First-Order Taylor Approximation in

Multiple Variables
1. Introduction
The First-Order Taylor Approximation is a fundamental concept in
multivariable calculus, allowing the approximation of a function near a
given point using a linear function. It generalizes the Taylor Series
Expansion from single-variable calculus to multiple dimensions, playing a
critical role in optimization, machine learning, and numerical analysis.

2. Differentiability in R
n

A function f : R n → R is differentiable at a point x ' if there exists a linear map

n
L :R → R such that:
lim f ( x ' + h )−f ( x ' )− L ( h )
h→0
=0
∥ h∥

where h=( h1 , h2 ,... , hn ) is a small perturbation vector, and ∥ h ∥ is its Euclidean

norm:

∥ h ∥=√ h21+ h22 +…+h 2n

2.1 Properties of Differentiability

 Linearity: The best linear approximation to f is given by a gradient
vector.

 Continuity: Differentiability implies continuity, but continuity does not

imply differentiability.

 Gradient Existence: The gradient exists and gives the direction of

maximum change.

3. First-Order Approximation Formula

If f is differentiable, the best linear approximation is given by:
L ( h )=∇ f ( x ' ) ⋅h

where ∇ f ( x ' ) is the gradient vector:

 ∇ f ( x ' )=
( ∂ f ∂f
,
∂ x1 ∂ x2
, ... ,
∂f
¿
∂ x n x' )
Thus, we obtain the First-Order Taylor Approximation:
f ( x '+h )=f ( x ' ) +∇ f ( x ' ) ⋅ h+ ϵ ( h )

where ϵ ( h ) is an error term satisfying:

lim ϵ ( h )
h→0
=0
∥h∥
This means that for sufficiently small h , we can approximate:
f ( x '+h ) ≈ f ( x ' ) + ∇ f ( x ' ) ⋅ h

4. Explicit Formulation in R
2

For a function f : R 2 → R , let:

( x , y ) =( x ' , y ' ) + ( h , k )
where h , k are small. Then:
f ( x ' , y ' ) +∇ f ( x ' , y ' ) ⋅ ( h , k ) ≈ f ( x '+ h , y ' + k )
Expanding the dot product:
∂f ∂f
f ( x , y )≈ f (x ' , y ' )+ ( x ' , y ' ) h+ ( x' , y ') k
∂x ∂y
Using vector notation:
f ( x , y ) ≈ f ( x ' , y ' ) +∇ f ( x ' , y ' ) ⋅ [ ( x , y )− ( x ' , y ' ) ]

5. Theoretical Framework and Principles

5.1 Principle of Local Linearity
The fundamental principle behind Taylor approximations is local linearity,
which states that a differentiable function behaves like a linear function
when viewed in a sufficiently small neighborhood.
5.2 Chain Rule and Partial Derivatives
The gradient of a function is built using partial derivatives, which
describe how the function changes with respect to each variable
independently.

5.3 Connection to the Total Differential

The total differential:
∂f ∂f
df = dx+ dy
∂x ∂y
is essentially the First-Order Taylor Approximation.

6. Patterns and Regularities

 Gradient as a Direction Indicator: The gradient always points in the
direction of maximum increase.

 Local vs. Global Behavior: The approximation holds only in a local

neighborhood.

 Higher-Order Corrections: The first-order term is the dominant

contributor in small perturbations.

7. Practical Relevance
7.1 Machine Learning & Optimization
 The gradient descent algorithm in machine learning relies on
Taylor approximations.

 Newton’s Method improves optimization using higher-order

derivatives.

7.2 Physics & Engineering

 Used to approximate functions in thermodynamics,
electromagnetism, and fluid dynamics.

 Applied in robotics and control systems to linearize nonlinear

models.

7.3 Computational Applications

 Essential in finite difference methods for numerical analysis.

 Used in scientific simulations and signal processing.

8. Historical Records and Expert Consultation
8.1 Historical Development
 Brook Taylor (1715) introduced Taylor Series.

 Joseph-Louis Lagrange extended Taylor’s work.

 Carl Gustav Jacobi and Augustin-Louis Cauchy formalized

gradient-based analysis.

