We apply the homotopy analysis method (HAM) for solving the
multipantograph equation. The analytical results have been obtained
in terms of convergent series with easily computable components. Several
examples are given to illustrate the efficiency and implementation
of the homotopy analysis method. Comparisons are made to confirm
the reliability of the homotopy analysis method.
1. Introduction
The delay differential equationy′(t)=λy(t)+∑i=1kμiy(fi(t)),t>0,y(0)=y0, where λ,μ1,μ2,…,μk,y0∈ℂ, has been studied by numerous authors (e.g.,
[1–8]). Second-order versions of this equation have also
been studied (e.g., [9, 10]). The enduring interest in this equation is due partially
to the number of applications it has found such as a current collection system
for an electric locomotive, cell growth models, biology, economy, control, and
electrodynamics (e.g., [10–13]). The focus of most of the
studies made in the complex plane (e.g., [12, 14]) was on solutions on the real line for either the
retarded case 0<q<1 or the advanced case q>1.
In 1999, Qiu et al. [15] have studied the delay
equationy′(t)=λy(t)+∑i=1kμiy(qit),y(0)=y0,where 0<qk<qk−1<⋯<q1<1 and λ,μ1,μ2,…,μk,y0∈ℂ, by transforming the proportional delay into
the constant delay. They got the sufficient condition of asymptotic stability
for the analytic solution, that is,Reλ<0,∑i=1k|μi|<−Reλ.
Liu and Li in [16, 17] proved the existence and
uniqueness of analytic solution of (1.2) for any λ,μ1,μ2,…,μk,y0∈ℂ,
and the analytic solution is asymptotically stable ifReλ<0,∑i=1k|μi|<|λ|.
In [17–19] the Dirichlet series solution of (1.2) is constructed,
and the sufficient condition of the asymptotic stability for the analytic
solution is obtained. It is proved that the θ-methods with a variable stepsize are
asymptotically stable if 1/2<θ≤1.
It is well known that for the multipantograph equationy′(t)=λy(t)+∑i=1kμiy(qit)+f(t),0<t<T,y(0)=α,where 0<qk<qk−1<⋯<q1<1, the collocation solution associated with
the mth degree collocation
polynomial possesses the optimal superconvergence order 2m+1 at the first step t=h,
provided that the collocation m parameters are properly chosen in (0,1) (e.g., [5] for f(t)=0,
and [20] for f(t)≠0).
Ishiwata and Muroya [21] proposed a piecewise (2m,m)-rational approximation with “quasiuniform meshes” which corresponds to the mth collocation method, and established the
global error analysis of O(h2m) on successive mesh points. This method is more
useful than the known collocation method when solving (1.5) in case that a long
time integration is needed, that is, if T is large, then the number of steps in the
method is less than that of the collocation method. Collocation method is
useful for computation, but in these mesh divisions, there are problems. For
example, if the end point t=T is larger, then the mesh size near the first
mesh point becomes too small, compared with the mesh size near the end point.
This implies that the total computational cost is higher (see also [22–25].)
In this paper, and in order to overcome such problems,
we propose an analytic solution of (1.5) by the HAM addressed in [26–36]. The HAM is
based on the homotopy, a basic concept in topology. The auxiliary parameter h is introduced to construct the so-called
zero-order deformation equation. Thus, unlike all previous analytic techniques,
the HAM provides us with a family of solution expressions in auxiliary parameter h.
As a result, the convergence region and rate of solution series are dependent
upon the auxiliary parameter h and thus can be greatly enlarged by means of
choosing a proper value of h.
This provides us with a convenient way to adjust and control convergence region
and rate of solution series given by the HAM.
2. Description of the Method
In order to obtain an analytic solution of the delay
differential equation (1.5), the HAM is employed. Consider the operator N,N[y(t)]=∂y(t)∂t−λy(t)−∑i=1kμiy(qit)−f(t)=0,where y(t) is unknown function and t the independent variable. Let y0(t) denote an initial guess of the exact solution y(t) that satisfies y0(0)=α, h≠0 an auxiliary parameter, H(t)≠0 an auxiliary function, and L an auxiliary linear operator with the property L[y(t)]=0 when y(t)=0.
