Mechanics of Materials 100 (2016) 148–153

Contents lists available at ScienceDirect

Mechanics of Materials
journal homepage: www.elsevier.com/locate/mechmat

An equivalence between generalized Maxwell model and fractional

Zener model
Rui Xiao∗, Hongguang Sun, Wen Chen
Institute of Soft Matter Mechanics, College of Mechanics and Materials, Hohai University, Nanjing, Jiangsu 210098, China

a r t i c l e i n f o a b s t r a c t

Article history: Both classic rheological models and fractional derivative models have been widely adopted to model the
Received 5 March 2016 viscoelastic behaviors of materials. In this work, we present a detailed comparison of the performance of
Revised 1 June 2016
the generalized Maxwell model and fractional Zener model. We first describe a method to determine the
Available online 24 June 2016
parameters of the generalized Maxwell model from the fractional Zener model based on the equivalence
Keywords: of complex modulus in the frequency domain of the two models. The two models are then applied to
Viscoelasticity investigating the stress response under constant strain rate, stress relaxation, cyclic and random load-
Rheological model ing conditions. The simulation results of the two models show excellent quantitatively equivalence. This
Fractional model finding can provide insight into choosing the more suitable model for specific conditions.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Viscoelasticity represents that the behavior of materials is in-
termediate between linear solids and viscous liquid (Ferry, 1980).
When deformed, most of polymers and biological tissues exhibit
this time-dependent viscous behavior represented as a stiffer stress
response at a higher loading rate and a more compliant response
at a lower loading rate. To understand the origin of viscoelasticity,
various physical-based models have been proposed (Rouse et al.,
1953; De Gennes, 1979; Doi and Edwards, 1988; Li et al., 2015). For
example, the Rouse model (Rouse et al., 1953) was used to explain
the properties of unentangled polymer solutions and melts. The
reptation model (De Gennes, 1979; Doi and Edwards, 1988) was
used to explain the relaxation and viscosity of entangled polymeric
materials. However, for engineering applications, the most widely
used method to describe the viscoelasticity is based on rheological
models.
The viscoelastic rheological models contain the elastic compo-
nents modeled as springs and the viscous components modeled as
dashpots (Ferry, 1980). Based on the arrangement of these compo-
nents, various models have been developed, such as the Maxwell
model, the Kelvin-Vogit model, the Zener model and more com-
plex generalized Maxwell model. The generalized Maxwell model
contains an elastic spring in parallel with multiple Maxwell mod-
els to represent the relaxation occurring at a broad distribution of
time. This model has been successfully applied to studying var-
ious viscoelastic solids (Del Nobile et al., 2007; Kaufman et al.,
∗
Corresponding author.
E-mail address: rxiao@hhu.edu.cn (R. Xiao).
2008; Yu et al., 2014; Xiao et al., 2015). For example, Del Nobile
et al. (2007) has used the generalized Maxwell model to fit the
experimental data of five different food matrices. Kaufman et al.
(2008) has applied the generalized Maxwell model to studying the
stress relaxation of hydrogels, while Yu et al. (2014) and Xiao et al.
(2015) have employed this model to describe the shape-memory
behaviors of amorphous polymers.
Though the classic viscoelastic models can well describe the ex-
perimental results, they typically involve excessive number of pa-
rameters. It is shown that the viscoelastic models can be gener-
alized into fractional derivative models (Koeller, 1984; Bagley and
Torvik, 1986; Schiessel et al., 1995). In recent years, fractional mod-
els have been widely adopted in the field of diffusion (Wu, 2012;
Wang et al., 2010; Zhao and Sun, 2011), heat transfer (Jiang and
Qi, 2012), chaos (Baleanu et al., 2015; Wu and Baleanu, 2015) and
nonlocal elasticity (Tarasov, 2014). However, the most extensive ap-
plication of fractional models still lies in the field of linear vis-
coelasticity (Mainardi, 2010). The general procedure to obtain the
fractional viscoelastic model is through replacing the derivative of
order 1 of the dashpot with the fractional derivative of order be-
tween 0 and 1. Through this process, various fractional deriva-
tive models can be obtained, such as fractional Maxwell model,
fractional Kelvin–Vogit model, fractional Zener model (Mainardi,
2010) and more complex model as shown in Arikoglu (2014). These
models have been widely adopted to describe the relaxation and
creep behaviors of elastomers (Di Paola et al., 2011) and natu-
ral materials (Cataldo et al., 2015), dynamic behavior of biologi-
cal tissue (Kohandel et al., 2005) and other solids (Rossikhin and
Shitikova, 2010), and visco-elastic Euler–Bernoulli beam (Di Paola
et al., 2013). In addition, Fan et al. (2015) and Yu et al. (2015) have

http://dx.doi.org/10.1016/j.mechmat.2016.06.016
0167-6636/© 2016 Elsevier Ltd. All rights reserved.
 R. Xiao et al. / Mechanics of Materials 100 (2016) 148–153 149

