Well Balanced CU Scheme for

the Euler Equations

Submitted by

Subhash Meena
Under the supervision of
Prof. Harish Kumar

Mathematics and Computing

Indian Institute of Technology, Delhi
Acknowledgement
I would first like to thank my thesis advisor Prof. Harish Kumar of Maths
Department at IIT Delhi. The door to Prof. Harish’s office was always open
whenever I ran into a trouble spot or had a question about my project or
writing. He consistently allowed this paper to be my own work, but steered
me in the right the direction whenever he thought I needed it.

I would also like to acknowledge Miss Asha Meena of Maths Department

at IIT Delhi and I am gratefully indebted her for very valuable suggestions
during the project.

Finally, I must express my very profound gratitude to my Parents and

Sister for providing me with unfailing support and continuous encouragement
throughout my years of study. This accomplishment would not have been
possible without them. Thank you.
Contents
1 Introduction 3

2 Second-Order Semi-Discrete Central-Upwind Scheme 4

3 One dimensional Method 4

4 Well Balanced Reconstruction 5

5 Well-Balanced Central-Upwind Scheme 6

6 Well-Balanced Evolution 8

7 Implementation 9

8 Second-Order Semi-Discrete Central-Upwind Scheme (2 D) 17

9 Well-Balanced Central-Upwind Scheme 17

10 Well Balanced Reconstruction 18

11 Well Balanced Evolution 21

12 Implementation 22

13 Implementation for Ten Moment Gaussian Closure Equation 30

14 Conclusion 39

15 References 40

2
1 Introduction
In this project we aim to implement a second order well balanced central
upwind scheme for the Euler equation of gas dynamics with gravitation. The
Euler equations of gas dynamics with gravitation, which can be written in
the two-dimensional (2-D) case as:-

qt + F (q)x + G(q)y = S(q);

where
q = (ρ, ρu, ρv, E)t is a vector of conservative variables,
F (q)x = (ρu, ρu2 + p, ρuv, u(E + p))t
G(q)y = (ρv, ρuv, ρv2 + p, v(E + p))t are the fluxes in the x- and y-directions
S(q) = (0, −ρΦx, −ρΦy, −ρuΦx − ρuΦy)t is the source term.
ρ = density
u = velocity in x direction
v = velocity in y direction
E = total energy
p = pressure
Φ = time-independent linear gravitational potential (in y direction g, in x
direction 0)

• This is a system of balance laws, in many physical applications, solu-

tions of the system are small perturbations of the steady states.

• Capturing such solutions numerically is a challenging task since the size

of these perturbations may be smaller than the size of the truncation
error on a coarse grid.

• To overcome this difficulty, one can use very fine grid, but in many
physically relevant situations, this may be unaffordable.

• It is important to design a well-balanced numerical method, that is,

the method which is capable of exactly preserving some steady state
solutions. Then, perturbations of these solutions will be resolved on a
coarse grid in a non-oscillatory way

The steady-state solutions of above equations can be obtained by solving the

time-independent system G(q)y = S(q). To this end, the source term −ρg in
incorporated into the flux, introduce a new global variable w,

3
 Z
w = p + R, R(y, t) = g ρ()d

and rewrite the system G(q)y = S(q) as

(ρv)y = 0,
(ρv 2 + w)y = 0,
(v(E + p))y = −ρvg.
Simplest steady state is the motionless one, for which v = 0 and w = Con-
stant. The corresponding 2d state is u=v=0 and w = Constant.

2 Second-Order Semi-Discrete Central-Upwind

Scheme
Well balanced semi-discrete CU scheme for 2-D Euler equations with gravi-
tation can be written as :-
ρt + (ρu)x + (ρv)y = 0,
(ρu)t + (ρu2 + p)x + (ρuv)y = 0,
(ρv)t + (ρuv)x + (ρv 2 + w)y = 0,
Et + (u(E + p))x + (v(E + p))y = −ρvg
where
q = (ρ, ρu, ρv, E)t
F (q)x = (ρu, ρu2 + p, ρuv, u(E + p))t
G(q)y = (ρv, ρuv, ρv 2 + w, v(E + p))t
S(q) = (0, 0, 0, −ρvg)t
and Z
w = p + R, R(x, y, t) = g ρ(x, , t)d

