Partial Differentials and Their Applications in Technology

Aldo1, Luqman Efendi1, M.Roihan Dzaki1

1
Physics Education, Faculty of Teacher Training and Educational Sciences, Sriwijaya University

Email: [email protected]

Abstract: Partial differential equations (PDEs)—encompassing the theory of partial derivatives and the
formulation and solution methodologies—are essential for the modeling of complex phenomena encountered
in the fields of science and engineering. This chapter offers a comprehensive review of the subject, tracing its
historical developments and mathematical foundations while delving into advanced computational
techniques. The topics covered include classical analytical methods such as separation of variables and
integral transform methods, alongside detailed and systematic calculations, with particular emphasis on the
Fourier transform method as applied to the heat equation. Furthermore, this chapter examines advanced
subjects such as fractional calculus, stability and bifurcation analysis, and contemporary data-driven
approaches, which include physics-informed neural networks and neural operator methods. Numerous case
studies are presented, accompanied by explicit computations. All references are formatted in accordance with
APA style.
Keywords: Analytical methods, Data-driven approaches, Fourier transform, Fractional calculus, Partial
differential equations
Introduction

Partial differential equations constitute the mathematical foundation for the analysis
of systems characterized by multiple variables. Partial derivatives quantify the
instantaneous rate of change of a function with respect to one variable while maintaining
the other variables at constant levels. When integrated into partial differential equations
(PDEs), these derivatives facilitate the modeling of phenomena such as heat conduction,
fluid dynamics, electromagnetic wave propagation, and chemical pattern formation
(Debnath, 2003; Podlubny, 1999). This chapter presents a thorough review of the theory
and applications of partial differential equations, accompanied by precise and accurate
calculations that can be readily transcribed into a Word document. Specifically, we offer a
comprehensive derivation of the Fourier transform method for solving the heat equation.

Historical Background and Mathematical Foundations

A. Historical Evolution
The origins of differential calculus can be traced back to the 17th century,
primarily due to the independent contributions of Sir Isaac Newton and Gottfried
Wilhelm Leibniz. Leibniz's introduction of differential notation, along with his
 conceptualization of infinitesimals, established the foundation for subsequent
advancements in the study of partial derivatives. By the 19th century, prominent
mathematicians such as Augustin-Louis Cauchy, Bernhard Riemann, and Joseph
Liouville rigorously formalized these foundational concepts. Notably, Liouville's work
on fractional derivatives in 1832 ultimately contributed to the development of the
contemporary field of fractional calculus, as highlighted by Kilbas, Srivastava, and
Trujillo (2006). Concurrently, Joseph Fourier's exploration of heat conduction and
Pierre-Simon Laplace's investigation of potential theory catalyzed the formulation of
partial differential equations (PDEs) as models for various physical phenomena, as
documented by Evans (2010) and Turing (1952).
B. Mathematical Definitions
Let f (x 1 , x 2 , … , x n) be a function of n independent variables. The partial derivative
of f with respect to xi is defined as:
∂f f ( x1 … , x i+ h ,… , x n )−f ( x 1 , … , x i , … , xn )
∂ xi
( x1 … , x n )=lim
h
h→ 0

This definition expresses the instantaneous rate of change of f in the xi -direction while
all other variables remain constant (Stewart, 2015).
C. Geometric Interpretation
∂f
Geometrically, the partial derivative represents the slope of the tangent line to
∂ xi
the curve obtained by intersecting the surface z=f (x 1 , x 2 ,… , x n ) with the plane in
which all variables except xi are held fixed. The gradient vector,

∇f= ( ∂∂xf , ∂∂ xy , … , ∂∂xf ) ,

1 n

points in the direction of the greatest rate of increase of f

D. Fundamental Differentiation Rules
Partial derivatives obey similar rules to single-variable calculus:
1) Linearity:
∂ ∂f ∂g
∂x
[ af ( x , y ) +bg ( x , y ) ] =a
∂x
( x , y )+ b
∂x
(x, y).
 2) Product Rule:
∂
∂x
[ f ( x , y ) g ( x , y ) ]= ∂∂ fx ( x , y ) g ( x , y )+ f ( x , y ) ∂∂ gx ( x , y ) .
3) Chain Rule: For z=f (u , v ) with u=g ( x , y ) and v=h ( x , y ) ,
∂ z ∂ f ∂ g ∂f ∂h
= + .
∂ x ∂u ∂ x ∂ v ∂ x
4) Example Calculation: Volume of a Cone
Consider the volume of a cone given by:
2
πr h
V ( r , h )=
3
∂V
Calculation of :
∂r
1. Write the function:
2
πr h
V ( r , h )= .
3
2. Differentiate with respect to r (treat h as constant):
∂ V πh d 2 πh 2 πrh
= ∙ ( r ) = ∙ 2 r= .
∂r 3 dr 3 3
∂V
Calculation of :
∂h
1. Differentiate with respect to h (treat r as constant):
2
∂V π r
= .
∂h 3
These calculations clearly demonstrate how changes in r or h affect the
volume.
E. Extensions: Fractional Calculus
Fractional calculus generalizes the concept of differentiation to noninteger orders,
which is particularly useful in modeling processes with memory and nonlocal effects.
1. Riemann–Liouville Fractional Derivative
For a function f (t) and order α such that n−1<α < n, the Riemann–Liouville
derivative is defined as:
 n t
1 d
n∫
α n−α −1
D f ( t )=
α ( t−τ ) f ( τ ) dτ .
Γ ( n−α ) dt α
1
For example, for α = and n=1:
2
1 t −1
1 d
2
D f ( t )= ∫ ( t−τ ) 2
f ( τ ) dr ,
()
0
1 dt 0
Γ
2

