Int. J. Appl. Comput.

Math (2021) 7:207

https://doi.org/10.1007/s40819-021-01153-9

ORIGINAL PAPER

Mixed-Type Discontinuous Galerkin Approach for Solving the

Generalized FitzHugh–Nagumo Reaction–Diffusion Model

Satyvir Singh1

Accepted: 16 September 2021

© The Author(s), under exclusive licence to Springer Nature India Private Limited 2021

Abstract
The generalized FitzHugh–Nagumo reaction–diffusion model has long been fascinating
topic in the field of the mathematical and physics. This paper is aimed to present a mixed-
type discontinuous Galerkin approach to solve the generalized FitzHugh–Nagumo reaction–
diffusion model. An auxiliary variable is introduced in the governing equation for handling
the higher-order term. The scaled Legendre polynomial functions are used for the spatial
discretization. The numerical technique transforms the problem into a system of semi-discrete
ordinary differential equation which is solved by an explicit three-stages, third-order SSP
Runge-Kutta scheme. For verifying the accuracy and efficiency, the numerical scheme is
tested on some different problems of generalized FitzHugh–Nagumo model with constant
and time-dependent coefficients. The obtained results are very close to the exact solutions
and better than those obtained by some other numerical schemes. This scheme plays as an
alternative option for solving the nonlinear reaction–diffusion type equations.

Keywords Mixed-type discontinuous Galerkin · Reaction–diffusion · Generalized

Fitzhugh–Nagumo · Explicit SSP-RK33 scheme

Introduction

The study of reaction–diffusion models which occur frequently in inhomogeneous media

associated with diffusion, dispersion, and dissipation phenomena has emerged as an inter-
esting topic from scientific and engineering applications. Typical examples include biology
[1], Turing patterns [2,3], nuclear reactors [4], chemical engineering [5], ecological invasions
[6], epidemics propagation [7] and tumour growth [8].
In 1952, Hodgkin and Huxley [9] proposed a reaction–diffusion mathematical model,
known as the Hodgkin–Huxley model based on the squid giant axon which describes the
excitation-propagation of waves in nerve cell membranes [10–13]. Later, FitzHugh [14]
and Nagumo et al. [15] derived the famous classical FitzHugh–Nagumo model based on the
Hodgkin–Huxley model. This model simplifies the knottiness of the Hodgkin–Huxley model

B Satyvir Singh
satyvir.singh@ntu.edu.sg
1 School of Physical and Mathematical Sciences, Nanyang Technological University, 21 Nanyang Link,
Singapore 637371, Singapore

0123456789().: V,-vol 123

Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 2 of 16 Int. J. Appl. Comput. Math (2021) 7:207

with preserving various excitation–propagation characteristics. In recent years, this model

has been paid much attention by many mathematicians and physicists because of its wide
variety in various applications [16–19].
In the recent past, a number of earlier works were carried out analytically and numerically
for solving the classical FitzHugh–Nagumo reaction–diffusion equation. The analytic solu-
tions through the Hirota approach in the classical FitzHugh–Nagumo equation were studied
to investigate the collisions of two travelling wave-fronts in [20]. The stability of travelling
solutions was investigated for the FitzHugh–Nagumo equation with eigenvalue technique in
[21]. The analytical solutions for the FitzHugh–Nagumo model was derived with the aid of
the non-classical method combined with Jacobbi-elliptic polynomials in [22]. The existence
of delay and explicit duck solutions were carried out for the classical FitzHugh–Nagumo
equation in [23]. The computational study of a perturbed FitzHugh–Nagumo equation was
investigated through the membrane cell in [24].
A series of analytic solutions for this reaction–diffusion model was investigated with the
first integral method in [19]. The soliton solutions for the classical FitzHugh–Nagumo equa-
tion were studied with a homotopy analysis approach in [16]. The rotating wave solutions for
the classical FitzHugh–Nagumo equation had been successfully examined for describing the
nerve cell behavior in [25]. Some different schemes such as Haar wavelet and pseudospectral,
were implemented for solving the classical FitzHugh–Nagumo equation in [26,27], respec-
tively. Later, the exact solutions for Nagumo telegraph as well as Nagumo reaction–diffusion
models were obtained by using the variational approach in [28]. Recently, a few works were
carried out to investigate the numerical solutions of the generalized Fitzhugh–Nagumo equa-
tion with time-dependent coefficients by various numerical approximation techniques, the
tanh method [29], the Jacobi–Gauss–Lobatto collocation method, [30] and the polynomial
differential quadrature method [31].
In this paper, a mixed-type approximation algorithm based on discontinuous Galerkin
method is presented for solving the generalized FitzHugh–Nagumo reaction–diffusion model.
Since past years, the discontinuous Galerkin (DG) methods have become popular as a fas-
cinating tool to solve the various partial differential equations in different contexts. This
was originally developed by Reed and Hill in 1973 for the neutron transport problem [32].
Later, DG methods have proven a significant room for the elaboration of state of the art
for numerical schemes in wide range of scientific and engineering applications including
computational fluid dynamics, computational aeroacoustics, and rarefied gas dynamics [33–
47]. The DG methods work as a hybridization approach of finite volume and finite element
methods as it incorporates the main properties commonly associated to these methods. The
DG methods have numerous important features such as hp adaptivity; robustness with strong
mathematical properties including conservation, stability, and convergence; well defined
for structured/unstructured meshes associated within complex geometries; well suited for
non-conforming elements having hanging nodes; very efficient for adopting time-stepping
algorithms and highly parallelizable
Toward our motivation, a mixed-type discontinuous Galerkin method is presented to solve
the generalized FitzHugh–Nagumo reaction–diffusion model along with prescribed initial
and boundary conditions. We organize this paper as follows: In Sect. “Mathematical Model”,
the generalized Fitzhugh–Nagumo reaction–diffusion model is introduced. In Sect. “Mixed-
Type Discontinuous Galerkin Method”, we describe the proposed mixed-type discontinuous
Galerkin method including spatial and temporal discretization for solving the equation. In
Sect. “Numerical Results and Discussion”, the numerical technique is applied to various
problems of our aimed model for authenticating the accuracy and efficiency of the method.
Finally, the concluding remarks are discussed in Sect. “Concluding Remarks”.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int. J. Appl. Comput. Math (2021) 7:207 Page 3 of 16 207