3 One dimensional Method

The computational domain is partitioned into uniform Cartesian cells Ck :=
[yk− 1 , yk+ 1 ] of size |Ck | = ∆y centered at (yk ) = (k∆y), k = kL , . . . .., kR , It
2 2
is assumed that at time level t, the cell averages of the numerical solution
are available.
1 Z
q¯k (t) = q(y, t)dy
∆y Ck

4
4 Well Balanced Reconstruction
Implementation of semi-discrete CU scheme is the following system of ODEs

dq̄ Gk+ 1 − Gk− 1

=− 2 2
+ S¯k
dt ∆y

b+
k+ 1
N
G(qK ) − b− GS
k+ 1 q+1 S
Gk+ 1 = 2 2
+ βk+ 1 (qk+1 − qkN )
2 b+
k+ 1
− b−
k+ 1
2
2 2

b+ b−
k+ 1 k+ 1 2 2
βk+ 1 =
2 b+
k+ 1
− b−
k+ 1
2 2

are numerical fluxes and S¯k = (0, −g ρ¯k , −g ρv

¯ k )T are approximations of the
cell averages of the source term qkN and qkS are the one-sided point values of
the computed solution at cell interfaces y = yk+ 1 . To construct a second-
2
order scheme, these variables are to be calculated using the piece wise linear
reconstruction:-
X
q(y)
e = (q̄k + (qy )k (y − yk )).χCk (y)
k

where χCk (y) is the characteristic function of the interval Ck and

N ∆y
qK e k+ 1 − 0) = q¯k +
= q(y (qy )k
2 2

S ∆y
qK+1 e k+ 1 + 0) = q¯k +
= q(y (qy+1 )k
2 2

(qy ) are computed using a nonlinear limiter applied to the cell averages (q¯k ).
!
q̄k+1 − q̄k q̄k+1 − q̄k−1 q̄k − q̄k−1
(qy )k = minmod θ , ,θ
∆y 2∆y ∆y
where the minmod function is defined by

if zi > 0 ∀i then min(zi )

5
 minmod(z1 , z2 . . . ) = if zi < 0 ∀i then max(zi )
else 0

Finally, the one-sided local speeds of propagation b+

k+ 1
and b−
k+ 1
are estimated
2 2
∂G
using the smallest and largest eigenvalues of the Jacobian ∂q
.

b+
k+ 1
= max(vkN + cN S S
k , vk+1 + ck+1 , 0)
2

b−
k+ 21
= min(vkN − cN S S
k , vk+1 − ck+1 , 0)

ρv
where the velocity vkN and speed of sound cN
K are obtained using v = ρ and
c2 = γp
ρ
and the pressure is obtained using EOS.
The mentioned implementation would not be able to preserve steady state
solutions.

5 Well-Balanced Central-Upwind Scheme

In this method the source term –ρg in incorporated into the flux, introduce
a new global variable w,
Z
w = p + R, R(y, t) = g ρ()d

and rewrite the system G(q)y = S(q) as

(ρv)y = 0,

(ρv 2 + w)y = 0,
(v(E + p))y = −ρvg.

The main idea is to reconstruct equilibrium variables (ρ; ρv; w) rather than
(ρ; ρv; E). For the first two components above mentioned scheme is used to
obtain the same piecewise linear reconstructions as before, ρ̄(y) and ρv(y)
¯
N,S N,S
and compute the corresponding point values of ρ and v and then obtain
N,S ρv N,S
v = ρN,S To reconstruct the third equilibrium variable w, The point values
of R is computed by integrating the piecewise linear reconstruction of ρ,

6
 X
q(y)
e = (q̄k + (qy )k (y − yk )).χCk (y)
k

which results in piece wise quadratic approximation of R.

Z
R(y)
e =g ρ(, t)d
k−1
X X (ρy )k
=g [∆y (ρ̄i + ρ¯k (y − yk− 1 ) + (y − yk− 1 )(y − yk+ 1 ))].χck (y)
k i=kL
2 2 2 2

Then, the point values of R at the cell interfaces and cell centers are
i=k
X
Rk+ 1 = g∆y ρ̄i
2
i=kL

and
i=k
X ∆y ∆y 2
Rk = g∆y ρ̄i + g ρ¯k − g (ρy )k
i=kL 2 8

and the values of w at the cell centers are set as

wk = pk + Rk

then the minmod reconstruction procedure is applied to wk and obtain the

point values of w at the cell interfaces
∆y
wkN = wk + (wy )
2
S ∆y
wk+1 = wk+1 − (wy )k+1
2
Where
!
w̄k+1 − w¯k w̄k+1 − w̄k−1 w̄k − w̄k−1
(wy )k = minmod θ , ,θ
∆y 2∆y ∆y