with τ ( 12 )=√ π .
2. Caputo Fractional Derivative
The Caputo derivative is defined as:
t
1
C α
D f ( t )=
α ∫
Γ ( n−α ) α
( t−τ )
n−α −1 n
f ( τ ) dr .

1
For α = and n=1
2
1 t −1
1
D f ( t )= ∫ ( t−τ ) 2 f ( τ ) dr .
C 2 '
0
√π 0
Partial Differential Equations (PDEs)

A. General Formulation
A PDE involves an unknown function u (x,y,...) and its partial derivatives. A
typical second-order PDE in two variables is written as:

( )
2 2 2
( ) ∂u ( ) ∂ u ( ) ∂ u ∂u ∂ y
A x,y 2
+2B x, y +C x , y 2
+F x , y,u, , =0.
∂x ∂x ∂ y ∂y ∂x ∂ y
B. Classification of PDEs
The type of a second-order PDE is determined by the discriminant:
2
D=B − AC .
1. Elliptic PDEs: D<0 (e.g., Laplace’s equation ∇ 2 u=0).
2. Parabolic PDEs: D=0 (e.g., the heat equation ut =α ∇2 u).
3. Hyperbolic PDEs: D>0 (e.g., the wave equation utt =c 2 ∇2 u ).
C. Analytical Methods
1. Separation of Variables (Detailed Example)
Consider the one-dimensional heat equation:
2
∂u ∂ u
=α 2 , 0< x< L, t> 0 ,
∂t ∂x
with boundary conditions:
u ( 0 , t ) =0 , u ( L , t )=0 ,
and initial conditional:
u ( x , 0 )=f ( x ) .
Step 1: Assume a Separable Solution:
u ( x , t )=X ( x ) T ( t ) .
Substitute into the PDE:
2
dT d X
X (x) =α T ( t ) 2
.
dt dx
Dividing byα X ( x ) T ( t ) yields:
2
1 dT 1 d X
= =− λ .
αT (t) dt X (x ) d X 2
Step 2: Solve the Temporal ODE:
The ODE for T (t):
dT
+αλT + 0.
dt
Its solution is:
−αλt
T ( t )=T 0 e ,
where T0 is an integration constant.
Step 3: Solve the Spatial ODE
The ODE for X (x):
2
d X
2
+ λX =0 ,
dx
with boundary conditions X (0) = 0 and X (L) = 0. The general solution is:
X ( x )= A sin ( √ λ x ) + B cos ( √ λ x ) .
Applying X (0) = 0 yields B = 0. Then X ( L )=0 implies:
 A sin ( √ λ L )=0.
For nontrivial A, we require:

( )
2
nπ
√ λ L=nπ → λ= , n=1, 2 , 3 , … ,
L

and: X n ( x )=sin ( nπxL ).

Step 4: Construct the General Solution
The full solution is given by the Fourier sine series:

nπx −α ( L )
2t
∞ nx

u ( x , t )=∑ b n sin
n=1
( )
L
e ,

where the Fourier coefficients bn are

L
2
b n= ∫ f ( x )sin
L 0
nπx
L
dx .( )
2. Fourier Transform Method(step by step)
Consider the heat equation on the entire real line:
ut ( x , t )=α u xx ( x , t ) ,−∞< x< ∞, t> 0
with initial condition:
u ( x , 0 )=f ( x ) .
The solution is :
2
−x
1
u ( x , t )= e 4 αt .
√ 4 παt
This function is the fundamental solution (or Green’s function) for the heat
equation with an impulsive initial condition.

4.Advanced topics

1. Fractional Partial Differential Equations

Fractional PDEs incorporate derivatives of noninteger order to capture memory
effects and nonlocal behavior. For instance, a time-space fractional diffusion
equation can be written as:
 ∂ au
= -K(-∆ ¿ ¿ β u
∂ ta
with 0 <α ≤1 and 0 < β ≤ 10 < . Numerical methods such as spectral techniques
are employed to solve these equations (Kilbas et al., 2006).