Then using q∈[0,1] as an embedding parameter, we construct such a
homotopy:(1−q)L[ϕ(t;q)−y0(t)]−qhH(t)N[ϕ(t;q)]=H^[ϕ(t;q);y0(t),H(t),h,q].
It should be emphasized that we have great freedom to
choose the initial guess y0(t),
the auxiliary linear operator L,
the nonzero auxiliary parameter h,
and the auxiliary function H(t).
Enforcing the homotopy (2.2) to be zero, that is,H^[ϕ(t;q);y0(t),H(t),h,q]=0,we have the so-called zero-order
deformation equation(1−q)L[ϕ(t;q)−y0(t)]=qhH(t)N[ϕ(t;q)].
When q=0,
the zero-order deformation equation (2.4) becomesϕ(t;0)=y0(t)and when q=1,
since h≠0 and H(t)≠0,
the zero-order deformation equation (2.4) is equivalent toϕ(t;1)=y(t).
Thus, according to (2.5) and (2.6), as the embedding
parameter q increases from 0 to 1, ϕ(t;q) varies continuously from the initial
approximation y0(t) to the exact solution y(t).
Such a kind of continuous variation is called deformation in homotopy.
By Taylor's theorem, ϕ(t;q) can be expanded in a power series of q as follows:ϕ(t;q)=y0(t)+∑m=1∞ym(t)qm,whereym(t)=1m!∂mϕ(t;q)∂qm|q=0.
If the initial guess y0(t), the auxiliary linear parameter L,
the nonzero auxiliary parameter h,
and the auxiliary function H(t) are properly chosen, so that the power series
(2.7) of ϕ(t;q) converges at q=1. Then, we have under these assumptions the solution
seriesy(t)=ϕ(t;1)=∑m=0∞ym(t).
For brevity, define the vectory→n(t)={y0(t),y1(t),y2(t),…,yn(t)}.
According to the definition (2.8), the governing
equation of ym(t) can be derived from the zero-order deformation
equation (2.4). Differentiating the zero-order deformation equation (2.4) m times with respect to q and then dividing by m! and finally setting q=0,
we have the so-called mth-order deformation
equationL[ym(t)−χmym−1(t)]=hH(t)ℜm(y→m−1(t)),ym(0)=0,whereℜm(y→m−1(t))=1(m−1)!∂m−1N[ϕ(t;q)]∂qm−1|q=0=ym−1′(t)−λym−1(t)−∑i=1kμiym−1(qit)−(1−χm)f(t),χm={0,m≤11,m>1.
3. ConvergenceTheorem 3.1.
As long as the series (2.9) converges, it must be the exact solution of
the multipantograph equation (1.5).
Proof.
If the
series (2.9) converges, we can writeS(t)=∑m=0∞ym(t)and it holds thatlimm→∞ym(t)=0.
We can verify that∑m=1n[ym(t)−χmym−1(t)]=y1+(y2−y1)+⋯+(yn−yn−1)=yn(t),which gives us, according to
(3.2),∑m=1∞[ym(t)−χmym−1(t)]=limn→∞yn(t)=0.
Furthermore, using (3.3) and the definition of the
linear operator L, we have∑m=1∞L[ym(t)−χmym−1(t)]=L[∑m=1∞[ym(t)−χmym−1(t)]]=0.
According to (2.11), we can obtain that∑m=1∞L[ym(t)−χmym−1(t)]=hH(t)∑m=1∞ℜm(y→m−1(t))=0,which gives, since h≠0 and H(t)≠0,∑m=1∞ℜm(y→m−1(t))=0.
