Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/364330648

Chapter 2: Numerical Investigation of Wave Pattern Evolution in Gray- Scott


Model Using Discontinuous Galerkin Finite Element Method

Chapter · February 2023


DOI: 10.1201/9781003367420-2

CITATIONS READS

0 44

1 author:

Satyvir Singh
RWTH Aachen University
50 PUBLICATIONS   335 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

High-Fidelity Computational Modeling of Stiff Multifluids Flows problems View project

Discontinuous Galerkin approach to the Boltzmann-Maxwell problem for ultrafast thermalisation/transport View project

All content following this page was uploaded by Satyvir Singh on 14 October 2022.

The user has requested enhancement of the downloaded file.


Chapter 2: Numerical Investigation of Wave Pattern Evolution in Gray-
Scott Model Using Discontinuous Galerkin Finite Element Method
Author:
*Satyvir Singh (School of Physical and Mathematical Sciences, Nanyang Technological
University, 21 Nanyang Link, Singapore - 637371, satyvir.singh@ntu.edu.sg,
https://orcid.org/0000-0001-6669-5296)

2.1 Introduction

In development biology, the evolution of pattern formation is a central but still unresolved

challenging research topic. Many biological and biomedical concerns include evolution and

form changes, which are the consequence of many nonlinear interactions (Harrison, 1993;

Meinhardt, 1982; Murray, 1989). Therefore, mathematical modeling and numerical computing

play a critical role in comprehending and forecasting the outcomes of such complicated

interactions. The reaction and diffusion process of biochemical components generate a wide

range of models that can be observed in nature: hence the coupled reaction-diffusion system

emerges.

According to Turing (1952), under some circumstances, a chemical reaction with diffusion

may generate complex spatial pattern formation of chemical concentration. There are many

prominent examples of coupled reaction-diffusion system, including the Gray-Scott model,

which plays an important role in developed biology. This model was proposed by Gray and

Scott (1983) as a substitution for the autocatalytic model of glycolysis initiated by Sel’kov

(1984). Pearson (1993) made a significant contribution to the study of spot pattern creation in

the two-dimensional Gray-Scott model by including the importance of space by removing the

restriction of a well-stirred tank. In his study, several complex spot type patterns were

illustrated for a Gray-Scott model via numerical experiments. Mazin et al. (1996) numerical

investigated a class of non-equilibrium phenomena in the bistable Gray-Scott model, including

stable localized structures, mixed Turing-Hopf modes, interacting fronts, global Turing

structures, and spatiotemporal chaos. Reynolds et al. (1994) investigated the one-dimensional
time-dependent wave replication which is filled in the domain with a periodic array of spots.

Doelman et al. (1997) studied numerically the one-dimensional Gray-Scott model and observed

various wave patterns, including self-replicating, stationary, and traveling waves. Tok et al.

(2019) investigated the wave pattern formation of the Gray-Scott reaction-diffusion model in

one-dimensional space using trigonometric quadratic B-spline functions.

In this work, our aim is to explore numerically the wave pattern evolution in the one-

dimensional Gray-Scott reaction-diffusion model using a high-order discontinuous Galerkin

finite element method (DG-FEM). This method is considered a hybrid approach of Finite

Element and Finite Volume methods. The DG-FEM method is rapidly being used in the last

decades as a computational tool for solving nonlinear partial differential equations that arise in

a broad spectrum of scientific and engineering problems (Singh, 2018; Singh, 2021; Singh,

2021a; Singh, 2021b; Singh & Battiato, 2021; Singh et al., 2022; Singh, 2022).

This work provides a one-dimensional mixed modal DG-FEM scheme for solving the Gray-

Scott model, which differs from earlier studies. The third order scaled Legendre basis functions

are adopted for DG spatial discretization, while a third-order TVD Runge-Kutta scheme is

employed as a temporal discretization. The remaining of this chapter is planned as follows. The

mathematical model for Gray-Scott reaction-diffusion equations is presented in Section 2.2. A

detailed description of the DG spatial discretization in one-dimensional space is illustrated in

Section 2.3. We apply the proposed numerical method to some test problems of the Gray-Scott

model, and their wave patterns evolution are discussed in Section 2.4. Finally, the concluding

remarks are discussed in Section 2.5.