Machine Learning Approaches

Physics-Informed Neural Networks (PINNs)
PINNs embed the governing PDE directly into the loss function of a neural
network. For the heat equation, the loss function is often defined as:
L(θ ) = Ldata +LPDE

Where
N
1
Ldata =
N ∑ ¿u ( xi , ti )−uNN ( xi , ti ,θ)¿2
I=1

M
1 ∂ uNN ∂2
LPDE =
M ∑ ¿ ∂ t ( xj , tj ,θ )−a ax 2 ( xj, tj , θ)¿ 2
J =1

Automatic differentiation is used to compute the required derivatives (Raissi,

Perdikaris, & Karniadakis, 2019).

Neural Operator Methods

Neural operators generalize the mapping between function spaces. The Fourier
Neural Operator (FNO) parameterizes the kernel integration in Fourier space:
(k ∅ v ¿ ( x ) =f −1 (R ∅ ( k ) . ( fv )( k ))( x )

Linear stability analysis reveals conditions for Turing instabilities, leading to

spatial pattern formation (Turing, 1952; Murray, 2003).
High-Performance Computing and Optimization
Reaction–Diffusion Systems and Pattern Formation
Reaction–diffusion systems model the interplay between chemical reactions
and diffusion:
∂u 2
=Du∇ u+ f ( u , v )
∂t
∂u 2
=Du∇ v + g ( u , v )
∂t

Large-scale PDEs require high-performance computing methods, including

parallel processing and domain decomposition. Techniques such as the finite
element method (FEM) and finite-difference time-domain (FDTD) are
implemented on GPUs and in parallel environments to accelerate computation.
PDE-constrained optimization using adjoint methods is crucial in design
applications (Antil et al., 2018).

5.Detailed Case Studies

Heat Conduction in Electronic Devices
Problem Statement
Consider a semiconductor chip with a two-dimensional domain:
Ω :0< x < L , 0< y < H
The steady-state heat conduction is governed by Laplace’s equation:

∇ 2 u ( x , y )+ 0
Boundary conditions:

U(0,y) = T0 and u(L,y) = TL (Dirichlet)

At y=H : convective boundary condition

∂u
-k ∨ y=H=h[u ( x , H )−T ∞ ]
∂y

where k is the thermal conductivity and h is the convection coefficient.

Analitycal derivation
1. Change of variabel define:

u(x,y) = v(x,y) + up(x)

Where

TL−T 0
up(x) = T0 +( )x
L

ensuring v(0,y) = v(L,y) = 0

2. Separation of variable for v(x,y) assume:

V(x,y) = X(x) Y(y)
Then
} (x)} over {X(x)} = {Y( y)
} left (x right ) Y left (y right ) +X left (x right ) {Y} ^ { X
X ( y )=0 ⇒ =0 ¿
Y ( y)

Set
{Y ( y)
X } (x)} over {X(x)} = -λ and
=λ ¿
Y ( y)
3. Solve the X(x) Equation:

X”(x) + λX ( x )=0 , X ( 0 ) =0 X ( L )=0

Solution:

(nπx) (nπ )2
X n ( x )=sin , λ n=
L (L)

4. Solve the Y(y) Equation:

Y”(y) - λ n Y ( y )=0

Solution

(nπy ) (nπy )
Y n ( y )= A n cosh + Bn sinh
(L) (L)
5. Apply convective boundary condition at y = H the deverivative of v (x,y)
with respect to y is :

∂v ( nπ ) ( nπy ) nπ ( nπy )
( x , y )= X ( x ) [ A n sinh + B n cosh ]
∂y ( L) ( L) L ( L)

At y = H subtitute u(x,H) = up(x) + v(x,H) into

∂u
-k ∨ y=H=h[u ( x , H )−T ∞ ]
∂y

(nπx )
And use the orthogonality of sin to determine An and Bn
L

6. Final series solution :

7.
N
(nπx)
U(x,y) = up(x) + ∑ sin ¿
n =1 ( L)
The series is truncated at N where the solution whitin a specified error
tolerance

Numerical Example

Babylonian clay tablet YBC 7289 (c. 1800–1600 BCE) with annotations.
The approximation of the square root of 2 is four sexagesimal figures, which
is about six decimal figures. 1 + 24/60 + 51/602 + 10/603 = 1.41421296...[1]
Before modern computers, numerical methods often relied on
hand interpolation formulas, using data from large printed tables. Since the
mid 20th century, computers calculate the required functions instead, but
many of the same formulas continue to be used in software algorithms.[5]
The numerical point of view goes back to the earliest mathematical writings.
A tablet from the Yale Babylonian Collection (YBC 7289), gives
a sexagesimal numerical approximation of the square root of 2, the length of
the diagonal in a unit square.
Numerical analysis continues this long tradition: rather than giving exact
symbolic answers translated into digits and applicable only to real-world
measurements, approximate solutions within specified error bounds are
used.