Mathematical Model

In current study, we highlight on the generalized FitzHugh–Nagumo reaction–diffusion

model
∂u ∂u ∂ 2u
+ ψ1 − ψ2 2 = ψ3 [u(1 − u)(u − α)] , (1)
∂t ∂x ∂x
in the domain  = [a, b] and t > 0, where u = u(x, t) is a time dependent unknown
function; ψ1 , ψ2 , ψ3 are the real-valued constant or time-dependent parameters; and α ∈  is
an arbitrary parameter that controls the global dynamics of the equation. The initial condition
is considered as

u(x, 0) = u 0 (x), x ∈ , (2)

while, the Dirichlet boundary are prescribed as

u(a, t) = f 1 (t),
(3)
u(b, t) = f 2 (t), t > 0.

If ψ1 = g1 (t), ψ2 = g2 (t), ψ3 = g3 (t), model (1) reduces to the generalized FitzHugh–

Nagumo reaction–diffusion model with time-dependent coefficients [30]. If ψ1 = 0, ψ2 =
ψ3 = 1 is considered, the mathematical model (1) becomes the classical FitzHugh Nagumo
reaction–diffusion model, originally introduced in the pioneer research of Fitzhugh [14] and
Nagumo et al. [15]

∂u ∂ 2u
− 2 = u(1 − u)(u − α). (4)
∂t ∂x
When α = −1, the classical FitzHugh–Nagumo model (4) turns out to the real Newell–
Whitehead model [48]. The generalized FitzHugh–Nagumo model (1) can be composed in
a compact form as
∂u
+ ∇ F c (u) + ∇ F d (∇u) = S(u), (5)
∂t
where ∇ is the Laplacian operator defined in one-dimensional space as ∇ = ∂∂x ; the symbol
F c denotes the convective term given as F c (u) = ψ1 u; the symbol F d represents the diffusive
term defined by F d (∇u) = −ψ2 ∇u; and S(u) is the reaction term defined as S(u) =
ψ3 [u(1 − u)(u − α)].

Mixed-Type Discontinuous Galerkin Method

In this section, we demonstrate the essentials ideas of the mixed-type discontinuous Galerkin
method in detail. The generalized FitzHugh–Nagumo reaction–diffusion model (5) which has
the high-order derivative in diffusive term, cannot be handled easily with the classical DG
method. In order to overcome with this numerical complexity, a mixed-type discontinuous
Galerkin approach developed in previous literature [33,37–39,45] is adopted to discretize
the generalized FitzHugh–Nagumo model. Following this approach, an auxiliary variable 
is brought for handling the high-order derivative occurring in diffusion term. The auxiliary
variable  is defined as the derivative of unknown variable. Thus, the model (5) can be

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 4 of 16 Int. J. Appl. Comput. Math (2021) 7:207

re-written as a first-order system for u and  in the mixed-type DG formulation:

 − ∇u = 0,
∂u (6)
+ ∇ F c (u) + ∇ F d () = S(u).
∂t

DG Spatial Discretization

For discretization the first-order system of the generalized FitzHugh–Nagumo model (6), the
domain  = [a, b] is partitioned into N uniform elements as follows:

a = x1/2 < x3/2 < · · · < x N +1/2 = b. (7)

The element is defined by In = [xn−1/2 , xn+1/2 ], ∀n ∈ 1, 2, . . . N , where xn =

2 (x n−1/2
+ xn+1/2 ) is the center point and x = xn+1/2 − xn−1/2 denotes the mesh size
1

of the element In . For the domain , we introduce the piecewise polynomial space of the
functions v :   −→  as

Vhk = {v ∈ L 2 (), v |  ∈ P k (In ), ∀n ∈ 1, 2, . . . N }, (8)

where L 2 () is the space function of the squared Lebesque integral over the domain  and
P k (In ) is the space of polynomial functions of degree at most k in element In . The solutions
of u and  are proper approximated in the finite element space Vhk , can be assigned as
follows:

Nk
u(x, t) = û ih (t)ϕi (x),
i=1
(9)

Nk
h (x, t) = ˆ ih (t)ϕi (x),
 x ∈ In ,
i=1

ˆ i denote the modal coefficients of u and  to be enhanced in time, the ϕi (x)

where û ih and  h
denotes the polynomial basis function of degrees k, and the number of polynomial basis
functions in the k-exact DG approximation is denoted Nk = k + 1 which depends on the
order of approximation k. In present study, the orthogonal scaled Legendre polynomial basis
function is used for function ϕi (x) in the reference element In = [−1, 1].

