lim 𝑓𝑖 𝑥 = 𝑓𝑖 𝑎 ⇔ 𝑓𝑖 continuous at 𝑎 , 𝑖 = 1, … , 𝑚
𝑛 𝑥→𝑎

𝒇 is continuous at 𝒂

8
CALCULUS I

5. Differentiability and Partial Derivatives

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ and 𝑎 ∈ 𝑖𝑛𝑡 𝐷𝑓 .

If 𝑓 is differentiable at 𝒂 , then the 𝒏 partial derivatives of 𝑓 at 𝑎 exist.

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ.
If 𝑓 has the 𝒏 1st order partial derivatives, 𝑓 is differentiable at all points at which at least 𝒏 − 𝟏 of the
partial derivatives are continuous.

9
CALCULUS I

Schwarz-Young Theorem

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ.

The partial derivatives 𝒇′′𝒙𝒊 𝒙𝒋 and 𝒇′′𝒙𝒋 𝒙𝒊 are the same in all points at which 𝒇′𝒙𝒊 and 𝒇′𝒙𝒋 are
differentiable.

If all partial derivatives (𝑓′𝑥𝑖 , 𝑖 = 1, … , 𝑛 ) are differentiable, we have a symmetric hessian matrix.

10
CALCULUS I

6. Differentiability and Directional Derivatives

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ.
If 𝑓 is differentiable at 𝑎, then the derivative of 𝑓 along vector 𝑢 at point 𝑎 may be written as a linear
combination of the partial derivatives of 𝑓 at 𝑎.

𝑓 ′ 𝑢 𝑎 = 𝑓 ′ 𝑥 𝑎 𝑢1 + 𝑓 ′ 𝑥 𝑎 𝑢2 +…+𝑓 ′ 𝑥 𝑎 𝑢𝑛
1 2 𝑛

𝑢1
𝜕𝑓 𝜕𝑓 𝜕𝑓 𝑢2
𝑓 ′𝑢 𝑎 = 𝑎 𝑎 … 𝑎 … = 𝛻𝑓 𝑎 . 𝑢
𝜕𝑥1 𝜕𝑥2 𝜕𝑥𝑛
𝑢𝑛

Therefore, if 𝑓 is differentiable at 𝑎 then not only all partial derivatives of 𝒇 at 𝒂 must exist, but also all directional
derivatives of 𝒇 at 𝒂 must exist (as these result of a linear combination of the partial derivatives).

11
CALCULUS I

Example:
Let 𝑓: ℝ2 → ℝ be defined by 𝑓 𝑥, 𝑦 = 𝑦𝑥 2 + 𝑦. Compute the directional derivative of 𝑓 at point 2,1 along
vector 1, −1 .

𝑓 is differentiable in ℝ2 as it is a polynomial function

𝜕𝑓
2,1 = 2𝑦𝑥 2,1 =4 𝑓 ′ (1,−1) 2,1 = 𝑓 ′ 𝑥 𝑎 𝑢1 + 𝑓 ′ 𝑥 𝑎 𝑢2 = 4 × 1 + 5 × −1 = −1
𝜕𝑥 1 2

𝜕𝑓 OR
2,1 = 𝑥 2 + 1 2,1 =5
𝜕𝑦 𝜕𝑓 𝜕𝑓 1 1
𝑓 ′ (1,−1) 2,1 = 2,1 2,1 = 4 5 = −1
𝜕𝑥 𝜕𝑦 −1 −1

12
CALCULUS I

𝑓 𝑎 + 𝑡𝑢 − 𝑓 𝑎 𝑓 𝑎1 + 𝑡𝑢1 , 𝑎2 + 𝑡𝑢2 , … , 𝑎𝑛 + 𝑡𝑢𝑛 − 𝑓 𝑎1 , 𝑎2 , … , 𝑎𝑛

Proof: 𝑓 ′
𝑎 = lim = lim
𝑢 𝑡→0 𝑡 𝑡→0 𝑡

Using the tangent plane of 𝑓 at 𝑎 :

