Journal of Discrete Mathematical Sciences and

Cryptography

ISSN: 0972-0529 (Print) 2169-0065 (Online) Journal homepage: https://www.tandfonline.com/loi/tdmc20

Partial differential equation calculation and

visualization

Li Li

To cite this article: Li Li (2017) Partial differential equation calculation and visualization,
Journal of Discrete Mathematical Sciences and Cryptography, 20:1, 217-229, DOI:
10.1080/09720529.2016.1178918

To link to this article: https://doi.org/10.1080/09720529.2016.1178918

Published online: 27 Dec 2016.

Submit your article to this journal

Article views: 93

View related articles

View Crossmark data

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tdmc20
Journal of Discrete Mathematical Sciences & Cryptography
Vol. 20 (2017), No. 1, pp. 217–229
DOI : 10.1080/09720529.2016.1178918

Partial differential equation calculation and visualization

Li Li *
College of Mathematics and Computer Science
Yan’an University
Yan’an 716000
China

Abstract
The calculation of partial differential equations has wide applied background. Many
of the problems in scientific research and engineering can be the mathematical model of
partial differential equations. Differential equation containing multiple independent vari-
ables is called as PDE (partial differential equation). As for PDE problem, it is very difficult
to solve. Apart from a few special cases, the vast majority of cases are difficult to obtain
exact solutions. For the majority of applications workers, from the point of partial differen-
tial equations, using the finite element method or finite difference method has to spend a
lot of effort to get numerical solution. By MATLAB PDEToolbox, the article introduced the
accurate solution process of two-dimensional problem with high-speed. In order to see the
relationship between graph, function value and independent variable from the mathematical
expression of partial differential equations by MATLAB programming, numerical solution,
and the results can be visualized. PDE Toolbox provides a powerful and flexible as well as
practical environment for research and problem solving about partial differential equations
in two-dimensional space.

Keywords: Partial differential equations, Difference method, Spectral method, MATLAB

1. Introduction
PDE mainly means the general variation law based on unknown
function and its derivatives to describe the physical quantities of objective
world. In theory, the solution of partial differential equations for research
has a long history. The initial research focused on physics, mechanics,
geometry and other aspects of specific issues; the classic representation

*E-mail: [email protected]

©
218 L. LI

is wave equation, heat conduction equation and potential equation

(harmonic equation). Through the study of these issues, an effective
method which is still used is formed, for example, separation of variables,
fourier transform method. Early research of partial differential equations
focused on theory, but in practice, research methods and findings are
difficult to be widely used. The main method to solve is: finite difference
method [5], the finite element method [3], spectral methods. However, for
the majority of applications workers, from the point of partial differential
equations, using the finite element method or finite difference method
has to spend a lot of effort to get numerical solution, in order to get
numerical solution. Now, MATLAB PDEToolbox has achieved high-speed
and accurate solution process of the spatial two-dimensional problem [2].
With the physical sciences studied phenomenon extended in both breadth
and depth, application scope of partial differential equations is more
extensive. From a mathematical point of view, to solve partial differential
equations can promote mathematical theory, variational method, series
expansions, various aspects of ordinary differential equations, algebra,
differential geometry, etc. development. From this perspective, partial
differential equations became the center of mathematics. Many of the
problems in scientific and engineering can be a mathematical model
of partial differential equations. Equations which contain multiple
independent variables are called PDE (partial differential equation). PDE
problem is very difficult to solve. Apart from a few special cases, the vast
majority of cases are difficult to obtain exact solutions.

2. Calculation of Partial Differential Equation

2.1 Theory of Partial Differential Equation
If the unknown function which appears in a differential equation
contain only one argument, the equation is called ordinary differential
equations, also referred to as the differential equation; if a differential
equation appears the partial derivative function of a pluralistic differential
equations, or if the unknown function and several variables are related,
and the unknown function of equation appear the derivative of several
variables, then this differential equation is a partial differential equation.
A partial differential equation is usually expressed as follows:

AΦ xx + BΦ xy + CΦ yy = f (x , y , Φ , Φ x , Φ y )
PDE CALCULATION AND VISUALIZATION 219

In the formula, A, B, Care the constant，called quasilinear number.