2n (n!)2 0,0
ϕn (ξ ) = P (ξ ), n ≥ 0,
(2n)! n (10)
2(x − x j )
ξ= , −1 ≤ ξ ≤ 1,
x
where Pn0,0 (ξ ) is the Legendre polynomial and x j is the centre of element. In this work,
the first four orthogonal polynomials which correspond to the third-order approximation, are
used.
1 3
ϕ0 (ξ ) = 1, ϕ1 (ξ ) = ξ, ϕ2 (ξ ) = ξ 2 − , ϕ3 (ξ ) = ξ 3 − ξ. (11)
2 5
To formulate the DG method, a test function which is selected as equivalent to the polyno-
mial function ϕi (x), is multiplied into the mixed system (6) and then performing an integration

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int. J. Appl. Comput. Math (2021) 7:207 Page 5 of 16 207

by parts over the element In . As a result, the following weak formulation for Eq. (6) is
obtained:
  
h ϕh d V + u h ∇ϕh d V − ϕh u h · nd
= 0,
In In ∂ In
  
∂
u h ϕh d V − ∇ϕh F (u h )d V +
c
ϕh F c (u h ) · nd
(12)
∂t
In In ∂ In
  
− ∇ϕh F d (h )d V + ϕh F d (h ) · nd
= S(u h )ϕh d V ,
In ∂ In In

where
(= ∂ In ) represents the face boundary of In , the V is the volume integral, and
n indicates the outward unit normal vector to the boundary ∂ In . The interface fluxes of
the numerical solution u h and h cannot be specified uniquely because of discontinuity at
interfaces of element. For handling the discontinuity occurring at interface, we need to define
numerical flux which relies on the discontinuous solutions at both faces of the interface. Here,
we adopt the local Lax-Friedrichs (LLF) scheme to the convective numerical flux F c · n in
Eq. (12),
 
Fc · n ≡ F L L F (u − , u + ) = 1 f (u − ) + f (u + ) − max| f  (u)|(u − + u + ) , (13)
h h h h h h
2
where, the superscripts (+) and (−) indicate the inside and outsides at interface of an ele-
mental. On the other side, the alternating scheme [34] is considered for the diffusion and
auxiliary numerical fluxes
1 −
u Alt (u −
uh · n ≡  +
(u + u +
h , uh ) =
− +
h ) + C 11 (u h − u h ),
2 h (14)
Fd · n ≡ F Alt (− , + ) = 1 (− + + ) − C11 (u − − u + ) − C12 (− − + ).
h h
2 h h h h h h

Utilizing this scheme at boundary interfaces requires that diffusion numerical flux at the
boundary nodes takes the following form:

u Alt
1 = u(a, t), 
u NAlt+ 1 = u(b, t),
2 2
1Alt = +1 − C11 (u −1 − u +1 ),
F (15)
2 2 2 2
Alt 1 = −1 − C11 (u −
F − u+ ).
N+ 2 2 N + 21 N + 21

When C11 = 0 and C12 = 1/2 or −1/2, the fluxes are called the alternating fluxes [34],
while the fluxes are called the central flux scheme or BR1 scheme [33]. Therefore, the weak
formulation of the mixed form (12) becomes,
  
h ϕh d V + u h ∇ϕh d V − ϕh u Alt d
= 0,
In In ∂ In
  
∂ L L F · nd

u h ϕh d V − ∇ϕh F c (u h )d V + ϕh F (16)
∂t
In In ∂ In
  
− ∇ϕh F d (h )d V + Alt d
=
ϕh F S(u h )ϕh d V .
In ∂ In In

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 6 of 16 Int. J. Appl. Comput. Math (2021) 7:207

In expressions (16), the boundary and volume integration can be calculated by the exact
formulation or numerical quadrature rule with polynomials of degree 2k − 1. Here, the
Gauss-Legendre quadrature rule with relevant number of Gaussian points is employed for
the evaluation of associated integrals [36].

Temporal Discretization

By combining all the computational assessments jointly, the mixed-type DG formulation

generates a system of semi-discrete ODEs in time for the generalized FitzHugh–Nagumo
model (1):
du h
M = L(u h ), (17)
dt
where M represents the orthogonal mass matrix which can be easily computed by hand and
L(u h ) denotes the residual function of the degrees of freedom. The semi-discrete form of
ODE (17) can be discretized in time using various time integrator methods. In present work,
an explicit three-stage, third-order accurate SSP Runge-Kutta method (SSP-RK33) proposed
by Shu and Osher [49] is employed,
(1)
u h = u nh + tM−1 R(u nh ),
(2) 3 1 (1) 1 (1)
u h = u nh + u h + tM−1 R(u h ), (18)
4 4 4
1 2 (2) 2 (2)
u n+1
h = u nh + u h + tM−1 R(u h ),
3 3 3
where R(u nh ) represents a numerical approximation of the solution at time tn and t denotes
the appropriate time-step.