8.2 Expert Insights

Mathematicians and engineers use Taylor approximations in fields like AI,
control systems, and finance. Experts emphasize the necessity of
considering higher-order terms for more accurate approximations.

9. Observation and Documentation

 Observing error behavior in numerical methods validates Taylor
expansions.

 Documenting convergence properties is critical in stochastic

gradient descent (SGD).

10. Variants and Extensions

 Second-Order Taylor Expansion: Includes the Hessian matrix for
quadratic approximation.

 Taylor Approximations in Function Spaces: Used in functional

analysis and PDE solutions.

 Multivariable Taylor Series: Extends beyond first-order to capture

curvature.

11. Practical Applications

 Control Theory: Designing PID controllers.

 Computer Graphics: Texture mapping approximations.

 Finance: Risk modeling in option pricing.

12. Conclusion
The First-Order Taylor Approximation provides a linear estimation of
differentiable functions and serves as a fundamental tool across
mathematics, engineering, and data science. Understanding its principles,
properties, and applications enhances problem-solving abilities in advanced
computational fields.

Understanding the First-Order Taylor Approximation in

Multiple Variables
1. Definition of Differentiability in Rn

A function f : R n → R is differentiable at a point x ' ∈ Rn if there exists a linear

map L :R n → R such that:
lim f ( x ' + h )−f ( x ' )− L ( h )
h→0
=0
∥ h∥
where h=( h1 , h2 ,… ,h n ) is a small perturbation vector, and ∥ h ∥ is its Euclidean
norm.
This definition ensures that the function can be well approximated by a linear
transformation when we zoom in close enough to x ' .
2. Best Linear Approximation via the Gradient
If f is differentiable at x ' , the best linear approximation is given by:
L ( h )=∇ f ( x ' ) ⋅h

where ∇ f ( x ' ) is the gradient vector, defined as:

∇ f ( x ' )= ( ∂∂xf , ∂∂xf , … , ∂∂xf ) ¿

1 2 n
x'

Thus, differentiability implies:

lim f ( x ' + h )−f ( x ' )−∇ f ( x ' ) ⋅h
h→0
=0 .
∥h∥
From this, for sufficiently small h , we can express:
f ( x '+h )=f ( x ' ) +∇ f ( x ' ) ⋅ h+ ε ( h ) ,

where the error term ε ( h ) satisfies:

 lim ε ( h )
h→0
=0 .
∥h∥
This means that as h approaches zero, the error term diminishes faster than
∥ h ∥, making the linear approximation highly accurate for small perturbations.
3. Explicit Formulation for R2

For a function f : R 2 → R , let:

( x , y ) =( x ' , y ' ) + ( h , k ) ,
where h and k are small perturbations. Differentiability ensures:
f ( x ' , y ' ) +∇ f ( x ' , y ' ) ⋅ ( h , k ) ≈ f ( x '+ h , y ' + k ) .
Expanding the dot product:
∂f ∂f
f ( x , y )≈ f (x ' , y ' )+ ( x ' , y ' ) h+ ( x' , y ') k .
∂x ∂y
Rewriting using vector notation:
f ( x , y ) ≈ f ( x ' , y ' ) +∇ f ( x ' , y ' ) ⋅ [ ( x , y )− ( x ' , y ' ) ] .

This is the first-order Taylor approximation in two variables.

4. Concrete Example
Consider the function:
2 2
f ( x , y )=x + xy+ y .
Let’s approximate f ( x , y ) near ( x ' , y ' )=( 1 ,2 ) .
1. Compute the partial derivatives:
∂f ∂f
=2 x + y , =x+ 2 y .
∂x ∂y
2. Evaluate at ( 1 , 2 ):
∂f ∂f
( 1, 2 )=2 ( 1 ) +2=4 , ( 1 ,2 ) =1+2 (2 )=5 .
∂x ∂y
3. The first-order approximation at ( x ' , y ' )=( 1 ,2 ) is:

f ( x , y ) ≈ f ( 1, 2 ) + 4 ( x−1 ) +5 ( y−2 ) .