Finally, the point values of p needed for computation of numerical fluxes are

pN N
k = wk − Rk+ 1 2

7
 pSk = wkS − Rk− 1
2

The value of R relates to each other and are computed recursively using the
relation
1
RkL − = 0,
2
Rk+ 1 = Rk− 1 + g∆y ρ¯k ,
2 2

∆y ∆y 2
Rk = Rk− 1 + g ρ¯k − g (ρy )k .
2 2 8
where k = kL , . . . .., kR

6 Well-Balanced Evolution
The cell-averages of q̄ are evolved in time according to the following system
of ODEs:
dq̄ Gk+ 1 − Gk− 1
=− 2 2
+ S¯k
dt ∆y
then, the second and third components of the numerical fluxes G are com-
puted the same way

−
b+
k+ 1
(ρN N 2 N S S 2 S
k (vk ) + wk ) − bk+ 1 (ρk+1 (vk+1 ) + wk+1 )
(2)
Gk+ 1 = 2 2
+βk+ 1 ((ρv)Sk+1 −(ρv)N
k )
2 b+
k+ 1
− b−
k+ 1
2
2 2

−
b+ v N (EkN + pN
k+ 1 k
S S S
k ) − bk+ 1 vk+1 (Ek+1 + pk+1 )
(3) S
Gk+ 1 = 2 2
+ βk+ 1 (Ek+1 − EkN )
2 b+
k+ 1
− b−
k+ 1
2
2 2

In order to preserve steady state the first component is modified by a smooth

function H accordingly

−
b+
k+ 1
(ρv)N S
k − bk+ 1 (ρv)k+1 |wk+1 − wk | ykR + 12 − ykL − 21
!
(1)
Gk+ 1 = 2 2
+βk+ 1 H · ((ρ)Sk+1 −(ρ)N
k )
2 b+
k+ 1
− b−
k+ 1
2 ∆y maxk {wk }
2 2

H is designed to be very small when the computed solution is locally

−wk
(almost) at steady state, that is, at the cell interfaces where wk+1
∆y
to be

8
very close to 1 elsewhere. This is done in order to guarantee the well-balanced
property of the scheme. On the other hand, the modification of the original
CU flux is quite minor since H is very close to 1 unless is very small.

7 Implementation
1. ShockTube Probelm :- We Solved the 1D System with g = 1 in the
computational domain [0,1] using the following initial data:-

(ρ(y, 0), v(y, 0), p(y, 0)) = (1, 0, 1) if y <= 0.5

(ρ(y, 0), v(y, 0), p(y, 0)) = (0.125, 0, 0.1) if y > 0.5
This boundary conditions are implemented using :-

ρkL −1 = ρkL , v kL −1 = −v kL , wkL −1 = wkL

ρkR +1 = ρkR , v kR +1 = −v kR , wkR +1 = wkR

Computed the solution for N=100, and plot the solution at T = 0.2 as
following :-

9
(Results for Well Balanced Scheme at T = 0.2)

10
 (Results for Non Well Balanced Scheme at T = 0.2)

In the implementation both the schemes are capable to capture the solu-
tions.

11
 2. Isothermal Equilibrium Solution :- In this implementation, the ability
of the proposed CU scheme to accurately capture small perturbations of the
steady state is tasted.

ρ(y) = e−y , v(y) = 0, p(y) = ge−y

Computational Domain is [0,1] and the boundary conditions are:-

ρkL −1 = ρkL , v kL −1 = v kL , wkL −1 = wkL

ρkR +1 = ρkR , v kR +1 = v kR , wkR +1 = wkR

now introduce a small initial pressure perturbation and consider the sys-
tem subject to following initial data:-
2
ρ(y, 0) = e−y , v(y, 0) = 0, p(y) = ge−y + τ e−100(y−0.5)

pressure perturbation constant is taken very small, first as 0.01 and then
0.0001 and obtained pressure perturbation is computed using N =200 and N
= 2000 as follows:-

12
(Results for Well Balanced Scheme at T = 0.25 and τ = 0.01)

13
(Results for Well Balanced Scheme at T = 0.25 and τ = 0.0001)

14
(Results for Non Well Balanced Scheme at T = 0.25 and τ = 0.01)

15
 (Results for Non Well Balanced Scheme at T = 0.25 and τ = 0.01)

Here, Well balanced schemes are able to capture the results in both the
cases for high perturbation constant as well as for lower perturbation constant
and in both the grid size n = 200 and n = 2000, where on the other side it
is difficult for non well balanced scheme to capture small perturbation when
n = 200.