By the definition (2.12) of ℜm(y→m−1(t)),
it holds that∑m=1∞ℜm(y→m−1(t))=∑m=1∞[ym−1′(t)−λym−1(t)−∑i=1kμiym−1(qit)−(1−χm)f(t)]=∑m=0∞ym′(t)−λ∑m=0∞ym(t)−∑m=0∞∑i=1kμiyn(qit)−f(t)=S'(t)−λS(t)−∑i=1kμiS(qit)−f(t).
From (3.7) and (3.8), we haveS′(t)=λS(t)+∑i=1kμiS(qit)+f(t)and, moreover, with the help of
(2.11), it holds thatS(0)=∑m=0∞ym(0)=y0(0)+∑m=1∞ym(0)=y0(0)=α.
In view of (3.9) and (3.10), S(t) must be the exact solution of (1.5).
4. Examples
The HAM provides an analytical solution in terms of an
infinite power series. However, there is a practical need to evaluate this
solution, and to obtain numerical values from the infinite power series. The
consequent series truncation, as well as the practical procedure conducted to
accomplish this task, transforms the otherwise analytical results into an exact
solution, which is evaluated to a finite degree of accuracy. In order to
investigate the accuracy of the HAM solution with a finite number of terms,
three examples were solved. The HAM
results were compared with the exact solutions. The impact of the term numbers
in the series solution and truncation process was assessed by evaluating the
HAM results for different terms in the series. By increasing the number of the
HAM terms, the percentage of error decreases. It is also observed that the HAM
results with 10 terms have acceptable accuracy compared to the exact solutions.
Therefore, it may be concluded that the use of 10 terms in the series yields
accurate results with HAM solution sufficiently. MATLAB 7 is used to carry out
the computations.
Defining that L[ϕ(t;q)]=∂ϕ(t;q)/∂t, with the
property L[C]=0, where C is the integral constant and using H(t)=1,
the mth-order deformation equations (2.11) for m≥1 becomesym(t)=χmym−1(t)+h∫0t[ym−1′(τ)−λym−1(τ)−∑i=1kμiym−1(qiτ)−(1−χm)f(τ)]dτ.
Example 4.1.
We
consider the following pantograph differential equation:y′(t)=−y(t)+14y(12t)−14e−0.5t,y(0)=1. The exact solution is y(t)=e−t.
Note that we still have freedom to choose the
auxiliary parameter h.
To investigate the influence of h
on the solution series (2.9), we can consider
the convergence of some related series such as y′(0),y′′(0),y′′′(0), and so on. However, y′′(0) is dependent of h.
Let Rh denote a set of all possible values of h by means of which the corresponding series of y′′(0) converges. According to Theorem 3.1, for each h∈Rh,
the corresponding series of y′′(0) converges to the same result. The curve y′′(0) versus h
contains a horizontal line segment above the
the valid region Rh. We call such a kind of curve the h-curve [33], which clearly indicates the the valid region Rh of a solution series. The so-called h-curve of y′′(0) is as shown in Figure 1. From Figure 1 it is clear
that the series of y′′(0) is convergent when −2≤h≤0. Using h=−1,
we have from (2.9) and (4.1) that the ten terms
approximate solution obtained by HAM are ∑m=010ym(t)=1−t+12t2−16t3+124t4−1120t5+1720t6−15040t7+140320t8−1362880t9+6.3×10−8t10≃∑k=010(−1)ktkk!. We see that HAM solution is
very close to the exact solution. It may be concluded that the use of 10 terms
in the homotopy series yields accurate results.
The h-curve of y′′(0). Solid line: 10th-order approximation of y′′(0).
Example 4.2.
Next, we consider the nonhomogeneous delay equation y′(t)=−y(t)+12y(12t)+cost+sint−12sin12t,0≤t≤2π,y(0)=0. By means of the h-curve, it is reasonable to choose h=−1.5.
We have for t>0
the ten terms approximate solution obtained by
HAM as follows: ∑m=010ym(t)=t−16t3+1120t5−15040t7+1362880t9+1.6×10−7t10≃∑m=09(−1)k(2k+1)!t2k+1.