The stress response is given by,

σ = E eq ε + E neq ε e , (5)

where E eqis the modulus of the equilibrium elastic spring and

E neq is the modulus of the spring in the non-equilibrium fractional
damping Maxwell element.
The evolution of ε v in the fractional damping element can be
described as Haupt et al. (20 0 0),
dα ε v ε − εv
Fig. 1. Rheological representative of (a) generalized Maxwell model and (b) frac-
tional Zener model. = , ε v (t = 0 ) = 0, (6)
dt α τα
dα ( · )
developed numerical algorithm to obtain the model parameters for where is the Riemann–Liouville fractional derivative with 0 <
dt α
fractional derivative models. Though both rheological models and α < 1, which is defined as (Haupt et al., 20 0 0; Nguyen et al., 2010),
fractional models can be applied to describe the viscoelastic be- 
dα ( f ) 1 d t
f (s )
haviors, limited work has been done to compare their performance = ds, f (t = 0 ) = 0. (7)
in detail. dt α (1 − α ) dt 0 (t − s )α
In this work, we present a numerical study to compare the
 (x) is the Eulerian Gamma function.
performance of generalized Maxwell model and fractional Zener
Eqs. (4)–(6) complete the fractional Zener model.
model. The model descriptions are pesented in Section 2. The fol-
lowing section describes the procedure of obtaining the model pa-
rameters of generalized Maxwell model from an approximation be- 3. Relaxation spectrum of generalized Maxwell model
tween the dynamic modulus of the two models. Finally, we com-
pare the performance of the two models under four different types The procedures of obtaining viscoelastic relaxation spectrum of
of loading conditions: constant strain rate, stress relaxation, cyclic generalized Maxwell model from fractional derivative model have
loading and a random loading condition. been discussed in detail in Haupt et al. (20 0 0); Nguyen et al.
(2010); Xiao et al. (2013) and Xiao and Nguyen (2015). Here we
2. Constitutive model briefly summarized the main processes.
The dynamic storage and loss modulus of the two models under
2.1. Generalized Maxwell model a small sinusoidal oscillations can be represented as,


N E j ω2 τ j2
The rheological representative of generalized Maxwell model is 
Egene (ω ) = E eq + ,
shown in Fig. 1a, which is composed of a spring to describe the j
1 + ω2 τ j2
equilibrium elastic response and multiple Maxwell elements ar-
 (ω ) = E eq + E neq ((ωτ )2α + (ωτ )α cos(απ /2 )
ranged in parallel to represent the viscoelastic response. Efrac ,
1 + (ωτ )2α + (ωτ )α cos(απ /2 )
In each Maxwell element, the total strain of the spring ε ej and
the dashpot ε vj should be equal to the strain ɛ in the elastic branch, 
N
E j ωτ j
 (ω ) =
Egene ,
which yields,
j
1 + ω2 τ j2
ε = ε ej + ε vj , j = 1..N, (1)
 (ω ) = E neq (ωτ )α sin(απ /2 )
Efrac , (8)
where N is the total number of Maxwell elements. 1 + (ωτ )2α + (ωτ )α cos(απ /2 )
The total stress is given by, where ω is the angular frequency, E (ω) denotes the storage mod-