16
8 Second-Order Semi-Discrete Central-Upwind
Scheme (2 D)
Well balanced semi-discrete CU scheme for 2-D Euler equations with gravi-
tation can be written as :-

ρt + (ρu)x + (ρv)y = 0,

(ρu)t + (ρu2 + p)x + (ρuv)y = 0,

(ρv)t + (ρuv)x + (ρv 2 + w)y = 0,

Et + (u(E + p))x + (v(E + p))y = −ρvg

where

q = (ρ, ρu, ρv, E)t

F (q)x = (ρu, ρu2 + p, ρuv, u(E + p))t
G(q)y = (ρv, ρuv, ρv 2 + w, v(E + p))t
S(q) = (0, 0, 0, −ρvg)t
and Z
w = p + R, R(x, y, t) = g ρ(x, , t)d

9 Well-Balanced Central-Upwind Scheme

The computational domain is partitioned into uniform Cartesian cells Cj , k :=
[xk− 1 , xk+ 1 ]∗[yk− 1 , yk+ 1 ] of size |Ck | = ∆x∆y centered at (xj , yk ) = (j∆x, k∆y), j =
2 2 2 2
jL , . . . .., jR ,, k = kL , . . . .., kR , It is assumed that at time level t, the cell av-
erages of the numerical solution are available.
1 1 Z Z
q¯k (t) = q(x, y, t)dxdy
∆x ∆y Ck

17
10 Well Balanced Reconstruction
In this scheme like in the case of 1-d, only the first three components of q
(ρ, ρu, ρv) are reconstructed :

qe(i) (x, y) = q̄ (i) (j, k) + (qx(i) )j,k (x − xj ) + (qy(i) )j,k (y − yk ),

where(x, y) belong to Cj,k i = 1, 2, 3,

and the point values at the cell interfaces are (xk+ 1 ,yk ) and (xj , yk+ 1 ) :
2 2

∆x (i)
(i)
(q (i) )E (i)
j,k = q (x(j+ 1 − 0, yk ) = q̄j.k +
g
(q )j.k
2 2 x
(i) ∆x (i)
(q (i) )W (i)
j,k = q (x(j− 12 + 0, yk ) = q̄j.k −
g
(q )j.k
2 x
(i) ∆y (i)
(q (i) )N (i)
j,k = q (xj , y(k+ 12 − 0) = q̄j.k +
g
(q )j.k
2 y
(i) ∆y (i)
(q (i) )Sj,k = qg
(i) (x , y
j (k− 1 + 0) = q̄j.k − (q )j.k
2 2 y
where the slopes (qx(i) )j,k and (qy(i) )j,k are computed using a nonlinear lim-
iter.

 
(i) (i) (i) (i) (i) (i)
q̄ − q̄j,k q̄j+1,k − q̄j−1,k q̄j,k − q̄j−1,k
(qx(i) )jk = minmod θ j+1,k , ,θ 
∆x 2∆x ∆x

 
(i) (i) (i) (i) (i) (i)
q̄ − q̄j,k q̄j,k+1 − q̄j,k−1 q̄j,k − q̄j,k−1
(qy(i) )jk = minmod θ j,k+1 , ,θ 
∆y 2∆y ∆y

where the minmod function is defined by

if zi > 0 ∀i then min(zi )

minmod(z1 , z2 . . . ) = if zi < 0 ∀i then max(zi )
else 0

18
 The one sided local speeds of propogation in the x and y direction respec-
tively are computed in further way :-

a+
j+ 1 ,k
= max(uE E W W
j,k + cj,k , uj+1,k + cj+1,k , 0)
2

a−
j+ 21 ,k
= min(uE E W W
j,k − uj,k , uj+1,k − cj+1,k , 0)

b+
j,k+ 1
N
= max(vj,k + cN S S
j,k , vj,k+1 + cj,k+1 , 0)
2

b−
j,k+ 12
N
= min(vj,k − cN S S
j,k , vj,k+1 − cj,k+1 , 0)