In view of (4.5), we can conclude that the
exact solution is y(t)=sint. Ishiwata and Muroya [21] proposed a piecewise (2m,m)-rational approximation Q2m,m(t)
with “quasiuniform meshes” which
corresponds to the mth collocation method. For m=2,
and h=2(6+n),n=0,1,…,4, the errors e(h)=|Q4,2(h)−y(h)| at the first mesh point t1=h
are shown in the third column of Table 1. In
Table 1, The accuracy of the HAM is examined by comparing (4.5) with the
available exact and the (2m,m)-rational approximation method.
Comparison of the results of the HAM and the (2m,m)-rational approximation.
n
HAM
(2m,m)-rational approximation
0
0
3.8391471⋯E−07
1
6.93⋯E−18
2.613675⋯E−08
2
3.46⋯E−18
1.70118⋯E−09
3
1.23⋯E−31
1.0844⋯E−10
4
4.04⋯E−36
6.83⋯E−12
Example 4.3.
In the last example, we consider the
pantograph equation y′(t)=−y(t)−e−0.5tsin(0.5t)y(0.5t)−2e−0.75tcos(0.5t)sin(0.25t)y(0.25t),y(0)=1. The exact solution is y(t)=e−tcost.
By means of the h-curve, it is reasonable to choose h=−1. We
have for t>0,∑m=010ym(t)=1−t+13t3−16t4+130t5−1630t7+12520t8−122680t9−13628800t10.The first nine terms of the
series (4.7) are coinciding with the first nine terms of the Taylor series of e−tcost. Figure 2 shows plots of ten and twenty terms
approximation of y(t).
Plots of ten “**” and twenty “oo” terms approximations for y(t) “−” versus t.
5. Discussion and Conclusion
In this paper, the HAM was employed to solve the
multipantograph differential equation. Unlike the traditional methods, the
solutions here are given in series form. The approximate solution to the equation
was computed with no need for special transformations, linearization, or
discretization. It was shown that the HAM solutions are very close to the exact
solutions. It may be concluded that the use of a few terms in the series yields
accurate results with HAM solution sufficiently. HAM is a powerful tool for
solving analytically nonlinear equations.
FredricksonP. O.UradbeM.Dirichlet series solution for certain functional differential equations1971243Berlin, GermanySpringer247254Lecture Notes in MathematicsFoxL.MayersD. F.OckendonJ. R.TaylerA. B.On a functional differential equation19718271307MR0301330ZBL0251.34045GrossF.On a remark of Utz19677411071108MR022601110.2307/2313627ZBL0186.21402IserlesA.LiuY.On neutral functional-differential equations with proportional delays199720717395MR143302610.1006/jmaa.1997.5262ZBL0873.34066IshiwataE.On the attainable order of collocation methods for the neutral functional-differential equations with proportional delays2000643207222MR176705410.1007/s006070050044ZBL0955.65098ObergR. J.Local theory of complex functional differential equations1971161269281MR028202610.2307/1995942ZBL0255.34067UtzW. R.The equation f′(x)=af(g(x))196571113810.1090/S0002-9904-1965-11252-6van BruntB.MarshallJ. C.WakeG. C.Holomorphic solutions to pantograph type equations with neutral fixed points20042952557569MR207203210.1016/j.jmaa.2004.03.019ZBL1060.34058van BruntB.WakeG. C.KimH. K.On a singular Sturm-Liouville problem involving an advanced functional differential equation2001126625644MR187734410.1017/S0956792501004624ZBL1017.34068WakeG. C.CooperS.KimH.-K.van-BruntB.Functional differential equations for cell-growth models with dispersion200044561573MR1843770ZBL1084.34545DerfelG.IserlesA.The pantograph equation in the complex plane19972131117132MR146936410.1006/jmaa.1997.5483ZBL0891.34072IserlesA.On the generalized pantograph functional-differential equation199341138MR120841810.1017/S0956792500000966ZBL0767.34054OckendonJ. R.TaylerA. B.The dynamics of a current collection system for an electric locomotive19713221551447468BuhmannM.IserlesA.Stability of the discretized pantograph differential equation199360202575589MR117670710.2307/2153103ZBL0774.34057QiuL.MitsuiT.KuangJ.