N ulus, and E  (ω) denotes the loss modulus.
σ = E eq ε + E j ε ej , (2) The viscoelastic spectra h(ν ) can be calculated from the com-
j plex moduli E ∗ (iω ) = E  (ω ) + iE  (ω ) using the inverse Stieltjes
where E eq is the equilibrium elastic modulus and Ej is the modulus transform (Christensen, 2003),

of the spring in jth Maxwell element. E ∗ (iω ) ∞
h (ν )
The following linear evolution equation is adopted for ε vj , = d ν, (9)
iω 0 ν + iω
ε − ε vj where ν is the relaxation frequency, which is inverse of the relax-
ε˙ vj = , ε vj (t = 0 ) = 0, (3)
τj ation time.
The cumulative relaxation spectra are defined as H (ν ) =
where τ j is the relaxation time of the dashpot in jth Maxwell ν
0 h (u )du, which yields the cumulative spectra of the two mod-
element. els,
Eqs. (1)–(3) complete the generalized Maxwell model, which
contains the following parameters: equilibrium elastic modulus E eq 
N  
and viscoelastic relaxation spectrum (τ j , Ej ). Hgene (ν ) = Ej ν − νj ,
j
 
2.2. Fractional Zener model (ντ )α + cos(απ )
E neq 1
Hfrac (ν ) = arctan −π −α , (10)
απ sin(απ ) 2
The 1D rheological representative of fractional Zener model is
shown in Fig. 1b, which is consisted of an equilibrium elastic where ν − ν j  = 1 for ν ≥ ν j and ν − ν j  = 0 for ν < ν j .
spring in parallel with a fractional damping Maxwell element. A power law distribution of relaxation frequencies is then
Similarly, the total strain in the non-equilibrium branch equals assumed,
to that of the equilibrium branch, which gives,  j−1

νmax
0 N−1

ε =ε +ε . e v
(4) ν =ν
0 0
. (11)
j min
νmin
0
150 R. Xiao et al. / Mechanics of Materials 100 (2016) 148–153

Fig. 2. (a) The viscoelastic relaxation spectra with τ = 0.1, (b) comparison of the storage modulus between generalized Maxwell model and fractional Zener model.

Fig. 3. Comparison of the stress response at constant strain rate of generalized Maxwell model and fractional Zener model, (a) τ = 0.1, α = 0.7, (b) τ = 0.01, α = 0.7,
(c) τ = 0.1, α = 0.5, (d) τ = 0.01, α = 0.5.

where νmax
0 and νmin
0 are the upper and low limit of relaxation fre- imation of that of the fractional Zener model, which gives,
quency.
To obtain the viscoelastic relaxation spectrum of the general- 1  
ized Maxwell model, we assume the cumulative relaxation spec- E1 = Hfrac ν10 + Hfrac ν20 ,
2
trum of the generalized Maxwell model forms a step-wise approx- 1  
Ej = Hfrac ν 0j+1 − Hfrac ν 0j−1 , 1 < j < N − 1,
2
 R. Xiao et al. / Mechanics of Materials 100 (2016) 148–153
Fig. 4. Comparison of the stress relaxation response of generalized Maxwell model and fractional Zener model, (a) τ = 0.1, α = 0.7, (b) τ = 0.1, α = 0.5.
Fig. 5. Strain as a function of time in the cyclic test.