ρv
where the velocities uE W N S
j,k , uj+1,k , vj,k , vj,k+1 are obtained using v = ρ
and
c2 = γp
ρ

The calculation for E in x directions are done by computing first the

point value of pj,k at the cell centres using EOS and then p is computed at
cell interfaces using minmod limiter :-

∆x ∆x
pE j, k = pj,k + 2
(px )j,k , pW j, k = pj,k − 2
(px )j,k

where

p̄j+1,k − p̄j,k p̄j+1,k − p̄j−1,k p̄j,k − p̄j−1,k

 
(px )jk = minmod θ , ,θ
∆x 2∆x ∆x
Equipped with these values the cell cell interfaces value of E is computed
by using EOS :-

pE
j,k ρE
E
Ej,k = + j,k ((uE 2 E 2
j,k ) + (vj,k ) )
γ−1 2
pW
j,k ρW
W
Ej,k = + j,k ((uW 2 W 2
j,k ) + (vj,k ) )
γ−1 2

19
 for computing E in y direcction it is done as the same way as in 1-D case.
First compute R at the cell interfaces and centres:-

Rj,kL − 1 = 0;
2

1
Rj,k+ 1 = Rj,k − + g∆y ρ̄j,k ,
2 2
1 g∆y g(∆y)2
Rj,k = Rj, k − + ρ̄j,k − (ρy )j,k
2 2 8
and wj,k is computed as follow :-
wj,k = pj,k + Rj,k .
Next reconstructing w in y-direction yields
N ∆y
wj,k = wj,k + (wy )j,k ,
2
S ∆y
wj,k = wj,k − (wy )j,k ,
2
where

!
w̄j,k+1 − w̄j,k w̄j,k+1 − w̄j,k−1 w̄j,k − w̄j,k−1
(wy )j,k = minmod θ , ,θ
∆y 2∆y ∆y
Finally the obtained point values are of w are used to evaluate the corre-
sponding point values of p by:
pN N
j,k = wj,k − Rj,k+ 1 2

pSj,k = wj,k
S
− Rj,k− 1
2

and E from equation of states :-

pN
j,k ρN
N
Ej,k = + j,k ((uN 2 N 2
j,k ) + (vj,k ) )
γ−1 2
pSj,k ρS
S
Ej,k = + j,k ((uSj,k )2 + (vj,k
S 2
))
γ−1 2

20
11 Well Balanced Evolution
The cell average of q̄ are evolved with the time according to the following
System of ODEs:-

dq̄ Fj+ 1 ,k − Fj− 1 , k Gj,k+ 1 − Gj,k− 1

=− 2 2
+− 2 2
+ S¯j,k
dt ∆x ∆y

Here F and G are numeric flux Introducing the notations :-

a+ a−
k+ 1 k+ 1 2 2
αk+ 1 =
2 a+
k+ 1
− a−
k+ 1
2 2

b+ b−
k+ 12 k+ 12
βk+ 1 =
2 b+
k+ 21
− b−k+ 12

and the component of F and G are :-

−
a+
j+ 1 ,k
(ρu)E W
j,k − aj+ 1 ,k (ρu)j+1,k |wj+1,k − wj,k | xkR + 21 − xkL − 12
!
(1)
Fj+ 1 ,k = 2 2
+αj+ 1 ,k H · (ρW E
j+1,k −ρj,k )
2 a+
j+ 12 ,k
− a−
j+ 12 ,k
2 ∆x maxj,k {wj,k }

−
a+ (ρE (uE )2 + pE
j+ 1 ,k j,k j,k
W W 2 W
j,k ) − aj+ 1 ,k (ρj+1,k (uj+1,k ) + pj+1,k )
(2)
Fj+ 1 ,k = 2 2
+αj+ 1 ,k ((ρu)W E
j+1,k −(ρu)j,k )
2 a+
j+ 1 ,k
− a−
j+ 1 ,k
2
2 2

a+ (ρE uE v E ) − a−
j+ 1 ,k j,k j,k j,k
(ρW uW v W )
j+ 1 ,k j+1,k j+1,k j+1,k
(3)
Fj+ 1 = 2 2
+αj+ 1 ,k ((ρv)W E
j+1,k −(ρv)j,k )
2 a+
j+ 1 ,k
− a−
j+ 1 ,k
2
2 2