𝑓 𝑎1 + 𝑡𝑢1 , 𝑎2 + 𝑡𝑢2 , … , 𝑎𝑛 + 𝑡𝑢𝑛 = 𝑓 𝑎 + 𝑓´𝑥1 𝑎 𝑡𝑢1 + 𝑓´𝑥2 𝑎 𝑡𝑢2 + ⋯ + 𝑓´𝑥𝑛 𝑎 𝑡𝑢𝑛 + 𝑅 𝑡𝑢

So, going back to the definition of directional derivative:

𝑓 𝑎 + 𝑓´𝑥1 𝑎 𝑡𝑢1 + 𝑓´𝑥2 𝑎 𝑡𝑢2 + ⋯ + 𝑓´𝑥𝑛 𝑎 𝑡𝑢𝑛 + 𝑅 𝑡𝑢 − 𝑓 𝑎 𝑅(𝑡𝑢)

lim =0
𝑓′
𝑢
𝑎 = lim = 𝑡→0 𝑡
𝑡→0 𝑡
𝑡 𝑓´𝑥1 𝑎 𝑢1 + 𝑓´𝑥2 𝑎 𝑢2 + ⋯ + 𝑓´𝑥𝑛 𝑎 𝑢𝑛 + 𝑅 𝑡𝑢
= lim = 𝑓´𝑥1 𝑎 𝑢1 + 𝑓´𝑥2 𝑎 𝑢2 + ⋯ + 𝑓´𝑥𝑛 𝑎 𝑢𝑛
𝑡→0 𝑡

𝑓 ′ 𝑢 𝑎 = 𝑓 ′ 𝑥 𝑎 𝑢1 + 𝑓 ′ 𝑥 𝑎 𝑢2 +…+𝑓 ′ 𝑥 𝑎 𝑢𝑛
1 2 𝑛

13
CALCULUS I

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ and 𝑎 ∈ 𝑖𝑛𝑡 𝐷𝑓 .

Greatest If 𝑓 is differentiable at 𝑎, then of all vectors with norm 1, the one which implies
Directional the greatest directional derivative at 𝑎 is the unit vector associated to the
Derivative gradient vector evaluated at that point.

max 𝑓 ′ 𝑢 𝑎 = 𝑓 ′ 𝛻𝑓(𝑎) 𝑎 = 𝑓 ′ 𝛻𝑓(𝑎)

෣ 𝑎
𝑢 =1 𝛻𝑓(𝑎)

Proof:

We know that 𝑓 ′ 𝑢 𝑎 = 𝛻𝑓 𝑎 . 𝑢 = 𝛻𝑓 𝑎 . 𝑢 . cos 𝜃 where 𝜃 is the angle between the gradient vector and
vector 𝑢. As 𝑢 = 1, 𝑓 ′ 𝑢 𝑎 has its greatest value when cos 𝜃 = 1, that is, when 𝜃 = 0. This means that 𝑓 ′ 𝑢 𝑎
is greatest when 𝑢 has the same direction and orientation as the gradient vector.

14
CALCULUS I

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ and 𝑎 ∈ 𝑖𝑛𝑡 𝐷𝑓 .

Lowest If 𝑓 is differentiable at 𝑎, then of all vectors with norm 1, the one which implies
Directional the lowest directional derivative at 𝑎 is the unit vector associated to the
Derivative symmetric of the gradient vector evaluated at that point.

min 𝑓 ′ 𝑢 𝑎 = 𝑓 ′ −𝛻𝑓(𝑎
෣ 𝑎
𝑢 =1

Proof:

We know that 𝑓 ′ 𝑢 𝑎 = 𝛻𝑓 𝑎 . 𝑢 = 𝛻𝑓 𝑎 . 𝑢 . cos 𝜃 where 𝜃 is the angle between the gradient vector and
vector 𝑢. As 𝑢 = 1, 𝑓 ′ 𝑢 𝑎 has its lowest value when cos 𝜃 = −1, that is, when 𝜃 = π. This means that 𝑓 ′ 𝑢 𝑎
is lowest when 𝑢 has the same direction and the opposite orientation of the gradient vector.