Typically, there are three kinds of quasi linear equations:

Hyperbolic equation: B2 − 4 AC > 0;

Parabolic equation: B2 − 4 AC = 0;

Elliptic equation: B2 − 4 AC < 0.

2.2 Partial differential equations solving by difference method

Difference method, also known as finite difference method or mesh
method, is one of the most widely used method in seeking the numerical
solution of partial differential equations about definite solution problems.
The basic idea is: first is mesh generation to solve the region, then
replace the finite discrete points (grid points) in continuously variable
region of independent variables, and replace the function of continuous
variable appears in problems with the discrete variables function in the
grid points; by replacing derivative with the difference coefficient of
function in grid point, then transform partial differential equation solving
problem with continuous variables into algebraic equations containing
only a finite number of unknowns (referred to as differential format).
If differential format has solution, and the grid point becomes smaller
and the solution is converges to the solution of the original differential
equation problem of definite solution, therefore the solution of difference
format is the approximation (numerical solution) of the original problem.
Therefore, using difference method for definite solution problem of
the partial differential equations generally needs to address the following
issues:

(1) Select the grid;

(2) 
Choose difference approximation for differential equations and
boundary conditions, list difference scheme;
(3) Solve the differential form;
(4) Discuss convergence and error estimate of difference scheme solu-
tions for differential equations.
220 L. LI

Here we use a simple example to illustrate the basic concepts of a

differential method and general process of differential methods for solving
partial differential equations.
Set a first order hyperbolic equation initial value problem.

∂u ∂u
 +a =0 t > 0, − ∞ < x < +∞
 ∂t ∂x (1)
u(x , 0) = ϕ (x )

Select Grid:
The difference schematic diagram is shown as Figure 1.
First, for a given area D = {(x , t) −∞ < x < +∞ , t ≥ 0}, proceed meshing
solution, the easiest used grid is equidistant linear axis parallel to x axis
and t axis respectively.

x = xk = kh, t = t j = jτ (k = 0, ±1, ±2 , j = 0,1, 2, )

Divide D into many small rectangular areas. These straight lines are
called grid lines, whose intersection are called mesh points, also referred
to as nodes, h and t are referred to the step length of direction x and t. This
grid is called a rectangular grid.

(1) 
Choose difference approximation for differential equations and
definite conditions, list difference scheme. If we use the forward
difference quotient represents the first-order partial derivatives,
namely:

Figure 1
Differenence schematic diagram
PDE CALCULATION AND VISUALIZATION 221

∂u u(xk +1 , t j ) − u(xk , t j ) h
= − ux′′2 (xk + θ 1h, t j )
∂x ( x ,t j )
h 2
k
(2)
∂u u(xk , t j +1 ) − u(xk , t j ) τ
= − ut′′2 (xk , t j + θ 2τ )
∂t ( x ,t j )
τ 2
k

In which 0 < θ 1 , θ 2 < 1.

Equation :

∂u ∂u
+a =0
∂t ∂x

The node (xk, tj) can be expressed as：

u(xk , t j +1 ) − u(xk , t j ) u(xk +1 , t j ) − u(xk , t j )

+a
τ h
τ ah
= ut′′2 (xk , t j + θ 2τ ) + ux′′2 (xk + θ 1h, t j )
2 2
= R(xk , t j ) (k = 0, ± 1, ± 2,  , j = 0,1, 2, )

Among them： u(xk , 0) = ϕ (xk ) (k = 0, ±1, ±2, ). Because when h,

t is small enough, the formula will omit R(xk, tj), you get a difference
equation approximating equation:

uk , j +1 −uk , j uk +1, j − uk , j
+a =0
τ h

here, uk, j can be seen as an approximation at node (xk, tj) when solving
problem. With initial conditions:

uk ,0 = ϕ (xk ) (k = 0, ± 1, ± 2, )