Since f ( 1 , 2 )=12 + ( 1 ) ( 2 ) +22=7, the final approximation is:

 f ( x , y ) ≈ 7+ 4 ( x−1 )+5 ( y −2 ) .

For small perturbations around ( 1 , 2 ), this linear function provides an accurate

approximation of f ( x , y ).

5. Expected Questions & Problems

Here are some types of problems you may encounter:
1. Conceptual Understanding

o Explain why differentiability implies a linear approximation.

o How does the error term ε ( h ) influence the approximation?

o Compare differentiability and continuity in the context of Taylor

approximations.

2. Computational Exercises

o Compute the first-order Taylor approximation of a given function

at a specific point.

o Evaluate the accuracy of the linear approximation for small

perturbations.

3. Theoretical Questions

o Prove that if f is differentiable, then it is continuous.

o Derive the first-order Taylor approximation in three or more

variables.

4. Applications

o Use the first-order approximation to estimate function values in

physics or engineering problems.

o Analyze the error of approximation in optimization and machine

learning contexts.

Example 1: Linear Approximation Near a Point

Function: f ( x , y )=ln ( xy )

Point of Approximation: ( x 0 , y 0 )=( 1 , e )

Objective: Approximate f ( 1.02 , e−0.03 )
Solution:
1. Compute Partial Derivatives at ( 1 , e ) :
∂f y 1
o = =
∂ x xy x
∂f
 At x=1: =1
∂x
∂f x 1
o = =
∂ y xy y
∂f 1
 =
At y=e:
∂y e
2. Calculate Displacements:

o Δx=1.02−1=0.02

o Δy=( e−0.03 )−e=−0.03

3. Apply First-Order Taylor Approximation:

∂f ∂f
f ( x , y ) ≈ f ( x0 , y0 )+ Δx+ Δy
∂x ∂y

( 1e )(−0.03)
f ( 1.02 , e−0.03 ) ≈ ln ( 1 ⋅e )+ ( 1 )( 0.02 ) +

f ( 1.02 , e−0.03 ) ≈ 1+0.02−(

e )
0.03

4. Compute Numerical Value:

f ( 1.02 , e−0.03 ) ≈ 1+0.02− ( 2.718

0.03
)
f ( 1.02 , e−0.03 ) ≈ 1.02−0.011
f ( 1.02 , e−0.03 ) ≈ 1.009

Example 2: Estimating Temperature Change

Function: T ( x , y )=x 2+ xy+ y 2

Point: ( 2 , 3 )
Objective: Estimate the temperature at ( 2.1 , 2.9 )
Solution:
1. Compute Partial Derivatives at ( 2 , 3 ):
∂T
o =2 x+ y
∂x
∂T
 At ( 2 , 3 ): =4+3=7
∂x
∂T
o =x +2 y
∂y
∂T
 =2+6=8
At ( 2 , 3 ):
∂y
2. Calculate Displacements:

o Δx=0.1

o Δy=−0.1

3. Approximate Temperature Change:

∂T ∂T
ΔT ≈ Δx+ Δy
∂x ∂y
ΔT ≈ ( 7 ) ( 0.1 ) + ( 8 )(−0.1 ) =0.7−0.8=−0.1
4. Estimate New Temperature:

o Original T ( 2 , 3 )=22 + ( 2 ) ( 3 ) +32 =4+ 6+9=19

o Estimated T ( 2.1 , 2.9 ) ≈ 19+ (−0.1 )=18.9

Example 3: Error Estimation in Engineering

Function: V ( l, w , h )=lwh (Volume of a Box)
Measurements:
 Length (l ) = 50 cm (± 0.2 cm)

 Width (w ) = 30 cm (± 0.1 cm)

 Height (h ) = 20 cm (± 0.1 cm)

Objective: Estimate the maximum error in volume.

