my Partial differential equations cheat sheet

Nasser M. Abbasi

May 28, 2020 Compiled on September 26, 2023 at 4:21pm

Contents
1 introduction 1

2 linear PDE’s 4
2.1 Elliptic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Parabolic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Hyperbolic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

3 hints 5

4 references 6

1 introduction
These are part of my study notes on PDE’s.
Trying to classify PDE’s, here is current diagram. It is very large, but it is meant to include a
summary of many methods in one place. Easier to view in a browser than in the pdf.
Some diagrams I made

1-D PDE

ut  a ux
a wave speed
IC: ux, 0  u 0x
BC: periodic
analytical solution ux, t  u 0x  at

Forward time, centered

Lax-Friedrich Lax-Wendroff Crank-Nicolson Leapfrog Beam-Warming Upwind
space (FTCS)

1
 2

1-D PDE

1-D Parabolic (heat equation or diffusion)

Stiff, Implicit, small time step

u t  Du xx
D diffusion constant  0
IC: ux, 0  u 0 x
BC: u0, t  gt, u1, t  ft

FTCS (Forward-time, centered space) Crank Nicholson

(second order in space, first order in time) (second order in space, second order in time)

i u i
u n1 n u ni1 2u ni u ni1 
k
D
h2
u n1
i  u ni  Dk
u ni1  2u ni  u ni1 
h2
u n1
i Dk n
u  1  2 Dk2 u ni  Dk
u ni1
h 2 i1 h h2
Let r  Dk2
h
u i  ru ni1  1
n1
 2ru ni  ru ni1

i1 2u i u i1

u n1 n1 n1
i u i
u n1 n u ni1 2u ni u ni1 
k
 D
2

h2 h2

u n1
i  Dk
u n1
i1  2u n1
i i1   u i  2h 2 u i1
 u n1 n Dk n
 2u ni  u ni1 
2h 2
 Dk2 u n1
i1  1  h 2
Dk
u n1
i  Dk2 u n1
i1  2h 2 u i1
Dk n
 1 Dk
u ni  Dk
u ni1
2h 2h h2 2h 2
Let r  Dk2
2h
ru i1  1  2ru n1
n1
i i1  ru i1  1  2ru i  ru i1
 ru n1 n n n

Au n1  Bu n

1D PDE, parabolic, heat/

diffusion, C-N

1  2r r u n1
2 1  2r r u n2
r 1  2r r u n1
3 r 1  2r r u n3

r  r  r  r 
r 1  2r u n1
N1 r 1  2r u nN1
 3

second order linear par-

tial differential equation

elliptic parabolic hyperbolic

No characteristic curves. diffusion pro-
cess reached equilibrium, steady state
temperature distribution. Numeri-
cally solved by relaxation methods.
2
∂2u 2
A ∂∂xu2 + B ∂x∂y + C ∂∂yu2 + D ∂u
∂x
+ E ∂u
∂y
+ Fu = G
B 2 − 4AC < 0
A = 1, C = 1, B = 0

F = 0 F = k2
a function that satisfies
Laplace is called harmonic

G = 0 G = g(x, y) G = 0 G = g(x, y)
Laplace in 2D Poisson in 2D homogeneous Helmholtz inhomogeneous Helmholtz
∂2u 2 ∂2u 2
∂2u ∂2u ∂2u 2
∂x2
+ ∂∂yu2 = 0 ∂x2
+ ∂∂yu2 = g(x, y) ∂x2
+ ∂y 2
+ k2 u = 0 ∂x2
+ ∂∂yu2 + k2 u = g(x, y)
eigenvalue equilibrium steady state oscillation

linear second order PDE classification, solution is ux, y

A  u
 C  D u  E u
2 2 2
u
 2B xy u
x y
 Fu  G
x 2 y 2

elliptic
parabolic hyperbolic
B2  AC B2  AC B2  AC

A  1, C  1, B  0

F0 F  k2

G0 G  gx, y G0 G  gx, y

Laplace in 2D Poisson in 2D Homogeneous Helmholtz inhomogeneous Helmholtz
2 u 2 u 2 u 2 u 2 u 2 u
 0   fx, y  k u  0 2 2 u
 2 u
 k 2 u  gx, y
x 2 y 2 x 2 y 2 x 2 y 2
x 2 y 2