N−1
EN = E neq − E j.
j relative difference is smaller than 0.2%. From the results, we can
We choose E eq = 1 MPa, E neq = 999 MPa to represent a typi-
cal thermoset polymer. The relaxation time is chosen as: τ = 0.1
or τ = 0.01, and fractional order is chosen as α = 0.7 or α = 0.5.
The shape of relaxation spectrum with the same fraction order re-
mains unchanged. Thus, in the following we only plot the relax-
ation spectra with relaxation time τ = 0.1 (Fig. 2a). As shown, the
relaxation spectrum of α = 0.5 is much broader than that of α =
0.7. We chose νmin
0 = 10−5 /s and νmax
0 = 105 /s for α = 0.7, while
νmin = 10 /s and νmax = 10 /s are chosen for α = 0.5. The num-
0 −8 0 8
ber of Maxwell elements N is chosen as 200 to guarantee the stor-
age modulus of the generalized Maxwell model showed excellent
agreement with that of fractional Zener model (Fig. 2b).
4. Results and discussions
As shown in Section 3, the two models predict almost identi-
cal dynamic response under certain frequency. In the following, we
perform a series of numerical simulations to compare the perfor-
mance of the two models under various loading conditions.
4.1. Numerical example I: constant strain rate
We first investigate the performance of the two models un-
der constant strain rate. In the simulation, the strain is ramped
to 0.1 under strain rate 0.0 0 01/s, 0.0 01/s or 0.01/s. As shown in
151
Fig. 3, the simulation results of two models agree excellently re-
garding all different values of τ and α . To quantitatively describe
the difference of the two models, we define the relative difference
= |σgene − σ f rac |/|σgene |, where σ gene and σ frac are the stress of
the generalized Maxwell model and fractional Zener model respec-
tively. In all the cases, the relative differences are smaller than
0.2% except for τ = 0.01, α = 0.7 and strain rate 0.0 0 01/s, where
the relative difference reaches 0.87%. These results suggest that the
two models show negligible difference in describing the stress re-
sponse at constant strain rate.
4.2. Numerical example II: stress relaxation
Following the real experimental setup and the discussion in
Di Paola et al. (2014), the stress relaxation test is performed by
an initial linear ramp and then the strain remains constant. Specif-
ically, the strain is increased to a certain level (0.01, 0.03 or 0.05)
at strain rate 0.01/s and held at that level for 60 s to allow stress
relaxation. Fig. 4 plots the stress response of the two models with
(12) different fractional order α = 0.5 and α = 0.7. In all the cases, the
see that a smaller fractional order results in a slower stress relax-
ation behavior indicating the relaxation occurs at a broader distri-
bution of relaxation time.
4.3. Numerical example III: cyclic test
In the cyclic test, the strain is increased to 0.05 at strain rate
0.01/s, then decreased to 0 at the same stain rate and held for
5 s, while this process is repeated four times as shown in Fig. 5.
Fig. 6 plots the corresponding stress response of the cyclic test as
a function of time. As shown, the stress response of the two mod-
els shows good agreement even after several cycles. The relative
difference does not show any increase with increasing cycles.
4.4. Numerical example IV: random loading test
We further simulate the stress response under a random load-
ing condition. In the simulation, the time step is chosen as t =
0.01. At each time step, the strain is increased by a random num-
ber between –0.0 0 01 and 0.0 0 01. Fig. 7a shows the strain as a
function of time during one random loading condition. The cor-
responding stress response of the two models with τ = 0.1 and
α = 0.7 is plotted in Fig. 7b and c, which shows identical pattern.
The simulation results of the last 80 steps are plotted in Fig. 7d.
As shown, the two models show excellent agreement. All of the
above results have demonstrated the two models are quantitatively
equivalent at various loading conditions.
152 R. Xiao et al. / Mechanics of Materials 100 (2016) 148–153

Fig. 6. Comparison of the stress response of generalized Maxwell model and fractional Zener model in the cyclic test, (a) τ = 0.1, α = 0.7, (b) τ = 0.1, α = 0.5.

Fig. 7. The random loading test: (a) strain as a function of time, (b) stress response of generalized Maxwell model, (c) stress response of the fractional Zener model,
(d) comparison the stress response of the two models.