Solution:
1. Compute Partial Derivatives:
 ∂V
o =wh
∂l
∂V 2
 =( 30 ) ( 20 )=600 cm
∂l
∂V
o =lh
∂w
∂V 2
 =( 50 ) (20 )=1000 cm
∂w
∂V
o =lw
∂h
∂V 2
 =( 50 ) (30 )=1500 cm
∂h
2. Calculate Maximum Possible Errors:

o Δl=0.2 cm

o Δw=0.1 cm

o Δh=0.1 cm

3. Estimate Maximum Error in Volume ( ΔV ):

ΔV ≈ |∂∂Vl Δl|+|∂∂ Vw Δw|+|∂V∂h Δh|

3
ΔV ≈ ( 600⋅ 0.2 )+ ( 1000⋅ 0.1 )+ ( 1500⋅ 0.1 )=120+100+150=370 cm
So, the maximum possible error in the volume measurement is 370 cm³.

Example 4: Approximating Economic Output

Production Function: Q ( K , L ) =K 0.3 L0.7
Initial Inputs: Capital K=100 , Labor L=200
Changes:
 ΔK =5

 ΔL=−10

Objective: Estimate the change in output ( ΔQ ).

Solution:
1. Compute Partial Derivatives at ( 100 , 200 ):
 ∂Q −0.7 0.7
o =0.3 K L
∂K
∂Q −0.7 0.7
 =0.3 ⋅ ( 100 ) ⋅ ( 200 )
∂K
∂Q 0.3 −0.3
o =0.7 K L
∂L
∂Q 0.3 −0.3
=0.7 ⋅ ( 100 ) ⋅ ( 200 )
∂L
2. Calculate Numerical Values (Using Logarithms or Calculators):

o For brevity, let’s assume:

∂Q
 ≈a
∂K
∂Q
 ≈b
∂L
3. Estimate ΔQ :
ΔQ ≈ aΔK +bΔL
4. Compute ΔQ Numerically:

o Substitute the computed values of a and b and the given ΔK and

ΔL .

Example 5: Pressure Change in Physics

nRT
Equation of State: P ( V ,T )=
V
Initial Conditions:
 Volume V =10 L

 Temperature T =300 K

Changes:
 ΔV =0.1 L

 ΔT =5 K

Objective: Estimate ΔP .
Solution:
 1. Compute Partial Derivatives:
∂ P −nRT
o =
∂V V
2

∂ P nR
o =
∂T V
2. Evaluate at Initial Conditions:
∂ P −nR ⋅ 300 −30 nR
o = = =−0.3 nR
∂V (10 )2 100

∂ P nR
o = =0.1 nR
∂ T 10
3. Estimate ΔP :

ΔP ≈ (−0.3 nR ) ( 0.1 ) + ( 0.1 nR ) ( 5 )=−0.03 nR+0.5 nR=0.47 nR

So, the pressure increases by approximately 0.47 nR units.

Example 6: Linearization in Biology

V max S
Enzyme Reaction Rate: R ( S , E )=
Km+ S

Initial Conditions:
 Substrate Concentration S=50 µM

 Enzyme Concentration E=10 µg/mL

Changes:
 ΔS =5 µM

 ΔE=1 µg/mL

Objective: Estimate ΔR .
Solution:
1. Assuming V max =kE , where k is a constant.

2. Compute Partial Derivatives:

∂ R V max K m
o =
∂ S ( K m+ S )2
 ∂R kS
o =
∂ E K m+ S

3. Evaluate at Initial Conditions and Estimate ΔR .

Example 7: Approximating a Change in Gravitational Force

m1 m2
Newton’s Law of Gravitation: F ( r )=G 2
r
Initial Distance: r =100 km
Change in Distance: Δr=−1 km (Objects move 1 km closer)
Objective: Estimate ΔF .
Solution:
1. Compute Derivative:

dF m1 m2
o =−2G 3
dr r
2. Evaluate at r =100 km:

dF m1 m2
o =−2G
dr ( 100 )3
3. Estimate ΔF :

dF m1 m2 2 G m1 m2
ΔF ≈ Δr=−2 G 3
(−1 )=
dr ( 100 ) 1 , 000 , 000

2 G m1 m2
So, the gravitational force increases by units.
1, 000 , 000

Example 8: Error in Approximating Distance

Function: D ( x , y )= √ x 2 + y 2

Point: ( 3 , 4 )
Changes:
 Δx=0.1

 Δy=−0.1
Objective: Estimate ΔD .
Solution:
1. Compute Partial Derivatives:
∂D x
o =
∂ x √ x 2+ y 2

∂D 3
 At ( 3 , 4 ): = =0.6
∂x 5
∂D y
o = 2 2
∂ y √x + y

∂D 4
 At ( 3 , 4 ): = =0.8
∂y 5
2. Estimate ΔD :

ΔD ≈ ( 0.6 ) ( 0.1 ) + ( 0.8 ) (−0.1 )=0.06−0.08=−0.02

So, the distance decreases by approximately 0.02 units.