2.2 Gray-Scott Reaction-Diffusion Model

The Gray-Scott reaction-diffusion system models the various spatiotemporal patterns,

including, self-replicating and pulse-splitting waves, spots, etc., appearing in nature. The

irreversible Gray-Scott model governs the chemical reactions (Gray and Scott; 1983)
W1  2W2  P, (2.1)
W2  P

in a gel reactor, where W2 catalyzes its own reaction with W1 and P is an inert product. In order

to study the wave pattern evolution mathematically, we analyze the irreversible Gray-Scott

reaction-diffusion model in one-dimensional space on the infinite line written as

w1 2w 2w (2.2)


 1 21  w1w22  F 1  w1   1 21  f  w1 , w2  ,
t z z
w2  w2
2
 2 w2
 2  w w 2
  F  k  2 2 2  g  w1 , w2  ,
w  
t z 2 z
1 2

where w1  z , t  and w2  z , t  denote the concentrations of two chemical species W1 and W2 ,

respectively. The constant F indicates the rate at which W1 is fed from the reservoir into the

reactor, and W2 is the overall rate of decay of W2 . To guarantee that the localized pulses can

propagate, the choice of F and k is small and 0   2  1. The Gray-Scott model (2.2) can be

written in a generalized form as

W (2.3)
 D 2 W  R  W  ,
t

with

w   0  f  w1 , w2   (2.4)
W   1, D   1  , R  W   .
 w2  0 2   g  w1 , w2  

2
Here, D is the diffusion constant matrix and  2  is the Laplacian operator. R  W 
z 2

denotes the reaction matrix of nonlinear reaction kinetics f and g.

2.3 Mixed modal DG-FEM scheme

To simulate the Gray-Scott reaction-diffusion model in one-dimensional space, a mixed modal

DG-FEM scheme is employed in this section. In this method, an auxiliary variable S is added

in this formulation to address the higher-order derivatives which are considered the derivative
of the solution variable (Singh, 2018). Thus, the Gray-Scott model (2.3) is re-written as a

coupled system of W and S for the mixed DG-FEM construction,

S  W  0, (2.5)
W
 F  S   R  W  ,
t

where F  S    DW. To formulate the mixed modal DG-FEM scheme, the following

Sobolev space

 
Vhl   h  L2    : h |I n  P k  I n  , (2.6)

where I n   zn 1/2 , zn 1/2  , n  N is the local element,  =  z L , z R  is the computational domain,

h is the one component for the vector-valued  h , P k  I n  is the space consisting of

discontinuous Galerkin polynomial functions of degree at most k. Now the exact solution of

W and S are approximated by the DG-FEM polynomial approximation of Wh Vhl and

S h Vhl , respectively as

Nk (2.7)
S h  z, t    Shi  t  bi  z  ,
i 0
Nk
Wh  z, t    Z hi  t  bi  z  ,
i 0

where Z hi and S hi denote the modal coefficients of W and S , and bi  z  represents the basis

function of degrees k. Here, the orthogonal scaled Legendre polynomials for bi  z  are adopted

in the computational element I n   1,1.

2i  n ! 0,0 (2.8)
2

bn    P   , n  0,
 2n  !
2  z  zi 
 ,  1    1,
z

where P 0,0   is the Legendre polynomial function and zi is the elemental center, respectively.
For the numerical simulations, the first four polynomial functions, which correspond to the

third-order DG-FEM approximation, are used in this work. These basis functions are defined

globally, implying that they are used by all local elements. The DG discretization of the coupled

system of Gray-Scot model (2.5) can be obtained by replacing the exact solutions with the

corresponding approximations defined in Eq. (2.7) and multiplying by a polynomial function

bh  z  and then integrated by parts over the local element I n . Taking bh Vhl , Wh Vhl , and

S h Vhl , we obtain

S
e
h bh dV   bh  S h dV 
e
bS
e
h h  n d   0, (2.9)


Wh bh dV   bh  F  S h  dV   bh F  S h   n d    bh R  Wh  dV .
t e e e e

Figure 2.1: Wave pattern evolution of the Gray-Scott model for Problem 1 with the
parameters 1  1, 2  0.01, F  0.01, k  0.053 at different time intervals.