Combined, you get a differential format on numerical solution of the

problem.

equation： R(xk , t j ) = τ u′′2 (xk , t j + θ 2τ ) + ah u′′2 (xk + θ 1h, t j ) = O(τ + h)

2 t 2 x
222 L. LI

are called truncation error of differential equations. If the truncation error

for a differential equation is R = O(τ q + h p ), it is called that a differential
equation is the order of accuracy q to t, and the order of accuracy p is for x.
Obviously, the larger order of truncation error is, the better approximation
of differential equations is on differential equations.
If the grid spacing tends to be 0, truncation error of differential
equations tends to 0 , then differential equations and corresponding
differential equations are compatible. This is a necessary condition
for using the finite difference method to solve the problem of partial
differential equations.
If the grid spacing tends to be 0, the solution of difference schemes
converges to the solution for corresponding problem of definite solution
by differential equations, and then such a difference scheme is convergent.

2.3 Spectral methods for solving partial differential equations

Spectral method is originated in Ritz-Galerkin method, which is
based on orthogonal polynomial (trigonometric polynomial, Chebyshev
polynomials, Legendre polynomials, etc.) as the basis functions’ Galerkin
method, Tau method or configuration method; they were called spectral
method, Tau method or the spectral method proposed (collocation method)
respectively, commonly known as spectral method. Spectral method is a
calculation method based on orthogonal function or intrinsic function as
approximated function. From the perspective of function approximation,
spectral can be divided into Fourier spectrum method, Chebyshev
method or Legendre method. The former applies to a cyclical problem;
the latter two methods are applicable to non-cyclical issues. And the basis
of these methods is to create space for group functions. Spectral method
is a calculation method based on orthogonal function or intrinsic function
as approximated function. Approximate spectrum can be divided into
two kinds of approximate methods, function approximation and equation
approximation. From the perspective of function approximation, spectral
spectrum can be divided into Fourier method, Chebyshev method or
Legendre method. The former applies to a cyclical problem; the latter two
methods are applicable to non-cyclical issues. From the perspective of the
approximate equation, spectral methods can be divided into Collocation
method in solving discrete physical space and Galerkin method in solving
a discrete spectrum space and Pseudo-spectral method in solving the
first discrete quadrature in physical space, and then transformed into the
spectrum space. Collocation method is suitable for nonlinear problems.
PDE CALCULATION AND VISUALIZATION 223

Galerkin method applies to linear problems, and Pseudo-spectral method

is for handling nonlinear terms of expanded equations. The characteristics
of spectral method are the spectrum precision to smooth exponential
function approximation; with fewer grid points to get high accuracy; no
phase error; for multiscale volatility issues; calculation accuracy better than
other methods. The changes of Fast Fourier proposed greatly promoted
the development of spectral method; so far, various spectral methods
of calculating format have been presented and were used to calculate
astronomy, electromagnetism, geography, and other issues.
When using the spectral method to solving partial differential
equations, you often need to use differential truncate function or
interpolation function and the following article will discuss this issue.
+∞

Assuming Su = ∑ ikuˆ kϕ k (x ) is the u’s Fourier series, therefore we can

obtain k =− ∞

PNu¢ = (PNu)¢,

Namely the cut-off and derivation are interchangeable, if u¢ ŒL2, then

Su¢ also converge to u¢ under the meaning of L2.

3. PDE visualization
3.1 PDE toolbox features
PDE Toolbox provides a powerful and flexible and practical
environment to research and solve the problem of partial differential
equations in two-dimensional space. PDE Toolbox features include:

(1) Set the PDE (PDE) definite solution of problem, which set a two-
dimensional area of ​​ a given solution, boundary conditions and
forms and coefficient of equations;
(2) Use finite element method (FEM) to solve the numerical solution of
PDE;
(3) Visualization of solutions.