Example 9: Approximating a Function in Thermodynamics

Function: U ( S ,V )=a Sb V c
Given Constants: a , b , c
Initial Conditions: S=100, V =50
Changes:
 ΔS =5

 ΔV =−2

Objective: Estimate ΔU .
Solution:
1. Compute Partial Derivatives:
∂U b−1 c
o =ab S V
∂S
∂U b c−1
o =ac S V
∂V
2. Evaluate at Initial Conditions and Compute ΔU :
 ΔU ≈ ( ab Sb−1 V c ) ΔS+ ( ac S b V c−1 ) ΔV
3. Substitute Numerical Values to Find ΔU .

Example 10: Electric Field Approximation

1 q
Function: E ( x , y ) = 4 π ϵ
0 ( x + y2)
2

Point: ( 1 ,1 )
Changes:
 Δx=0.05

 Δy=−0.05

Objective: Estimate ΔE .
Solution:
1. Compute Partial Derivatives:

o
∂E
=
−2 x
(1
∂ x ( x2 + y 2 )2 4 π ϵ 0
q
)
∂E
o Similarly for .
∂y

2. Evaluate at ( 1 ,1 ) and Compute ΔE .

Example 11: Approximation in Financial Mathematics

Function: P ( R ,T )=P 0 e RT

Initial Conditions:
 Interest Rate R=0.05

 Time T =10 years

Changes:
 ΔR=0.001

 ΔT =0

Objective: Estimate ΔP .
Solution:
1. Compute Partial Derivative:
∂P RT
o =P0 T e
∂R
2. Compute ΔP :
∂P RT
ΔP ≈ ΔR=P0 T e ΔR
∂R
3. Substitute Numerical Values and Calculate ΔP .

Example 12: Estimating Change in Demand

Demand Function: Q ( P , I )=c P−d I e
Variables:
 Price P

 Income I

 Constants c , d ,e

Objective: Estimate ΔQ when P increases by 2 % and I increases by 1 %.

Solution:
1. Compute Elasticities:

o E P=−d

o E I =e

2. Estimate Percentage Change in Q :

ΔQ ΔP ΔI
≈ EP + EI
Q P I
ΔQ
≈ (−d ) ( 0.02 )+ e ( 0.01 )
Q
3. Compute ΔQ Using Q=c P−d I e .

Example 13: Change in Surface Area

Function: S ( r , h )=2 πrh+2 π r 2 (Surface Area of Cylinder)
Initial Measurements:
 Radius r =5 cm

 Height h=10 cm

Changes:
 Δr=0.1 cm

 Δh=−0.2 cm

Objective: Estimate ΔS .
Solution:
1. Compute Partial Derivatives:
∂S
o =2 πh+ 4 πr
∂r
∂S
o =2 πr
∂h
2. Evaluate at Initial Conditions:
∂S
o =2 π ( 10 ) + 4 π ( 5 )=20 π + 20 π =40 π
∂r
∂S
o =2 π ( 5 )=10 π
∂h
3. Estimate ΔS :

ΔS ≈ ( 40 π ) ( 0.1 ) + ( 10 π )(−0.2 )=4 π −2 π =2 π cm2

So, the surface area increases by approximately 6.283 cm2.