Here n denotes the outward unit normal vector;  and V are the boundary and volume of the

element e , respectively. The interface fluxes are not uniquely defined because of the
discontinuity in the solution Wh and S h at the elemental interfaces. The functions F  S h   n

and Wh  n emerging in Eq. (2.9) can be substituted by the numerical fluxes at the elemental

interfaces denoted by H BR1 and H aux , respectively. Here, the central flux or Bassi-Rebay

(BR1) scheme is employed for the viscous and auxiliary numerical fluxes to calculate the flux

at elemental interfaces.

H aux  Whint , Whext  


1 (2.10)
 Whint  Whext  ,
2
H BR1  Sint
h , Sh  
F  Sint
h   F S h  .
1
ext ext

2 

Here, the superscripts  int  and  ext  deliver the interior and exterior states of the elemental

interface. As a result, the DG-FEM weak formulation is obtained as

S
e
h bh dV   bh  S h dV 
e
b
e
h H aux d   0, (2.11)


Wh bh dV   bh  F  S h  dV   bh H BR1 d    bh R  Wh  dV .
t e e e e

In the above-mentioned expression, the emerging volume and surface integrals are

approximated by the Gaussian-Legendre quadrature rule within the elements to ensure the high

order accuracy. Finally, the DG-FEM spatial discretization (2.11) may be expressed in semi-

discrete ODEs form as

dWh (2.12)
M  L  Wh  ,
dt

where M and L  Wh  are the orthogonal mass matrix, and the residual function, respectively.

Here, an explicit form of Strongly Stability Preserving Runge-Kutta method with third-order

accuracy (Shu and Osher, 1988) is adopted as the time marching scheme.
Wh(1)  Whn  t M 1L  Wh  , (2.13)

Wh  Wh  t M 1L  Wh(1)  ,
3 n 1 (1) 1
Wh(2) 
4 4 4
1 2 2

Whn 1  Whn  Wh(2)  t M 1L Wh  ,
3 3 3
2

 
where L Whn is the residual approximation at time tn , and t is the suitable time-step value.

Figure 2.2: Space-time plots of concentrations w1 and w2 in the Gray-Scott model for
Problem 1 with the parameters 1  1, 2  0.01, F  0.01 with different k values
 k  0.047, 0.053, 0.060  .

2.4 Results and discussion

In present section, we present the wave pattern evolution of the Gray-Scott model, which can

be easily seen in real life, such as butterfly wings, damping, gastrulation, embryos, multiple

spots, Turing patterns, and many more. For this purpose, three test problems of Gray-Scott

model are examined by simulating the numerical experiments with the third-order mixed modal

DG-FEM scheme. All the following simulations are done with grid points of 301, and time step
of z  0.01.

Figure 2.3: Wave pattern evolution of the Gray-Scott model for Problem 2 with parameters
1  2 105 , 2  1105 , F  0.023, k  0.050 at different time intervals.