Beginners and advanced researchers will all feel very convenient

when using PDE Too1box. As long as the reference PDE problem of definite
solution is correct, then after starting MATLAB, typing pdetool in the
command line of MATLAB workspace, the system of partial differential
equations will generate graphical user interface (Graphical User Interface,
224 L. LI

Jane immediately referred to as GUI), namely PDE solution graphical

environment from Toolbox (PDE Toolbox) and then you can draw on it
definite solution area, set the equations and boundary conditions, doing
meshing, solving, mapping, etc.

3.2 PDE toolbox solving equations type

The basic equations for solving PDE Toolbox have elliptic equations,
parabolic equations, hyperbolic equations, eigen value equation, elliptic
equations and nonlinear elliptic equations.

Elliptic equation: -∇ ⋅ (c∇u) + au = f , in Ω

The W is bounded plane region; c, a, f and unknown number u are

real (or complex) function defined on W.

∂u
Parabolic Equation: d − ∇ ⋅ (c∇u) + au = f , in Ω
∂t

∂ 2u
Hyperbolic equation: d − ∇ ⋅ (c∇u) + au = f , in Ω
∂t 2

Eigen value equation: −∇(c∇u) + au = λ du, in Ω

There d is a complex function defined on W, l is the unknown eigen

values. In parabolic equations and hyperbolic equations, the coefficients c,
a, f and d may be dependent on the time t.
We can solve nonlinear elliptic equations:

−∇ ⋅ (c(u)∇u) + a(u) = f (u), in Ω

In which c, a, f may be a function of the unknown function u.

PDE can also solve the following equations;

 ⋅ (c11 (u)∇u1 ) − ∇ ⋅ (c12 (u)∇u2 ) + a11u1 + a12 u2 = f1

−∇

−∇ ⋅ (c21 (u)∇u1 ) − ∇ ⋅ (c22 (u)∇u2 ) + a21u1 + a22 u2 = f1
PDE CALCULATION AND VISUALIZATION 225

You can use the command line to solve high-order equations. For
elliptic equations, you can use adaptive mesh algorithm and it also can be
combined and used with the nonlinear solution together.

3.3 Boundary conditions

(1) Dirichlet conditions: hu = r


(2) Neumann conditions: n ⋅ (c∇u) + qu = g

There n is a unit outside the normal vector on the boundary ∂W of W,
g, q, h and r is function defined in ∂W. For eigen value problems are only
limited to second condition: g = 0 and r = 0. For nonlinear case, factor g,
q, h and r depend on u; for parabolic equations and hyperbolic equations,
coefficients may depend on the time t.

For the case of equations and boundary conditions are:

(1) Dirichlet conditions:

h11u1 + h12 u2 = r1

h21u1 + h22 u2 = r2

(2) Neumann conditions:

 
n ⋅ (c11∇u1 ) + n ⋅ (c12∇u2 ) + q11u1 + q12 u2 = g1

 
n ⋅ (c21∇u1 ) + n ⋅ (c22∇u2 ) + q21u1 + q22 u2 = g2

(3) Mixed boundary conditions: h11u1 + h12 u2 = r1

 
n ⋅ (c11∇u1 ) + n ⋅ (c12∇u2 ) + q11u1 + q12 u2 = g1 + h11µ

 
n ⋅ (c21∇u1 ) + n ⋅ (c22∇u2 ) + q21u1 + q22 u2 = g2 + h12 µ

In which calculation of m should make the Dirichlet condition met.

In the finite element method, Dirichlet condition is also called essential
226 L. LI

boundary conditions; Neumann condition is called natural boundary

conditions.

4. Application
Solving Poisson equation on boundary problems on the unit circle:

{
- ∆u= 1, Ω= (x , y ) x 2 + y 2 < 1

}
 u ∂Ω = 0

Exact solution to this problem is:

(1 − x 2 − y 2 )
u(x , y ) =
4

If you use a graphical user interface (Graphical User Interface,

abbreviated as GUI), the first step of work is typing pdetool in the MATLAB
window, press the Enter key OK, and then PDE Toolbox window appears.
If you need to coordinate grid, click the Grid option to Options menu.
Proceed as follows:
Set boundary conditions (Figure 2); set equation (Figure 3), a solution
can be obtained after the decomposition equation with cross-sectional
grid (Figure 4).