Numerical Results and Discussion

In this section, we elaborate the applications of the mixed-type DG method on some test
problems of the generalized Fitzhugh–Nagumo model. All the test problems have analytic
solutions which help to measure the accuracy of present numerical method rigorously. So
far, all examples are solved by the third-order DG scheme, namely k = 2 in Eq. (9). All
the computations have been performed with the help of Visual studio 2019 and Fortran 90
platform. To illustrate the performance of this method, the accuracy is measured with L ∞ ,
L 2 error-norms and absolute error, defined as
 
L ∞ − error = max (û(xi , t n ) − u h (xi , t n )),
1≤i≤N

N
  
L 2 − error =
û(xi , t n ) − u h (xi , t n )2 , (19)
i=1
 
Absolute error = û(xi , t n ) − u h (xi , t n )
where û(xi , t n ) and u h (xi , t n ) denote the exact and the approximate solution, respectively.
The convergence rate of the numerical scheme is determined as
log(Err (N1 /Err (N2 ))
Or der = (20)
log(N2 /N1 )

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int. J. Appl. Comput. Math (2021) 7:207 Page 7 of 16 207

The Err (N1 ) and Err (N2 ) represent the error-norms (L 2 and L ∞ ) at the numbers of elements
N and 2N , respectively.

Problem 4.1 (Constant Coefficient Model)

For a first test problem, the following classical Fitzhugh–Nagumo reaction–diffusion model
is considered [50,51]:
∂u ∂ 2u
= + u(1 − u)(u − α), 0 < α < 1. (21)
∂t ∂x2
For this problem, the initial and boundary conditions are provided as
 
1 1 (1 − α)
u(x, 0) = (1 + α) + (1 − α)tanh √ x , (22a)
2 2 2 2
lim u(x, t) = 0.2, lim u(x, t) = 1.0. (22b)
x→−∞ x→∞

The analytic solution is specified as

 
1 1 (1 − α) (1 − α 2 )
u(x, t) = (1 + α) + (1 − α)tanh √ x+ √ t . (23)
2 2 2 2 2 2
For numerical computations, the domain is considered as [a, b] = [−22, 22] and par-
titioned with N = 401. The obtained results are illustrated in Figs. 1 and 2. Figure 1
demonstrates a comparison between exact and numerical solutions for α = 0.2, 0.5, 0.75
at t = 0.01. It is observed from Fig. 1 that present numerical solution is quite similar to
the exact solution. This shows that the present mixed-type DG solver is capable to compute
the solution-profile accurately. The space–time plot of numerical solutions are also plotted
for α = 0.2, 0.5, 0.75 in shown in Fig. 2. For error comparison with the existing results
[50,51], we exhibit the absolute errors for α = 0.2, t = 0.0001 at three distinct time levels
t = 0.001, 0.002, 0.003 as shown in Table 1. In addition, we conduct the convergence test
for the applied method by varying number of grid points N (20, 40, 80, 160). Table 2 reports
the errors and rate of convergence of the employed DG approach and their graphical repre-
sentation is illustrated in Fig. 3. From the results, it is affirmed that the present numerical
scheme achieved the desired order of accuracy (k + 1).

Problem 4.2 (Constant Coefficient Model)

Consider another example of the classical Fitzhugh–Nagumo reaction–diffusion model with

constant coefficients [30,52]:
∂u ∂ 2u
= + u(1 − u)(u − α), (24)
∂t ∂x2
with initial and boundary conditions
 
1 1 x
u(x, 0) = + tanh √ , (25a)
2 2 2 2
 
1 1 1 2α − 1
u(a, t) = + tanh √ a − √ , (25b)
2 2 2 2 2
 
1 1 1 2α − 1
u(b, t) = + tanh √ b − √ . (25c)
2 2 2 2 2

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 8 of 16 Int. J. Appl. Comput. Math (2021) 7:207

Fig. 1 Comparing the numerical 1

and exact solutions for Problem
4.1 with α = 0.2, 0.5, 0.75 at
t = 0.01
0.8

u(x,t)
0.6

0.4 Numerical sol; α = 0.2

Exact sol; α = 0.2
Numerical sol; α = 0.5
Exact sol; α = 0.5
Numerical sol; α = 0.75
Exact sol; α = 0.75
0.2
-20 -10 0 10 20
x

Fig. 2 Space–time plot of numerical solutions for Problem 4.1 at a α = 0.2, b α = 0.5, and c α = 0.75 with
t ∈ [0, 0.01]

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Int. J. Appl. Comput. Math

Table 1 Solutions comparison for Problem 4.1 in terms of absolute errors with ρ = 0.2 at different time levels
(2021) 7:207

t = 0.001 t = 0.0002 t = 0.0003

x Ref. [50] Ref. [51] Present Ref. [50] Ref. [51] Present Ref. [50] Ref. [51] Present

−22 1.68083E−7 1.36767E−7 1.43598E−7 3.36253E−7 2.88809E−7 2.95513E−7 5.0451E−7 4.56151E−7 4.09754E−7
−6 5.18408E−6 4.61128E−6 4.84401E−7 1.03737E−5 9.74011E−6 9.99052E−6 1.55688E−5 1.53881E−5 1.56710E−5
2 2.95696E−5 2.66947E−5 2.75120E−5 5.91566E−5 5.63389E−5 5.81123E−5 8.87612E−5 8.87583E−5 8.62094E−5
10 1.07392E−6 9.15268E−7 1.00667E−7 2.14679E−6 1.93128E−6 2.01140E−6 3.21863E−6 3.04774E−6 4.99235E−6
18 5.45852E−7 4.27262E−7 5.13045E−7 1.09127E−6 9.01634E−7 9.21599E−7 1.63624E−6 1.42300E−6 1.61024E−6
Page 9 of 16
207