Problem 1. As the first test problem of the Gray-Scott model, the following initial

conditions are considered:

w1  z, 0   1  0.5sin100  z / L  , (2.14)
w2  z, 0   0.25sin100  z / L  .

in the limited interval [0, L] which is considered long enough to ensure that the wave dynamics

of the model are not influenced by boundaries. The homogenous Neumann conditions are

chosen for the boundary conditions. The simulation parameters are taken as

1  1, 2  0.01, F  0.01, and k  0.047, 0.053, 0.060. (2.15)

Figure 2.1 illustrates the wave pattern evolution of the Gray-Scott model for the parameters

1  1, 2  0.01, F  0.01, k  0.053. From the wave profiles, it can be observed that the two

solitary pulses generated from the beginning circumstance was broke into further four pulses
and moved apart as time progressed until the equilibrium state was obtained. As the k value

decreases, there are fewer peaks over the domain. The parameters F and k are frequently used

to assess how many peaks a domain can support. We can observe more peaks if we expand the

domain to [0,400]. The initial conditions of the simulations differ from those in Figure 2.1. We

run the simulation of the Gray-Scott model till time t =15,000 to see the traveling pulses' self-

replications on a global scale. Figure 2.2 shows the space-time plots of the concentrations

w1 , and w2 in the Gray-Scott model for different k values. For k = 0.047, 0.053, and 0.060, the

pulses can grow into 14, 9, and 2 peaks, respectively.

Figure 2.4: Space-time plots of concentrations w1 and w2 in the Gray-Scott model for
Problem 2 with parameters 1  2 105 , 2  1105 , F  0.023 with different k values
 k  0.047, 0.050, 0.053 .

Problem 2. For the second test problem of the Gray-Scott model, the following initial

conditions are considered with the homogenous boundary conditions:


0.5, 1.2  z  1.3 (2.16)
w1  z, 0   
1, else.
0.25, 1.2  z  1.3
w2  z , 0   
0, else.

in the bounded interval [0, 2.5]. The simulation parameters are taken as

1  2 105 , 2  1105 , F  0.023, and k  0.047, 0.050, 0.053. (2.17)

Figure 2.3 shows the wave pattern evolution of the Gray-Scott model with the parameters

1  2 105 , 2  1105 , F  0.023, k  0.050. New pulses are produced on the trailing edges

of previous pulses, and growth-wave patterns and self-splitting pulses have been observed.

Figure 2.4 displays the space-time plots of the considered concentrations in the Gray-Scott

model for the different k values  k  0.047, 0.050, 0.053 . When the simulations are repeated

for a longer time, it is clear that no pattern emerges for the smaller k value. Interestingly, only

two stationary pulses can be seen in periodic patterns for larger k value.

Problem 3. Finally, the last test problem of the Gray-Scott model is considered with the

following initial conditions in the bounded interval [0, 2.5]:

0.5, 0.6  z  0.7 (2.18)



w1  z , 0   0.25, 1.8  z  1.9
1,
 else
0.25, 0.6  z  0.7

w2  z , 0   0.75, 1.8  z  1.9
0,
 else.

The simulations are run using Neumann boundary conditions that are homogeneous, and the

parameters are set similarly to Problem 2. The wave pattern evolutions of the Gray-Scott model

for the parameters 1  2 105 , 2  1105 , F  0.023, k  0.050 are displayed in Figure 2.5.

At longer time, both initial waves are merged and make a more complex pattern. Simulations

are run till t =5000, and the corresponding space-time plots for different k values of the model
is shown in Figure 2.6.

Figure 2.5: Wave pattern evolution of the Gray-Scott model for Problem 3 with parameters
1  2 105 , 2  1105 , F  0.023, k  0.050 at different time intervals.

2.5. Concluding remarks

This study focuses on the numerical investigation of wave pattern evolution in the Gray-Scott

reaction-diffusion model. For this purpose, a mixed modal discontinuous Galerkin finite

element method in one-dimensional space is employed to simulate the Gray-Scott model. In

this numerical method, a mixed formulation is introduced by adding an extra variable to handle

the high-order derivative occurred in diffusion term. For spatial discretization, a hierarchal

modal basis function based on scaled Legendre polynomials is utilized, whereas for temporal

discretization, an explicit third order Strongly Stability Preserving Runge-Kutta approach is

used. Three different test problem with different initial conditions are adopted to illustrate the

wave pattern evolution in the Gray-Scott model. Numerical results reveal that different patterns

appear when a modest value in the parameters is changed. The proposed numerical technique,
on the other hand, reveals that it is an effective strategy for finding numerical solutions in a

wide range of linear and nonlinear physical models. In the future, this method could be used to

efficiently solve higher-dimensional reaction-diffusion models.