4.5. General remarks Kelvin model and the Maxwell model, fail to describe the long-
term relaxation response (Xiao et al., 2016). To accurately de-
From Section 2, we can see that the classic dashpot is a lo- scribe the real materials behaviors, an excess number of param-
cal component while the fractional dashpot is inherently a global
eters need to be determined, which could be time-consuming. The
component. There should be fundamental differences between the
fractional models have the potential to overcome this disadvan-
two models. However, our results clearly demonstrate that the two tages, which require fewer parameters to fully describe the ex-
models show equivalences in various loading conditions. We are
perimental observations. In this work, we clearly show that the
not quite clear about the physical mechanism behind this equiva-
three parameter fractional Zener model has the ability to represent
lence. However, this finding can been extremely useful for the en-
a similar response of generalized Maxwell model with multiple
gineering applications.
parameters.
Viscoelastic materials, especially polymers, usually exhibit
However, the fractional models exhibit nonlocal properties,
broad relaxation spectrum, which requires long time to reach
which need take into account all the previous deformation
the equilibrium state. The simple rheological models, such as the
 R. Xiao et al. / Mechanics of Materials 100 (2016) 148–153
history. The fractional viscoelastic models may be difficult to be
applied for long-term loading conditions. In this case, the classic
rheological models involving local calculations may be more prac-
tical to use. Especially, in this work we demonstrate that the gener-
alized Maxwell model and fractional Zener model are numerically
equivalent. To our knowledge, no existing work has demonstrated
this equivalence. Thus, this work could arouse the interest of re-
searchers to explore the equivalence or non-equivalence of using
rheological models and fractional derivative models. Based on the
excellent agreement of the two models, this work may also provide
some insight into the numerical algorithm for solving the fractional
differential equations.
For fractional model, the physical meaning of fractional order is
far from settled. From the Section 3, we can see the fractional or-
der of fractional Zener model has a clear physical meaning, which
represents the breadth of relaxation spectrum. A smaller fractional
order means the relaxation behavior occurs in a broader time re-
gion, which may directly relate with the microstructure of materi-
als. Thus, this work can also shed light on the physical significance
behind fractional viscoelastic models.
5. Conclusion
The generalized Maxwell model and fractional Zener model
have been widely used to model viscoelastic solids. The fractional
Zener model is based on fractional calculus and exhibits global cor-
relation, while the generalized Maxwell model exhibits only local
property. In this work, we have compared the performance of the
two models under various loading conditions. Our results clearly
demonstrate that the two models are quantitatively equivalent re-
garding all different values of relaxation time and fractional order.
This finding could potentially assist choosing the suitable model
for practical applications. We also find that a smaller fractional or-
der represents the relaxation occurring at a broader distribution
of time, which results in a slower relaxation behavior. This finding
can shed light on understanding the physical meaning of fractional
order.
Acknowledgments
This work is supported by the National Natural Science Founda-
tion of China (Grant Nos. 11502068, 11572112) and the Fundamen-
tal Research Funds for Central Universities, Hohai University (Grant
No. 2016B01414).
References
Arikoglu, A., 2014. A new fractional derivative model for linearly viscoelastic ma-
terials and parameter identification via genetic algorithms. Rheologica Acta 53
(3), 219–233.
Bagley, R.L., Torvik, P.J., 1986. On the fractional calculus model of viscoelastic behav-
ior. J. Rheol. (1978-present) 30 (1), 133–155.
Baleanu, D., Magin, R.L., Bhalekar, S., Daftardar-Gejji, V., 2015. Chaos in the frac-
tional order nonlinear bloch equation with delay. Commun. Nonlinear Sci. Nu-
mer. Simul. 25 (1), 41–49.
Cataldo, E., Di Lorenzo, S., Fiore, V., Maurici, M., Nicoletti, F., Pirrotta, A., Scaf-
faro, R., Valenza, A., 2015. Bending test for capturing the vivid behavior of giant
reeds, returned through a proper fractional visco-elastic model. Mech. Mater.
89, 159–168.
Christensen, R.M., 2003. Theory of Viscoelasticity. Dover Publications, Mineola, NY.
De Gennes, P.-G., 1979. Scaling Concepts in Polymer Physics. Cornell University
Press, Ithaca, New York, USA.
153
Del Nobile, M., Chillo, S., Mentana, A., Baiano, A., 2007. Use of the generalized