linear second order PDE classification, solution is ux, y

No characteristic curves
(diffusion process reached elliptic
parabolic hyperbolic
equilibrium, steady state
A  u
 C  D u  E u
2 2 2
temperature distribution,
u
 B xy u
 Fu  G
A  2u  B tx
 u
 C  D u  E u A  2u  B tx
 u
 C  D u  E u
2 2 2 2 2 2

Numerically, solved by
x 2 y 2 x y u
t x
 Fu  G u
t x
 Fu  G
t x 2 t x 2
relaxation methods) B2  4AC  0
B2  4AC  0 B2  4AC  0
A  1, C  1, B  0
A  0, B  0, C  k, D  1 A  1, B  0, C  c 2
 u
 c2 
2 2
u
 k
2 u
u
t x 2 t 2 x 2
c is wave speed
F0 Fk 2
k is called thermal diffusivity
Eigenvalue Evolution problem,
equilibrium propagation of waves
problem Initial disturbances
Steady state oscillations transition from the hyperbolic
PDE to the elliptic PDEs. propagate along the
G0 G  gx, y G0 G  gx, y diffusion problems, transient chracteristics
Laplace in 2D Poisson in 2D Homogeneous Helmholtz inhomogeneous Helmholtz flow of heat Two distinct
2 u 2 u 2 u 2 u 2 u 2 u Diffusion problems. characteristic directions
 0   fx, y   k2u  0 2 u
 2 u
 k 2 u  gx, y temperature in rod
x 2 y 2 x 2 y 2 x 2 y 2 supports solutions with
x 2 y 2
ONE characteristic discontinuitie: shock wave.
directions. Any t=0 is mechanical oscillator,
such line vibrating strings, convection
Solution methods driven transport problem
solutions to linear hyperbolic
equations as smooth as
Analytical Numerical boundary and initial
conditions
nonlinear hyperbolic PDE with
On square/rectangular, using separation of variables smooth boundary and initial
method, with essential BC, leads to series solution. conditions give rise to nonsmooth
See http://en.wikibooks.org/wiki/Partial_Differential_Equations/The_Laplacian_and_Laplace’s_Equation
solutions. stability issues.

Use Matrix splitting method to obtain the iterative equation

Direct solver Ax=b Matrix splitting methods Finite difference Finite elements

Generate A,b from

Ax  f To do
Rewritten as
scheme uised
In 2D, using standard 5 point laplacian. This gives
x k1  Tx k  Cf approximation of each grid point at each iteration.
where T is the iteration matrix More iterations improves the solution. Use residual
and k is the iteration number norm and epsilon to determine when to stop.

There are different

methods to generate the
iteration matrix T above,
based on different stationary solvers
1D splitting methods. non-stationary solvers
1-D, using 3-point central difference formula Different algorithms uses Drawn by Nasser M. Abbasi
different splitting. 7/17/2013
2
d u
 fx , u0  a, u1  b Jacobi Gauss- Gauss- SOR
Pdes.vsd
dx 2 SOR With Without
Seidel Seidel With
Chebyshev preconditioner preconditioner
Red/
black (best among
Multigrid methods
stationary Preconditioner method can
solvers) be NONE, SSOR, ILU0, ILUT,
Steepest
ILUTP, and others
descent

Conjugate GMRES
gradient BiCGSTAB

dirichlet Neumann boundary

boundary conditions (use image point
conditions method)

To derive the PDE, we start by setting up the state quantities and the flow quantities, and
relate these to each others by the use of the constitutive law. Then substiting this into the local
conservation law, lead to the PDE.
 4

state quantity constitutive law flow quantity

density tension
tempreture velocity
pressure ⇐⇒ momentum
specific internal energy heat flux
entropy