Example 14: Linearizing a Multivariable Function

Function: f ( x , y , z )=x sin ( yz )
Point: ( 1 , 0 , π )
Objective: Find the linear approximation near this point.
Solution:
1. Compute Partial Derivatives at ( 1 , 0 , π ) :
 ∂f
o =sin ( yz )
∂x

 At point: sin ( 0 )=0

∂f
o =xz cos ( yz )
∂y

 At point: ( 1 ) ( π ) cos ( 0 )=π

∂f
o =xy cos ( yz )
∂z

 At point: ( 1 ) ( 0 ) cos ( 0 ) =0
2. Linear Approximation:

f ( x , y , z ) ≈ f ( 1 ,0 , π ) +0 ⋅ ( x−1 ) + π ⋅ ( y−0 )+ 0 ⋅ ( z−π )

f ( x , y , z ) ≈ 0+ π ( y )

Example 15: Estimating Profit Change

Profit Function: Π ( p ,q )= pq−C ( q ), where C ( q )=a q2
Initial Conditions:
 Price p=$ 50

 Quantity q=100

 Cost coefficient a=0.2

Changes:
 Δp=$ 1

 Δq=2

Objective: Estimate ΔΠ .
Solution:
1. Compute Partial Derivatives:
∂Π
o =q
∂p
∂Π
o =p−2 aq
∂q
2. Evaluate at Initial Conditions:
 ∂Π
o =100
∂p
∂Π
o =50−2 ( 0.2 ) ( 100 )=50−40=10
∂q
3. Estimate ΔΠ :

ΔΠ ≈ (100 )( 1 ) + ( 10 )( 2 )=100+20=$ 120

So, profit increases by approximately $120.

Example 16: Change in Air Pressure with Altitude

Function: P ( h , T )=P 0 e− Mgh/ RT

Variables:
 h : Altitude

 T : Temperature

 Constants: P0 , M , g , R

Objective: Estimate ΔP when h increases by 100 m and T decreases by 2 K.

Solution:
1. Compute Partial Derivatives:
∂ P −Mg
o = P
∂h RT
∂ P Mgh
o = P
∂T RT2

2. Estimate ΔP Using Given Changes.

Example 17: Estimating Error in Area Measurement

Function: A ( r )=π r 2
Measurement: r =10 cm (± 0.05 cm)
Objective: Estimate the maximum error in area.
Solution:
 1. Compute Derivative:
dA
o =2 πr
dr
2. Estimate ΔA :
dA
ΔA ≈ Δr=( 2 π ⋅10 ) ( 0.05 ) =π
dr
So, the maximum error in area is approximately π cm².

Example 18: Linear Approximation in Meteorology

Function: P ( T , H )=a e−bT H c (Pressure depends on temperature T and
humidity H )
Objective: Estimate ΔP given small changes in T and H .
Solution:
1. Compute Partial Derivatives:
∂P −bT c
o =−ab e H
∂T
∂P −bT c−1
o =ac e H
∂H
2. Estimate ΔP Using Given Changes.

Example 19: Approximation in Optics

Lens Maker’s Equation:

1
f
= ( n−1 )(1
−
1
R1 R2 )
Objective: Estimate the change in focal length ( Δf ) when the radii of
curvature R1 and R2 change slightly.
Solution:
1. Compute Partial Derivatives with Respect to R1 and R2.

2. Use First-Order Approximation to Estimate Δf .

Example 20: Estimating Change in Sound Intensity
P
Function: I ( d )= 2
4πd
Objective: Estimate ΔI when distance d increases by 1 m.
Solution:
1. Compute Derivative:
dI −2 P −P
o = =
dd 4 π d 3 2 π d3
2. Estimate ΔI :
dI
ΔI ≈ Δd
dd

1. Linear Function in R2
 Function: f ( x , y )=3 x+ 4 y +5 .

 Point: ( x ' , y ' )=( 1 ,2 ) .

 Gradient: ∇ f =( 3 , 4 ) (constant).

 Taylor Approximation:

f ( 1+ h ,2+ k )=16+3 h+ 4 k .

 Error Term: ε ( h , k )=0.

0
 Differentiability: The limit lim ¿(h , k ) → 0 =0 ¿, so f is differentiable.
√ h +k 2
2

2. Quadratic Function
 Function: f ( x , y )=x 2+ y 2.

 Point: ( x ' , y ' )=( 1 ,1 ) .