Figure 2
The Condition of Set boundary conditions
PDE CALCULATION AND VISUALIZATION 227

Figure 3
The specification of set equation

Figure 4
A solution can be obtained after the decomposition equation
with cross-sectional grid

Compare with accurate solution. Click the Plot Parameters ... menu
option, open Plot Selection dialog, choose user entry in Height (3-D plot)
line Property drop-down box, and type u-(1-x.^2-y.^2)/4 in the line
of User entry input box, click the plot button to see the absolute error
graphics solution, as shown in Figure 5. We can see that at the boundary
the error is zero.
228 L. LI

Figure 5
The absolute error graphics solution after click the plot button
5. Conclusion
Solution of partial differential equations has become the core content
of scientific and engineering computing, including some large-scale
computing and many of them have become routine calculations. As for a
high-level technical computing language and interactive environment of
algorithm development, data visualization, data analysis and numerical
calculation, the MATLAB can solve mathematical problems faster. The
PDE is a very practical subject. For the issue of partial differential equations
in theoretical research and practical application, due to its complexity and
other reasons of proposed boundary and boundary conditions, to seek
analytical solution is very difficult or even impossible. It is essential to
take advantage of computer research corresponding numerical solution of
the problem. Programming for solving the whole process requires a good
theoretical basis and programming techniques from numerical partial
differential equations, and partial differential equations Toolbox provides
a powerful and flexible and practical environment to research and solve
partial differential equations in two-dimensional space. By means of this
tool, we can learn from cumbersome and common steps for solutions freed
to focus on core issues namely description of the problem, definitions and
simplifying the determination of the boundary conditions, solving and
precision control method selection, greatly improving the computational
efficiency.
PDE CALCULATION AND VISUALIZATION 229

References

[1] Li Ming. MATLAB Solution on partial differential equations[J]. Hu-

nan Agricultural Machinery, 2010, 37(3):89-91.
[2] Feng Guilian. Numerical Solution and Visualization by MATLAB of
Partial Differential Equations [J]. Compter Technology and Development,
2013, 23(13):120-123.
[3] Feng Guilian. Reality and Visualization of Finite Element Numerical
Method in MATLAB [J]. Software Engineer, 2015, 18(1):61-64.
[4] Jiang Jun, Zhu Shaoming, Shu Shi, A Software Offers Solution for
Partial Differential Equations-UG[J]. Computer Engineering and Appli-
cations. 2010(3):63-66.
[5] Pan Shulong, Sun Weifu.Multistep Finite Difference Methods for
Parabolic Partial Differential Equations[J]. Journal of Yantai Vocational
College, 2011, 17(1):76-80.
[6] Wang Ziyu, Ren Xingmin.Solution of Partial Differential Equations
in Friction Damping Computation[J]. Mechanical Science and Technol-
ogy.2010, 29(12):1669-1672.
[7] Shang Yueqiang. FreeFem + + and finite element computations of par-
tial differential equations [J]. Journal of Guizhou Normal College.2011,
27(12):1-5.
[8] Li Wuming, Characteristic method and its some applications in sol-
cing partial differential equations [J].Journal of Henan Polytechnic Uni-
versity, 2012, 31(2):235-238.
[9] 
YueqiangShang, YinnianHe.Parallel iterative finite element algo-
rithms based on full domain partition for the stationary Navier-Stokes
equations[J].Applied Numerical Mathematics, 2010, 60(7):719-737.
[10] YueqiangShang.New stabilized finite element method for time-de-
pendent incompressible fl-owproblems[J].International Journal for Nu-
merical Methods in Fluids, 2010, 62(2):166-187.

Received October, 2015

Revised January, 2016