2 linear PDE’s
2.1 Elliptic
Some properties
1. Solution to the PDE represents steady state of u.
2. Only boundary conditions are used to solve. No initial conditions.
3. Relation to complex analytic functions: If f (z) = φ(x, y) + iψ(x, y) is analytic, then φ(x, y)
and ψ(x, y) are solutions to Laplace pde’s
4. Solutions to Laplace PDE are called harmonic functions.
5. constitutive law: Either consider them as stationary process, or take the time dependent
pde, and set those terms in that which depend on time to zero.
Examples of elliptic PDE’s
1. Laplace uxx = 0 or in general ∇2 u = 0
2. Poisson uxx = −f (x)
3. Helmholtz in 1D uxx + λu(x) = −f (x)
4. Helmholtz in 2D uxx + uyy + λu(x, y) = −f (x, y)

2.2 Parabolic
Some properties
1. Diffusion. Material spread is one specific example of diffusion. Here the state variable is the
concentration of the diffusing matrial. The flow quantity is its flux. The constitutive law is
Fickś law.
2. Heat spread. Here the state variable is the temprature, and the flow quantity is the heat
flux. The constitutive law is Fourierś law.
3. Stiff PDE, hence requires small time step, solved using implicit methods, not explicit for
stability.
4. Numerically, use Crank-Nicleson, in 2D, can use ADI.
5. Requires initial and boundary conditions to solve.
Examples of parabolic PDE’s
 5

1. Diffusion. ut − Duxx = 0 where D is the diffusion constant, must be positive quantity. For
heat PDE, D is the thermal diffusivity D = κ/cp ρ where κ is thermal conductivity, cp is
specific heat capacity, ρ is density of medium.
2. In higher spatial dimension ut − D∇2 u = 0
3. Foller-Plank, Black-Sholes PDEs
4. Diffusion-Reaction ut − Duxx = F (u(x, t)) where F (u(x, t)) is the reaction term, which can
be stiff or not. Examples
(a) Fischer equation, nonlinear PDE for modeling population growth. ut −Duxx = ru(x, t)(1−
u(x,t)
K ) where K is carrying capacity, and r is growth rate.

2.3 Hyperbolic
Some properties
1. Advection PDE (or Transport or convection?). ut + aux = 0, Transport or drift of conserved
substance (pollutant) in Fluid or Gas where a is speed of fluid. Analytic solution is u(x, t) =
f (x − at) where f (x) = u(x, 0) is the initial conditions.
2. The state variable is the concentration u of the contaminant, and the flow quantity is its
flux φ. The constitutive law is φ = cu.
1
3. Wave equation utt = c2 uxx . Analytic solution is u(x, t) = 2 [f (x − ct) + f (x + ct)] +
1
R x+ct
2c x−ct g(y) dy where f (x) = u(x, 0) and g(x) = ut (x, 0).
Examples
1. Advection, Wave (See above)
2. non-homogenouse advection and wave: ut + aux = f (x, t) and utt = c2 uxx + f (x, t).
3. Klein-Gordon utt = c2 uxx − bu
4. Telegraphy utt + kut = c2 uxx + bu

3 hints
reference
Characteristics are curves in the space of the independent variables along which the
governing PDE has only total differentials
 6

4 references
1. Elements of partial differential Equations, Pavel Drabek and Gabriela Holubova, 2007.
2. Applied partial differental equations. 4th edition, Richard Haberman
3. http://www.phy.ornl.gov/csep/pde/node3.html
4. http://www.me.metu.edu.tr/courses/me582/files/PDE_Introduction_by_Hoffman.pdf
5. http://en.wikibooks.org/wiki/Partial_Differential_Equations/Introduction_and_
Classifications
6. http://www.scholarpedia.org/article/Partial_differential_equation
7. http://how.gi.alaska.edu/ao/sim/chapters/chap3.pdf good discussion on classifica-
tion via Characteristics lines
8. http://gwu.geverstine.com/pde.pdf table on classification, diagram for discriminant sign