Figure 2.6: Space-time plots of concentrations w1 and w2 in the Gray-Scott model for
Problem 3 with parameters 1  2 105 , 2  1105 , F  0.023 with different k values
 k  0.047, 0.050, 0.053 .

Acknowledgements

The author gratefully acknowledges the financial support provided by the Nanyang

Technological University, Singapore through the NAP-SUG grant program.


References

Harrison, L.G. (1993). Kinetic Theory of Living Pattern. Cambridge University Press.

Meinhardt, H. (1982). Models of Biological Pattern Formation. Academic Press.

Murray, J.D. (1989). Mathematical Biology. Springer-Verlag.

Turing, A.M. (1952). The chemical basis of morphogenesis. Philosophical Transactions of the

Royal Society B, 237, 37-72.

Gray, P. & Scott, S.K. (1983). Autocatalytic reactions in the isothermal, continuous stirred tank

reactor: isolas and other forms of multistability. Chemical Engineering Science, 38(1),

29-43.

Sel'Kov, E.E. (1984). Self-Oscillations in Glycolysis 1. A Simple Kinetic Model. European

Journal of Biochemistry, 4(1), 79-86.

Pearson, J.E. (1993). Complex patterns in a simple system. Science, 261(5118), 189-192.

Mazin, W., Rasmussen, K.E., Mosekilde, E., Borckmans, P., & Dewel, G. (1996). Pattern

formation in the bistable Gray-Scott model. Mathematics and Computers in Simulation,

40(3-4), 371-396.

Reynolds, W.N., Pearson, J.E., & Ponce, D.S. (1994). Dynamics of self-replicating patterns in

reaction-diffusion systems. Physical review letters, 72(17), 2797.

Doelman, A., Kaper, T.J., & Zegeling, P.A. (1997). Pattern formation in the one-dimensional

Gray-Scott model. Nonlinearity, 10(2), 523.

Tok-Onarcan, A., Adar, N., & Dag, I. (2019). Wave simulations of Gray-Scott reaction-

diffusion system. Mathematical Methods in the Applied Sciences, 42(16), 5566-5581.

Singh, S. (2018). Development of a 3D discontinuous Galerkin method for the second-order

Boltzmann-Curtiss-based hydrodynamic models of diatomic and polyatomic gases.

(Doctoral dissertation, Gyeongsang National University South Korea. Department of

Mechanical and Aerospace Engineering)


Singh, S. (2021). Numerical investigation of thermal non-equilibrium effects of diatomic and

polyatomic gases on the shock-accelerated square light bubble using a mixed-type modal

discontinuous Galerkin method. International Journal of Heat and Mass Transfer, 169,

121708.

Singh, S. (2021a). Mixed-type discontinuous Galerkin approach for solving the generalized

FitzHugh-Nagumo reaction-diffusion model. International Journal of Applied and

Computational Mathematics,7, 207.

Singh, S. (2021b). A mixed-type modal discontinuous Galerkin approach for solving nonlinear

reaction-diffusion equations. AIP Conference Proceedings (accepted).

Singh, S., & Battiato, M. (2021c). An explicit modal discontinuous Galerkin method for

Boltzmann transport equation under electronic nonequilibrium conditions. Computers &

Fluids, 224, 104972.

Singh, S., Karchani, A., Chourushi, T., & Myong, R.S. (2022). A three-dimensional modal

discontinuous Galerkin method for second-order Boltzmann-Curtiss constitutive models

of rarefied and microscale gas flows. Journal of Computational Physics,457, 111052.

Singh, S. (2022b). Computational Modeling of Nonlinear Reaction-Diffusion Fisher-KPP

Equation with Mixed Modal Discontinuous Galerkin Scheme. Mathematical Modeling

for Intelligent Systems, CRC Press.

View publication stats

You might also like