-X.The numerical stability of the θ-method for delay differential equations with many variable delays1999175523532MR1712643ZBL0942.65087LiD.LiuM. Z.The properties of exact solution of multi-pantograph delay differential equation20003113LiuM. Z.LiD.Properties of analytic solution and numerical solution of multi-pantograph equation20041553853871MR207869610.1016/j.amc.2003.07.017ZBL1059.65060BellenA.GuglielmiN.TorelliL.Asymptotic stability properties of θ-methods for the pantograph equation1997242-3279293MR146472910.1016/S0168-9274(97)00026-3ZBL0878.65064BellenA.ZennaroM.2003Oxford, UKOxford University Pressxiv+395Numerical Mathematics and Scientific ComputationMR1997488ZBL1038.65058MuroyaY.IshiwataE.BrunnerH.On the attainable order of collocation methods for pantograph integro-differential equations20031521-2347366MR199130110.1016/S0377-0427(02)00716-1ZBL1023.65146IshiwataE.MuroyaY.Rational approximation method for delay differential equations with proportional delay20071872741747MR232307910.1016/j.amc.2006.08.086ZBL1117.65105BrunnerH.200415Cambridge, UKCambridge University Pressxiv+597Cambridge Monographs on Applied and Computational MathematicsMR2128285ZBL1059.65122BrunnerH.HuQ.LinQ.Geometric meshes in collocation methods for Volterra integral equations with proportional delays2001214783798MR186741810.1093/imanum/21.4.783ZBL1014.65143BellenA.Preservation of superconvergence in the numerical integration of delay differential equations with proportional delay2002224529536MR193651810.1093/imanum/22.4.529ZBL1031.65089TakamaN.MuroyaY.IshiwataE.On the attainable order of collocation methods for delay differential equations with proportional delay2000402374394MR176574210.1023/A:1022351309662ZBL0965.65101AbbasZ.HayatT.Radiation effects on MHD flow in a porous space2008515-61024103310.1016/j.ijheatmasstransfer.2007.05.031AbbasbandyS.Soliton solutions for the Fitzhugh-Nagumo equation with the homotopy analysis method200832122706271410.1016/j.apm.2007.09.019HayatT.SajidM.On analytic solution for thin film flow of a fourth grade fluid down a vertical cylinder20073614-531632210.1016/j.physleta.2006.09.060IncM.Application of homotopy analysis method for fin efficiency of convective straight fins with temperature-dependent thermal conductivity200879218920010.1016/j.matcom.2007.11.009IncM.On exact solution of Laplace equation with Dirichlet and Neumann boundary conditions by the homotopy analysis method20073655-6412415MR230878110.1016/j.physleta.2007.01.069IncM.On numerical solution of Burgers' equation by homotopy analysis method20083724356360MR238182010.1016/j.physleta.2007.07.057SajidM.HayatT.AsgharS.Comparison between the HAM and HPM solutions of thin film flows of non-Newtonian fluids on a moving belt2007501-22735MR234492910.1007/s11071-006-9140-yLiaoS.20042Boca Raton, Fla, USAChapman & Hall/CRC Pressxii+322CRC Series: Modern Mechanics and MathematicsMR2058313ZBL1051.76001LiaoS.On the homotopy analysis method for nonlinear problems20041472499513MR201258910.1016/S0096-3003(02)00790-7ZBL1086.35005LiaoS.An explicit analytic solution to the Thomas-Fermi equation20031442-3495506MR199408710.1016/S0096-3003(02)00423-XZBL1034.34005TanY.AbbasbandyS.Homotopy analysis method for quadratic Riccati differential equation200813353954610.1016/j.cnsns.2006.06.006ZBL1132.34305