 Gradient: ∇ f =( 2 x , 2 y ) =( 2 ,2 ) .

 Taylor Approximation:

f ( 1+ h ,1+k ) ≈ 2+2 h+2 k .

 Error Term: ε ( h , k )=h2 +k 2.

 2 2
h +k
=√ h +k →0 .
2 2
 Limit Analysis:
√h +k
2 2

3. Differentiable Non-C 1 Function

 Function:

f ( x , y )=
{ ( x 2+ y 2) sin
0
( x +1 y )
2 2
( x , y ) ≠ (0 , 0)

( x , y )=( 0 , 0 ) .

 Gradient at (0,0): ∇ f ( 0 , 0 )= ( 0 ,0 ).

 Differentiability:

lim ( h +k ) sin
2 2
( 2
1
2 ) =√ h +k sin
( h +k1 )→ 0 .
( h , k ) →0 h +k 2 2

√h +k
2 2 2 2

 Conclusion: Differentiable at ( 0 , 0 ) , but partial derivatives are

discontinuous.

4. Non-Differentiable Function with Partial Derivatives

 Function:

{
x3
( x , y ) ≠ ( 0 , 0)
f ( x , y )= x 2 + y 2
0 ( x , y )= ( 0 ,0 ) .

∂f ∂f
 Partial Derivatives at (0,0): ( 0 , 0 )=1, ( 0 , 0 )=0.
∂x ∂y
 Limit Analysis:

h3
lim 2 2
−h lim −h k 2
( h , k ) →0 h + k
= h →02 2 3/ 2 (depends on path).
√h +k
2 2
(h + k )
 Conclusion: Not differentiable at ( 0 , 0 ) .
5. Exponential Function
 Function: f ( x , y )=e x + y .

 Point: ( 0 , 0 ) .

 Gradient: ∇ f =( e x+ y , e x+ y )=( 1 ,1 ).

 Taylor Approximation:

f ( h ,k ) ≈ 1+h+ k .

( h+k )2
 Error Term: ε ( h , k )= +⋯.
2

( h+k )2
 Limit: →0 .
2 √ h2 + k 2

6. Trigonometric Function
 Function: f ( x , y )=sin ( x ) +cos ( y ) .

 Point: ( π /2 , 0 ).

 Gradient: ∇ f =( cos ( π /2 ) ,−sin ( 0 ) )=( 0 , 0 ) .

 Taylor Approximation:

f ( π2 +h , 0+ k) ≈ 2 .
2 2
−h k
 Error Term: − +⋯ .
2 2

−h 2+ k 2
 Limit: 2 .
→0
√h + k
2 2

7. Absolute Value Function (Non-Differentiable)

 Function: f ( x , y )=|x|+| y|.

 Point: ( 0 , 1 ).
∂f
 Partial Derivatives: ( 0 ,1 ) does not exist.
∂x
  Conclusion: Not differentiable at ( 0 , 1 ).

8. Polynomial with Cross Terms

 Function: f ( x , y )=x 2 y + x y 2.

 Point: ( 1 ,1 ).

 Gradient: ∇ f =( 2 xy + y2 , x 2+2 xy ) =( 3 ,3 ).

 Taylor Approximation:

f ( 1+ h ,1+k ) ≈ 2+3 h+3 k .

 Error Term: h2 + k 2+ 4 hk + ⋯.
2 2
h +k + 4 hk
 Limit: → 0.
√ h2 + k 2

9. Rational Function
x+ y
 Function: f ( x , y )= 2 2.
1+ x + y
 Point: ( 0 , 0 ) .

 Gradient: ∇ f ( 0 , 0 )= (1 , 1 ).

 Error Term:

h+k −( h+k ) ( h2+ k 2 )

ε ( h , k )= 2 2
−( h+ k )= 2 2
.
1+ h + k 1+h + k
ε (h , k )
 Limit: → 0.
√ h2+ k 2

10. Logarithmic Function

 Function: f ( x , y )=ln ( 1+ x + y ).

 Point: ( 0 , 0 ) .

 Gradient: ∇ f = ( 1+ 1x+ y , 1+ x1+ y )=( 1, 1).