maxwell model for describing the stress relaxation behavior of solid-like foods.
J. Food Eng. 78 (3), 978–983.
Di Paola, M., Fiore, V., Pinnola, F.P., Valenza, A., 2014. On the influence of the initial
ramp for a correct definition of the parameters of fractional viscoelastic mate-
rials. Mech. Mater. 69 (1), 63–70.
Di Paola, M., Heuer, R., Pirrotta, A., 2013. Fractional visco-elastic euler–bernoulli
beam. Int. J. Solids Struct. 50 (22), 3505–3510.
Di Paola, M., Pirrotta, A., Valenza, A., 2011. Visco-elastic behavior through fractional
calculus: an easier method for best fitting experimental results. Mech. Mater.
43 (12), 799–806.
Doi, M., Edwards, S.F., 1988. The Theory of Polymer Dynamics, Vol. 73. Oxford Uni-
versity Press, Oxford, UK.
Fan, W., Jiang, X., Qi, H., 2015. Parameter estimation for the generalized fractional
element network zener model based on the bayesian method. Physica A Stat.
Mech. Appl. 427, 40–49.
Ferry, J.D., 1980. Viscoelastic Properties of Polymers. John Wiley and Sons, New
York, NY.
Haupt, P., Lion, A., Backhaus, E., 20 0 0. On the dynamic behaviour of polymers un-
der finite strains: constitutive modelling and identification of parameters. Int. J.
Solids Struct. 37 (26), 3633–3646.
Jiang, X., Qi, H., 2012. Thermal wave model of bioheat transfer with modified rie-
mann–liouville fractional derivative. J. Phys. A: Math. Theor. 45 (48), 485101.
Kaufman, J.D., Miller, G.J., Morgan, E.F., Klapperich, C.M., 2008. Time-dependent me-
chanical characterization of poly (2-hydroxyethyl methacrylate) hydrogels using
nanoindentation and unconfined compression. J. Mater. Res. 23 (05), 1472–1481.
Koeller, R., 1984. Applications of fractional calculus to the theory of viscoelasticity.
J. Appl. Mech. 51 (2), 299–307.
Kohandel, M., Sivaloganathan, S., Tenti, G., Darvish, K., 2005. Frequency dependence
of complex moduli of brain tissue using a fractional zener model. Phys. Med.
Biol. 50 (12), 2799.
Li, Y., Tang, S., Kröger, M., Liu, W.K., 2015. Molecular simulation guided constitutive
modeling on finite strain viscoelasticity of elastomers. J. Mech. Phys. Solids.
Mainardi, F., 2010. Fractional calculus and waves in linear viscoelasticity: an intro-
duction to mathematical models. World Scientific.
Nguyen, T.D., Yakacki, C.M., Brahmbhatt, P.D., Chambers, M.L., 2010. Modeling the
relaxation mechanisms of amorphous shape memory polymers. Adv. Mater. 22
(31), 3411–3423. Sp. Iss. SI
Rossikhin, Y.A., Shitikova, M.V., 2010. Application of fractional calculus for dynamic
problems of solid mechanics: novel trends and recent results. Appl. Mech. Rev.
63 (1), 010801.
Rouse Jr., E.P., 1953. A theory of the linear viscoelastic properties of dilute solutions
of coiling polymers. J. Chem. Phys. 21 (7), 1272–1280.
Schiessel, H., Metzler, R., Blumen, A., Nonnenmacher, T., 1995. Generalized viscoelas-
tic models: their fractional equations with solutions. J.Phys. A: Math. Gener. 28
(23), 6567.
Tarasov, V.E., 2014. Lattice model of fractional gradient and integral elastic-
ity: long-range interaction of grünwald–letnikov–riesz type. Mech. Mater. 70,
106–114.
Wang, H., Wang, K., Sircar, T., 2010. A direct o (nlog 2 n) finite difference method
for fractional diffusion equations. J. Comput. Phys. 229 (21), 8095–8104.
Wu, G.-C., 2012. Applications of the variational iteration method to fractional diffu-
sion equations: local versus nonlocal ones. Int. Rev. Chem. Eng. 4 (5), 505–510.
Wu, G.-C., Baleanu, D., 2015. Discrete chaos in fractional delayed logistic maps. Non-
linear Dyn. 80 (4), 1697–1703.
Xiao, R., Choi, J., Lakhera, N., Yakacki, C., Frick, C., Nguyen, T., 2013. Modeling the
glass transition of amorphous networks for shape-memory behavior. J. Mech.
Phys. Solids 61 (7), 1612–1635.
Xiao, R., Guo, J., Nguyen, T.D., 2015. Modeling the multiple shape memory effect and
temperature memory effect in amorphous polymers. RSC Adv. 5 (1), 416–423.
Xiao, R., Nguyen, T.D., 2015. An effective temperature theory for the nonequilibrium
behavior of amorphous polymers. J. Mech. Phys. Solids 82, 62–81.
Xiao, R., Yakacki, C.M., Guo, J., Frick, C.P., Nguyen, T.D., 2016. A predictive parameter
for the shape memory behavior of thermoplastic polymers. J. of Polym. Sci. Part
B: Polym. Phys. 54, 1405–1414.
Yu, B., Jiang, X., Qi, H., 2015. An inverse problem to estimate an unknown order of
a riemann–liouville fractional derivative for a fractional stokes’ first problem for
a heated generalized second grade fluid. Acta Mechanica Sinica 31 (2), 153–161.
Yu, K., Ge, Q., Qi, H.J., 2014. Reduced time as a unified parameter determining fixity
and free recovery of shape memory polymers. Nat. Commun. 5 (3066). doi:10.
1038/ncomms4066.
Zhao, X., Sun, Z.z., 2011. A box-type scheme for fractional sub-diffusion equation
with neumann boundary conditions. J. Comput. Phys. 230 (15), 6061–6074.