15
CALCULUS I

Seja 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ and 𝑎 ∈ 𝑖𝑛𝑡 𝐷𝑓 .

Zero
If 𝑓 is differentiable at 𝑎, then of all vectors, the ones which implie a zero
Directional directional derivative at 𝑎 are the vectors which are perpendicular to the
Derivative gradient vector evaluated at that point.

𝑓 ′ 𝑢 𝑎 = 0 ⇔ 𝑢 ⊥ 𝛻𝑓 𝑎 ⇔ 𝑢. 𝛻𝑓 𝑎 =0

Proof:

We know that 𝑓 ′ 𝑢 𝑎 = 𝛻𝑓 𝑎 . 𝑢 = 𝛻𝑓 𝑎 . 𝑢 . cos 𝜃 where 𝜃 is the angle between the gradient vector and
𝜋
vector 𝑢. If we want that 𝑓 ′ 𝑢 𝑎 = 0, if 𝑢 > 0, we need that cos 𝜃 = 0, that is, that 𝜃 = . This means that
2
𝑓 𝑢 𝑎 = 0 when vector 𝑢 is perpendicular to the gradient vector.
′

16
CALCULUS I

17
CALCULUS I

7. 𝑪𝒌 Function
Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ.
𝒇 is a 𝑪𝒌 function at 𝒂 if all its partial derivatives up to the order 𝒌 are continuous at 𝒂. If
𝑓 is 𝐶 𝑘 , it also is 𝐶 𝑘−1 , 𝐶 𝑘−2 , … , 𝐶 1 .

Example: 𝑔: 𝐷𝑓 ⊂ ℝ2 → ℝ

𝜕𝑔 𝜕𝑔
If 𝜕𝑥 , 𝜕𝑥 are continuous functions, then 𝑔 is 𝐶 1 .
1 2

𝜕2 𝑔 𝜕2 𝑔 𝜕2 𝑔 𝜕2 𝑔
If, besides that, , , , (second order partial derivatives) are also continuous, then
𝜕𝑥1 𝑥1 𝜕𝑥1 𝑥2 𝜕𝑥2 𝑥1 𝜕𝑥2 𝑥2
𝑔 is 𝐶 2 .

18
CALCULUS I

𝑪∞ Functions

The following families of functions are 𝑪∞ at any interior point of their domains, as all their
partial derivatives of any order are continuous at that point.

- Polynomial Functions
- Rational Functions
- Exponential Functions
- Logarithmic Functions

19
CALCULUS I

Let 𝑓: 𝐷𝑓 ⊂ ℝ𝑛 → ℝ and 𝑎 ∈ 𝑖𝑛𝑡 𝐷𝑓 .

If 𝑓 is 𝑪𝟏 at 𝑎, then 𝑓 is differentiable at 𝑎.

Proof: If 𝑓 is 𝐶 1 at 𝑎, all its first order partial derivatives are continuous at 𝑎. Well, to ensure the
differentiability of 𝑓 at 𝑎, we only need that all partial derivatives exist and 𝑛 − 1 of them are
continuous. In this case, ALL partial derivatives are continuous at 𝑎.

If 𝑓 is 𝑪𝒌 at 𝑎, then 𝑓 is differentiable at 𝑎.

Observation: If 𝑓 is 𝐶 𝑘 at 𝑎, it also is 𝐶 1 at 𝑎, therefore it is differentiable at 𝑎.

If 𝑓 is 𝑪𝒌 at 𝑎, its partial derivatives of order lower than or equal to 𝒌 (evaluated at

𝑎) do not depend on the derivation order.

20