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 10 of 16 Int. J. Appl. Comput. Math (2021) 7:207

Table 2 Errors and estimated k N L 2 -error order L ∞ -error order

orders of convergence for
Problem 4.1 1 20 6.89292E−02 – 4.16916E−02 –
40 1.79328E−02 1.91 1.06198E−02 1.86
80 4.53598E−03 1.87 2.65698E−03 2.13
160 1.13801E−03 1.79 6.63269E−04 2.11
2 20 1.04224E−03 – 1.24147E−03 –
40 1.30558E−04 3.01 1.58800E−04 2.89
80 1.63282E−05 3.10 2.00245E−05 2.81
160 2.04129E−06 3.02 2.51331E−06 2.90
3 20 1.76257E−05 – 2.70453E−05 –
40 1.10354E−06 4.02 1.70007E−06 3.88
80 6.89895E−08 4.05 1.06066E−07 3.79
160 4.31213E−09 4.00 6.62179E−09 4.01

0
10
0 10

1
P
1 P
2
P
2 P
-2
10-2 P3 10 P3

-4
10
-4 10
Linf - error
L2 - error

10-6 10-6

-8
10
-8 10

10-10 1 10-10 1
10 102 103 10 102 103
Number of elements Number of elements

Fig. 3 Accuracy test for Problem 4.1: L 2 and L ∞ -errors with number of elements

The exact solution was found in [52] as

 
1 1 1 2α − 1
u(x, t) = + tanh √ x − √ . (26)
2 2 2 2 2

Here, the domain is taken as [a, b] = [−10, 10] with partitioned 401 elements. The
numerical simulations are performed for α = −2, −1 and 4. Figure 4 displays the numerical
results with exact solutions and their space–time plots for three different α values at distinct
time levels. Table 3 reports the comparison between numerical and exact solutions for ρ = −1
( i.e. for the real Newell–Whitehead model) at different times t = 0.1, 0.5, 1.0. It is cleared
that the obtained results demonstrate the good accuracy of the applied mixed-type DG scheme.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Int. J. Appl. Comput. Math

Table 3 Solutions comparison for Problem 4.2 in terms of absolute errors with α = −1 at different time levels
t = 0.1 t = 0.5 t = 1.0
(2021) 7:207

x Numerical Exact Abs. error Numerical Exact Abs. error Numerical Exact Abs. error

−10 0.123147E−2 0.123172E−2 2.97419E−6 0.273299E−2 0.273123E−2 1.08635E−6 0.639771E−2 0.639709E−2 4.54461E−6
−7 0.800058E−2 0.799954E−2 1.00502E−6 0.149662E−1 0.149620E−1 1.08635E−6 0.639771E−2 0.639709E−2 4.54461E−6
−3 0.122808E+0 0.122814E+0 1.3525E−6 0.198340E+0 0.198331E+0 2.98439E−6 0.351533E+0 0.351539E+0 9.54360E−3
0 0.537585E+0 0.537579E+0 1.55031E−6 0.680007E+0 0.680017E+0 8.28788E−6 0.819048E+0 0.819036E+0 4.715131E−6
3 0.909004E+0 0.909059E+0 2.44252E−6 0.948717E+0 0.9487161E+0 9.41522E−6 0.975967E+0 0.975932E+0 3.34350E−6
7 0.994331E+0 0.994343E+0 2.6837E−6 0.997569E+0 0.997526E+0 9.821601E−7 0.999694E+0 0.999603E+0 1.90080E−6
10 0.999372E+0 0.999376E+0 9.1381E−7 0.100002E+1 0.9999983E+0 7.363843E−7 0.100001E+1 0.999998E+0 1.93001E−7
Page 11 of 16
207

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 12 of 16 Int. J. Appl. Comput. Math (2021) 7:207

(a) 1

0.8

0.6
u(x,t)

Numerical sol; t = 0
Exact sol; t = 0
Numerical sol; t = 1
0.4 Exact sol; t = 1
Numerical sol; t = 2
Exact sol; t = 2
Numerical sol; t = 3
Exact sol; t = 3
0.2 Numerical sol; t = 4
Exact sol; t = 4
Numerical sol; t = 5
Exact sol; t = 5
0
-10 -5 0 5 10
x

(b) 1

0.8

0.6
u(x,t)

Numerical sol; t = 0.0

Exact sol; t = 0.0
0.4 Numerical sol; t = 0.2
Exact sol; t = 0.2
Numerical sol; t = 0.4
Exact sol; t = 0.4
Numerical sol; t = 0.6
0.2 Exact sol; t = 0.6
Numerical sol; t = 0.8
Exact sol; t = 0.8
Numerical sol; t = 1.0
Exact sol; t = 1.0
0
-10 -5 0 5 10
x

(c) 1
Numerical sol; t = 0.0
Exact sol; t = 0.0
Numerical sol; t = 0.2
Exact sol; t = 0.2
Numerical sol; t = 0.4
0.8
Exact sol; t = 0.4
Numerical sol; t = 0.6
Exact sol; t = 0.6
Numerical sol; t = 0.8
Exact sol; t = 0.8
0.6
Numerical sol; t = 1.0
Exact sol; t = 1.0
u(x,t)