−
a+ uE (Ej,k
j+ 1 ,k j,k
E
+ pE W W W
j,k ) − aj+ 1 ,k uj+1,k (Ej+1,k + pj+1,k )
(4) W E
Fj+ 1 ,k = 2 2
+αj+ 1 ,k (Ej+1,k −Ej,k )
2 a+
j+ 1 ,k
− a−
j+ 1 ,k
2
2 2

21
 −
b+
j,k+ 1
(ρv)N S
j,k − bj,k+ 1 (ρv)j,k+1 |wj,k+1 − wj,k | ykR + 12 − ykL − 12
!
(1)
Gj,k+ 1 = 2 2
+βj,k+ 1 H · ((ρ)Sj,k+1 −(ρ)N
j,k )
2 b+
j,k+ 21
− b−
j,k+ 12
2 ∆y maxj,k {wj,k }

−
b+
j,k+ 1
(ρN N N S S S
j,k uj,k vj,k ) − bj,k+ 1 (ρj,k+1 uj,k+1 vj,k+1 )
(2)
Gj,k+ 1 = 2 2
+βj,k+ 1 ((ρu)Sj,k+1 −(ρu)N
j,k )
2 b+
j,k+ 12
− b−
j,k+ 12
2

−
b+
j,k+ 1
(ρN N 2 N S S 2 S
j,k (vj,k ) + wj,k ) − bj,k+ 1 (ρj,k+1 (vj,k+1 ) + wj,k+1 )
(3)
Gj,k+ 1 = 2 2
+βj,k+ 1 ((ρv)Sj,k+1 −(ρv)N
j,k )
2 b+
j,k+ 12
− b−
j,k+ 12
2

−
b+ v N (Ej,k
j,k+ 1 j,k
N
+ pN S S S
j,k ) − bj,k+ 1 vj,k+1 (Ej.k+1 + pj,k+1 )
(4) S N
Gj,k+ 1 = 2 2
+βj,k+ 1 (Ej,k+1 −Ej,k )
2 b+
j,k+ 1
− b−
j,k+ 1
2
2 2

12 Implementation
1. Isothermal Equilibrium Solution : In this implementation we solved the
2-D system with subjected to initial conditions:-
ρ0 gy
−
ρ(x, y, 0) = ρ0 e p0

ρ0 gy
−
p(x, y, 0) = p0 e p0

u(x, y, 0) = v(x, y, 0) = 0
where ρ0 = 1.21 and p0 = 1 and the solid boundary conditions imposed
at the edges of unit square [0 , 1]*[0 , 1].
A smalll perturbation to the initial pressure is being added
−p0 gy −100p0 g
( ((x−0.3)2 +(y−0.3)2 )
p(x, y, 0) = p0 e p0
+ ηe p0

The results are following for the different cases :-

22
(Results for Well Balanced Scheme at T = 1.5 and n = 200*200)

(Results for Well Balanced Scheme at T = 1.5 and n = 200*200)

23
(Results for Well Balanced Scheme at T = 1.5 and n = 100*100)

(Results for Well Balanced Scheme at T = 1.5 and n = 100*100)

24
(Results for Well Balanced Scheme at T = 1.5 and n = 50*50)

(Result for Well Balanced Scheme at T = 1.5 and n = 50*50)

25
(Results for Non Well Balanced Scheme at T = 1.5 and n = 100*100)

(Results for Non Well Balanced Scheme at T = 1.5 and n = 100*100)

Thus We observe that well balanced schemes are capable to capture the
small perturbations more accurately than non Well balanced scheme.

26
 2. Explosion :- In the second 2-D example, performance of well balanced
scheme and non well balance scheme is compared in an explosion setting and
demonstrated non well balanced shock wave generated by non well balanced
scheme with g = 0.118 and in the computational domain [0,3]*[0,3] subject
to the following initial data:-

ρ(x, y, 0) = 1
u(x, y, 0) = 0
p(x, y, 0) = 1 − gy + 0.005, given(x − 1.5)2 + (y − 1.5)2 < 0.01
p(x, y, 0) = 0otherwise
at different times expansion of gas due to heat generation in captured in
this manner :-

27
(Results for Well Balanced)

(Results for Non Well Balanced)

28
(Results for Well Balanced)

(Results for Non Well Balanced)

29
 For Isothermal Equilibrium problem Well Balanced Scheme can capture
the small perturbation much accurately in all the cases implemented where
Non Well Scheme tends to get the small perturbation and becomes better
in case of using 200 cells but still Well Balanced is more accurate. and for
the explosion example also Well balanced scheme accurately captures the
behaviour of expansion of gas while Non well balanced scheme shifted from
results.