 2
−( x+ y )
 Error Term: +⋯ .
2
2
−( x + y )
 Limit: 2 .
→0
√x +y
2 2

11. Discontinuous Function

 Function:

{
2
f ( x , y )= 1 y=x , x ≠ 0
0 otherwise .

 Point: ( 0 , 0 ) .

 Analysis: Discontinuous at ( 0 , 0 ) ; hence, not differentiable.

12. Homogeneous Function

1/ 3
 Function: f ( x , y )=( x 2 + y 2 ) .

 Point: ( 0 , 0 ) .

 Partial Derivatives: Do not exist at ( 0 , 0 ) .

 Conclusion: Not differentiable.

13. Directional Derivatives Exist but Not Differentiable

 Function:

{
x3
( x , y ) ≠ ( 0 , 0)
f ( x , y )= x 2 + y 2
0 ( x , y )= ( 0 ,0 ) .
3
a
 Directional Derivatives: D v f ( 0 , 0 ) = 2 2 for v= ( a , b ).
a +b
 Conclusion: Directional derivatives exist but f is not differentiable.
14. Hyperbolic Function
 Function: f ( x , y )=sinh ( x ) cosh ( y ) .

 Point: ( 0 , 0 ) .

 Gradient: ∇ f =( cosh ( x ) cosh ( y ) , sinh ( x ) sinh ( y ) )=( 1 , 0 ).

 Taylor Approximation:

f ( h ,k ) ≈ h .
3
h 2
 Error Term: + h k +⋯.
6
Error
 Limit: → 0.
√ h2+ k 2

15. Product of Functions

 Function: f ( x , y )=e x ln ( 1+ y ).

 Point: ( 0 , 0 ) .

 Gradient: ∇ f =( 0 ,1 ).

 Taylor Approximation:

f ( h ,k ) ≈ k .
 Error Term: hk +⋯ .
hk
 Limit: → 0.
√ h2+ k 2

16. Additive Exponential Function

 Function: f ( x , y )=x e y + y e x .

 Point: ( 0 , 0 ) .

 Gradient: ∇ f =( 1, 1 ).

 Error Term: 2 hk +⋯ .
2 hk
 Limit: → 0.
√ h2+ k 2
17. Square Root Function
 Function: f ( x , y )=√ 1+ x+ y .

 Point: ( 0 , 0 ) .

 Gradient: ∇ f = ( 12 , 12 ).
2
−( x+ y )
 Error Term: +⋯ .
8
2
−( x + y )
 Limit: 8 .
→0
√x +y
2 2

18. Function with Oscillatory Behavior

 Function:

{ x 2 sin ( 1x )+ y 2
x≠0
f ( x , y )=
2
y x =0 .

 Point: ( 0 , 0 ) .

 Gradient: ∇ f ( 0 , 0 )= ( 0 ,0 ).

 Error Analysis:
x2 sin ( 1x )+ y → 0.
2

√ x2 + y2
 Conclusion: Differentiable at ( 0 , 0 ) .

19. Function with Isolated Non-Differentiability

 Function: f ( x , y )=√ x 2+ y2 .

 Point: ( 0 , 0 ) .

 Analysis: Not differentiable at ( 0 , 0 ) (sharp point).

20. Function with All Directional Derivatives Zero
 Function:

{
x2 y 2
( x , y )≠ ( 0 , 0)
f ( x , y )= x 4 + y 4
0 ( x , y ) =( 0 , 0 ) .

 Directional Derivatives: Dv f ( 0 , 0 ) =0 for all v .

f (h , k )
 Differentiability: lim ¿(h , k ) → 0 ¿ does not exist (path-dependent).
√ h2 +k 2
 Conclusion: Not differentiable.

Key Takeaways:
1. Differentiability requires the error term ε ( h ) to vanish faster than ∥ h ∥.
2. Gradient existence (partial derivatives) is necessary but insufficient
for differentiability.
3. Discontinuous partials or directional derivative mismatches can
invalidate differentiability.
4. The first-order Taylor approximation is valid iff the function is
differentiable at the point.