0.4

0.2

0
-10 -5 0 5 10
x

Fig. 4 Profile comparison between exact and numerical solutions and the space–time plot for Problem 4.2 at
a α = −2, b α = −1, and c α = 4. Arrow symbol represents the direction of traveling wave solutions

Problem 4.3 (Time-Dependent Coefficient Model)

Consider the generalized Fitzhugh–Nagumo reaction–diffusion model with time-dependent

coefficients [29–31]:
∂u ∂u ∂ 2u
+ cos(t) = cos(t) 2 + 2cos(t) (u(1 − u)(u − α)) , (27)
∂t ∂x ∂x
with the prescribed initial and boundary conditions
α α  αx 
u(x, 0) = + tanh , (28a)
2 2 2

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int. J. Appl. Comput. Math (2021) 7:207 Page 13 of 16 207

Fig. 5 Comparison between 0.5 Numerical sol; t = 0.0

Exact sol; t = 0.0
exact and numerical solution for Numerical sol; t = 2.2
Problem 4.3 with α = 0.5 at Exact sol; t = 2.2
several time levels Numerical sol; t = 2.7
0.4 Exact sol; t = 2.7
Numerical sol; t = 3.0
Exact sol; t = 3.0

0.3

u(x,t)
0.2

0.1

0
-10 -5 0 5 10
x

Table 4 Solutions comparison L ∞ −error L 2 −error

for Problem 4.3 in terms of L ∞
and L 2 error-norms with t Ref. [31] Present Ref. [31] Present
α = 0.75 at different time levels
0.2 1.2350E−5 1.1228E−5 1.3122E−6 7.1547E−6
0.5 5.1986E−4 7.5845E−5 6.0995E−6 7.4891E−6
1.0 6.3283E−4 9.1278E−5 2.1213E−5 8.6891E−5
1.5 8.5383E−4 9.2423E−5 3.2340E−5 7.0278E−5
2.0 9.9123E−4 1.7724E−5 4.8342E−5 2.4271E−5
3.0 1.7904E−3 1.4609E−4 6.1098E−5 6.6466E−5

α α α 
u(a, t) =
+ tanh (a − (3 − α)sin(t)) , (28b)
2 2  2 
α α α
u(b, t) = + tanh (b − (3 − α)sin(t)) . (28c)
2 2 2
The exact solution was studied in [29,30] as
α α α 
u(x, t) = + tanh (x − (3 − α)sin(t)) . (29)
2 2 2
Here, the domain is considered as [a, b] = [−10, 10] with 401 cells for the computations.
The numerical simulations are performed for the parameters; α = 0.25, 0.5, 0.75, 3.0 and
t ∈ [0, 3]. Figure 5 depicts a comparison between numerical and exact solutions for α = 0.5
at different times t = 0, 2.2, 2.7, 3. The space-time plots of numerical solutions at distinct
α values are illustrated in Fig. 6. The numerical solution at α = 3.0 travels only in space-
direction and independent of time-direction as shown in Fig. 6b. Table 4 illustrates the result
comparison with the result obtained from differential quadrature method [31] with α = 0.75
at several time levels. As seen in Table 4, the present results are improved than the results
gained in Ref. [31].

Concluding Remarks

A mixed-type discontinuous Galerkin method is presented for solving the generalized

FitzHugh–Nagumo reaction–diffusion model. The numerical scheme is based on the concept

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 14 of 16 Int. J. Appl. Comput. Math (2021) 7:207

Fig. 6 Space–time plots of numerical solution for Problem 4.3 at a α = 0.25, b α = 0.5, c α = 0.75, and d
ρ = 3 with t ∈ [0, 3]

of introducing an auxiliary unknown for high order derivative involved in diffusion term. A
third-order DG method with scaled Legendre polynomials is used for the spatial discretiza-
tion. An explicit three-stages, third-order SSP-RK (SSP-RK33) scheme is implemented for
temporal discretization to the resulting semi-discrete ordinary differential equation. The
present scheme is applied on three different examples of the generalized FitzHugh–Nagumo
model with prescribed initial and boundary conditions. The numerical solution are com-
pared with the exact solution and shown in L ∞ , L 2 , and absolute error form. This presented
scheme gives fairly good solution when compared results with exact solution. Relying on the
successful implementation of the mixed-type DG numerical scheme, the extension to high-
dimensional nonlinear partial differential equations will be the future topic of the current
study.

Acknowledgements The author would like to acknowledge the financial support from the NAP-SUG grant
program funded by the Nanyang Technological University, Singapore.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Int. J. Appl. Comput. Math (2021) 7:207 Page 15 of 16 207

Data Availability The data that support the findings of this study are available from the corresponding author
upon reasonable request.