13 Implementation for Ten Moment Gaus-

sian Closure Equation
1. Contact Discontinuity Test :- This test is generalized from gravitational
field problem. We consider the domain [0,1] with initial discontinuity at
x=0.5 and Initial conditions for Contact discontinuity test are :-

We consider the the potential W(x) = x with solid wall boundary condi-
tions. In the observations we get that well balanced scheme as well as non
well balanced scheme is able to capture the results. :-

30
 (Density)

(Velocity)

31
(Pressure)

(Temperature)

32
 2. Small Perturbation Test :- Well balanced scheme advantage is that it
is able to capture the solution close to stationary solution with significant
accuracy. The initial profile for this implementation is,
x
− 2RT
ρ(x) = ρ0 e 0 ,
v1 = v2 = 0

−x 100(x−0.5)2
− 2RT −
p11 = ρ0 RT0 e 0 + e 2RT0
, p12 = 0, p22 = 1
,
where p0 = 1 , R = 1, T0 = 1 and  = 10−5
The numeric results at the final time t = 0.25 are presented using 100,200,400,800,1000
cells. Well balanced scheme is able to capture the solutions accurately in this
test.

(Small Perturbation Test)

3. Small Perturbation Test with Discontinuous potential :- To observe

the impact of discontinuous potential on the solution close to steady state,
we consider a steady state solution with small perturbation. In particular,

33
we consider a small perturbation of stationary isothermal initial data in p1 1
component ie:-
W (x)
− 2RT
ρ(x) = ρ0 e 0 , v1 = v2 = 0

W (x) 100(x−0.5)2
− 2RT −
p11 = ρ0 RT0 e 0 + e 2RT0
, p12 = 0, p22 = 1
,
with discontinuous potential,
W(x) = 2 when 0.4 < x < 0.6 , 1 otherwise.
with p0 = 1, R = 1, T0 = 1 . The perturbation parameter has been
taken smaller for different comparisons. The numerical results are present
for 100,500,1000 cells with the final time t = 0.25  = 10−5

(Well Balanced Scheme)

34
 (Non Well Balanced Scheme)

4. Sod ShockTube Test :- In the Sod ShockTube case the domain is con-
sidered to be [0,1] with solid boundary conditions and the initial discontinuity
conditions are as following :-

W(x) = x for ensuring non-vanishing source term. The solutions are

presented at t = 0.125. The numerical solutions are computed using 100
cells and 2000 cells.
We observe that velocity v1 takes the negative value and p,p11 and p22
increasing near x = 0.

35
(Density)

(Pressure p11 )

36
(Pressure p12 )

(Pressure p22 )

37
(Velocity in x direction)

(Velocity in y direction)

38
14 Conclusion
In this work we have scrutinized various implementations for well balanced
scheme as well as non well balanced schemes for Euler equations and Ten Mo-
ment Gaussian Closure and compared their significant results. We observed
that Well balanced schemes were able to capture solutions more accurately
preserving steady state conditions while Non well balanced schemes were
leading to oscillations and were not able to capture the small added uncer-
tainties

39
15 References
• Asha Meena, Harish Kumar, A Well-balanced Scheme for Ten-Moment
Gaussian Closure Equations with Source term

• A. Kurganov, S. Noelle and G. Petrova: Semi-discrete central-upwind

scheme for hyperbolic conservation laws and Hamilton-Jacobi equa-
tions, 2001

• J. M. Gallardo, C. Parfes, and M. Castro, On a well-balanced high-

order finite volume scheme for shallow water equations with topography
and dry areas, Journal of Computational Physics, (2007)

• A. Kurganov, G. Petrova and B. Popov: Second-order semi-discrete

central-upwind schemes for hyperbolic conservation laws with non-
convex fluxes or equations of state.

• A. Kurganov and E. Tadmor: New high-resolution central schemes for

nonlinear conservation laws and convection-diffusion equations. (2003)

• B. Wang, H. Glaz: Second order Godunov-like schemes for gas dynam-

ics with a non-convex equation of state, 1999

40