References
1. Murray, J.D.: Mathematical Biology: I. An Introduction. Springer (2007)
2. Turing, A.M.: The chemical basis of morphogenesis. Bull. Math. Biol. 52, 153 (1990)
3. Pearson, J.E.: Complex patterns in a simple system. Science 261, 189 (1993)
4. Canosa, J.: On a nonlinear diffusion equation describing population growth. IBM J. Res. Dev. 17, 307
(1973)
5. Luther, R.: Propagation of chemical reactions in space. J. Chem. Educ. 64, 740 (1987)
6. Andow, D.A., Kareiva, P.M., Levin, S.A., Okubo, A.: Spread of invading organisms. Landsc. Ecol. 4, 177
(1990)
7. Murray, J.D., Stanley, E.A., Brown, D.L.: On the spatial spread of rabies among foxes. Proc. R. Soc.
Lond. 229, 111 (1986)
8. Chaplain, M.A.J.: Reaction–diffusion prepatterning and its potential role in tumour invasion. J. Biol. Syst.
3, 929 (1995)
9. Hodgkin, A.L., Huxley, A.F.: A quantitative description of membrane current and its application to
conduction and excitation in nerve. J. Physiol. 117, 500 (1952)
10. Argentina, M., Coullet, P., Krinsky, V.: Head-on collisions of waves in an excitable FitzHugh–Nagumo
system: a transition from wave annihilation to classical wave behavior. J. Theor. Biol. 205, 47 (2000)
11. Zorzano, M.P., Vázquez, L.: Emergence of synchronous oscillations in neural networks excited by noise.
Phys. D 179, 105 (2003)
12. Gong, P.L., Xu, J.X., Hu, S.J.: Resonance in a noise-driven excitable neuron model. Chaos Soliton Fract.
13, 885 (2002)
13. Margerit, D., Barkley, D.: Selection of twisted scroll waves in three-dimensional excitable media. Phys.
Rev. Lett. 86, 175 (2001)
14. FitzHugh, R.: Impulses and physiological states in theoretical models of nerve membrane. Biophys. J. 1,
445 (1961)
15. Nagumo, J., Arimoto, S., Yoshizawa, S.: An active pulse transmission line simulating nerve axon. Proc.
IRE 50, 2061 (1962)
16. Abbasbandy, S.: Soliton solutions for the FitzHugh–Nagumo equation with the homotopy analysis method.
Appl. Math. Model. 32, 2706 (2008)
17. Abdusalam, H.A.: Analytic and approximate solutions for Nagumo telegraph reaction diffusion equation.
Appl. Math. Comput. 157, 515 (2004)
18. Aronson, D.G., Weinberger, H.F.: Multidimensional nonlinear diffusion arising in population genetics.
Adv. Math. 30, 33 (1978)
19. Li, H., Guo, Y.: New exact solutions to the FitzHugh–Nagumo equation. Appl. Math. Comput. 180, 524
(2006)
20. Kawahara, T., Tanaka, M.: Interactions of traveling fronts: an exact solution of a nonlinear diffusion
equation. Phys. Lett. A 97, 311 (1983)
21. Yanagida, E.: Stability of travelling front solutions of the FitzHugh–Nagumo equations. Math. Comput.
Model. 12, 289 (1989)
22. Nucci, M.C., Clarkson, P.A.: The nonclassical method is more general than the direct method for symmetry
reductions. An example of the FitzHugh–Nagumo equation. Phys. Lett. A 164, 49 (1992)
23. Kakiuchi, N., Tchizawa, K.: On an explicit duck solution and delay in the FitzHugh–Nagumo equation.
J. Differ. Equ. 141, 327 (1997)
24. Shih, M., Momoniat, E., Mahomed, F.M.: Approximate conditional symmetries and approximate solutions
of the perturbed FitzHugh–Nagumo equation. J. Math. Phys. 46, 023503 (2005)
25. Alford, J.G.: Bifurcation structure of rotating wave solutions of the FitzHugh–Nagumo equations. Com-
mun. Nonlinear Sci. Numer. Simul. 14, 3282 (2009)
26. Hariharan, G., Kannan, K.: Haar wavelet method for solving FitzHugh–Nagumo equation. Int. J. Comput.
Math. Sci. 2, 2 (2010)
27. Olmos, D., Shizgal, B.D.: Pseudospectral method of solution of the FitzHugh–Nagumo equation. Math.
Comput. Simulat. 79, 2258 (2009)
28. Van, G., Robert, A., Vajravelu, K.: A variational formulation of the Nagumo reaction–diffusion equation
and the Nagumo telegraph equation. Nonlinear Anal. Real World Appl. 11, 2957 (2010)

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 207 Page 16 of 16 Int. J. Appl. Comput. Math (2021) 7:207

29. Triki, H., Wazwaz, A.M.: On soliton solutions for the FitzHugh–Nagumo equation with time-dependent
coefficients. Appl. Math. Model. 37, 3821 (2013)
30. Bhrawy, A.H.: A Jacobi–Gauss–Lobatto collocation method for solving generalized FitzHugh–Nagumo
equation with time-dependent coefficients. Appl. Math. Comput. 222, 255 (2013)
31. Jiwari, R., Gupta, R.K., Kumar, V.: Polynomial differential quadrature method for numerical solutions
of the generalized FitzHugh–Nagumo equation with time-dependent coefficients. Ain Shams Eng. J. 5,
1343 (2014)
32. Reed, W.H., Hill, T.R.: Triangular mesh methods for the neutron transport equation. Tech. Rep., Los
Alamos Scientific Lab., N. Mex., USA (1973)
33. Bassi, F., Rebay, S.: A high-order accurate discontinuous finite element method for the numerical solution
of the compressible Navier–Stokes equations. J. Comput. Phys. 131, 267 (1997)
34. Cockburn, B., Shu, C.W.: The local discontinuous Galerkin method for time-dependent convection–
diffusion systems. SIAM J. Numer. Anal. 35, 2440 (1998)
35. Cockburn, B., Shu, C.W.: The Runge–Kutta discontinuous Galerkin method for conservation laws V:
multidimensional systems. J. Comput. Phys. 141, 199 (1998)
36. Cockburn, B., Shu, C.W.: Runge–Kutta discontinuous Galerkin methods for convection-dominated prob-
lems. J. Sci. Comput. 16, 173 (2001)
37. Raj, L.P., Singh, S., Karchani, A., Myong, R.S.: A super-parallel mixed explicit discontinuous Galerkin
method for the second-order Boltzmann-based constitutive models of rarefied and microscale gases.
Comput. Phys. 157, 146 (2017)
38. Singh, S., Myong, R.S.: A computational study of bulk viscosity effects on shock-vortex interaction using
discontinuous Galerkin method. J. Comput. Fluids Eng. 22, 86 (2017)
39. Singh, S.: Development of a 3D discontinuous Galerkin method for the second-order Boltzmann-Curtiss
based hydrodynamic models of diatomic and polyatomic gases. Ph.D. Thesis, Gyeongsang National
University, South Korea (2018)
40. Singh, S., Karchani, A., Myong, R.S.: Non-equilibrium effects of diatomic and polyatomic gases on the
shock-vortex interaction based on the second-order constitutive model of the Boltzmann–Curtiss equation.
Phys. Fluids 30, 016109 (2018)
41. Singh, S.: Role of Atwood number on flow morphology of a planar shock-accelerated square bubble: a
numerical study. Phys. Fluids 32, 126112 (2020)
42. Singh, S., Battiato, M.: Effect of strong electric fields on material responses: the Bloch oscillation reso-
nance in high field conductivities. Materials 13, 1070 (2020)
43. Singh, S., Battiato, M.: Strongly out-of-equilibrium simulations for electron Boltzmann transport equation
using explicit modal discontinuous Galerkin method. Int. J. Appl. Comput. Math. 6, 1 (2020)
44. Singh, S., Battiato, M.: An explicit modal discontinuous Galerkin method for Boltzmann transport equa-
tion under electronic nonequilibrium conditions. Comput. Phys. 224, 104972 (2021)
45. Singh, S.: 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.
Int. J. Heat Mass Transf. 179, 121708 (2021)
46. Singh, S., Battiato, M.: Behavior of a shock accelerated heavy cylindrical bubble under nonequilibrium
conditions of diatomic and polyatomic gases. Phys. Rev. Fluids 6, 044001 (2021)
47. Singh, S., Battiato, M., Myong, R.S.: Impact of the bulk viscosity on flow morphology of shock-bubble
interaction in diatomic and polyatomic gases. Phys. Fluids 33, 066103 (2021)
48. Newell, A.C., Whitehead, J.A.: Finite bandwidth, finite amplitude convection. J. Fluid Mech. 38, 279
(1969)
49. Shu, C.W., Osher, S.: Efficient implementation of essentially non-oscillatory shock-capturing schemes.
J. Comput. Phys. 77, 439 (1988)
50. Arora, G., Joshi, V.: A computational approach for solution of one dimensional parabolic partial differ-
ential equation with application in biological processe. Ain Shams Eng. J. 9, 1141 (2018)
51. Dhiman, N., Tamsir, M.: A collocation technique based on modified form of trigonometric cubic B-spline
basis functions for Fisher’s reaction-diffusion equation. Multidiscip. Model. Mater. Struct. 14, 923 (2018)
52. Wazwaz, A.M.: The tanh-coth method for solitons and kink solutions for nonlinear parabolic equations.
Appl. Math. Comput. 188, 1467 (2007)

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

123
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center
GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers
and authorised users (“Users”), for small-scale personal, non-commercial use provided that all
copyright, trade and service marks and other proprietary notices are maintained. By accessing,
sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of
use (“Terms”). For these purposes, Springer Nature considers academic use (by researchers and
students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and
conditions, a relevant site licence or a personal subscription. These Terms will prevail over any
conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription (to
the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of
the Creative Commons license used will apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may
also use these personal data internally within ResearchGate and Springer Nature and as agreed share
it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not otherwise
disclose your personal data outside the ResearchGate or the Springer Nature group of companies
unless we have your permission as detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial
use, it is important to note that Users may not:

1. use such content for the purpose of providing other users with access on a regular or large scale
basis or as a means to circumvent access control;
2. use such content where to do so would be considered a criminal or statutory offence in any
jurisdiction, or gives rise to civil liability, or is otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association
unless explicitly agreed to by Springer Nature in writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a
systematic database of Springer Nature journal content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a
product or service that creates revenue, royalties, rent or income from our content or its inclusion as
part of a paid for service or for other commercial gain. Springer Nature journal content cannot be
used for inter-library loans and librarians may not upload Springer Nature journal content on a large
scale into their, or any other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not
obligated to publish any information or content on this website and may remove it or features or
functionality at our sole discretion, at any time with or without notice. Springer Nature may revoke
this licence to you at any time and remove access to any copies of the Springer Nature journal content
which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or
guarantees to Users, either express or implied with respect to the Springer nature journal content and
all parties disclaim and waive any implied warranties or warranties imposed by law, including
merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published
by Springer Nature that may be licensed from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a
regular basis or in any other manner not expressly permitted by these Terms, please contact Springer
Nature at

onlineservice@springernature.com