The third-order conditional Lie–Bäcklund symmetries of nonlinear reaction-diffusion equations are constructed due to the method of linear determining equations. As a consequence, the exact solutions of the resulting equations are derived due to the compatibility of the governing equations and the admitted differential constraints, which are resting on the characteristic of the admitted conditional Lie–Bäcklund symmetries to be zero.
Key Scientific and Technological Project of Higher Education of Henan Province16A1100321. Introduction
The conditional Lie–Bäcklund symmetry (CLBS) was independently proposed by Zhdanov [1] and Fokas and Liu [2]. Exact solutions possessing the form of nonlinear separation of variables [3] and exact solutions defined on the invariant subspaces [4] for nonlinear heat conductivity equations were constructed due to the corresponding symmetry reductions in [1], which cannot be obtained within the framework of Lie’s classical symmetry [5], conditional symmetry [6, 7], and Lie–Bäcklund symmetry [8]. Multishock and multisoliton solutions of nonintegrable equations were obtained by using CLBS in [2], which are regarded as remnants of integrability for the considered nonintegrable equations.
As a consequence, CLBSs were applied to a number of particular subclasses of evolution equations [9, 10]. It has been noted that CLBS is very effective in studying classifications and reductions of diffusion equations [11–16]. CLBS can also be used to consider the reductions of initial value problems for nonlinear evolution equations [17–19]. CLBS for evolution systems were studied in [20, 21].
The procedure for determining whether or not a given Lie–Bäcklund symmetry is conditionally invariant of the considered equations is straightforward; however, the determination of the most general CLBS admitted by a given differential equation is a very difficult, if not impossible, problem. CLBSs are generalizations of conditional symmetries, in the same way that Lie–Bäcklund symmetries are generalizations of Lie’s classical symmetry. Therefore the procedure for computing CLBSs is about the same as that for the conditional symmetries. We need to determine the form of CLBS presumably. It is proved that CLBSs related to separation of variables [13], sign-invariants [14], and invariant subspaces [15, 16] are powerful to study different types of diffusion equations. Are there other choices of CLBS which may yield new interesting results for these considered equations?
Once the CLBS and the governing equation are determined, exact solutions of the discussed equations can be easily constructed due to the compatibility of the considered equations and the additional differential constraints obtained by setting the characteristic of the corresponding CLBS to be zero. Thus, CLBS can be reformulated by using the technicalities of the method of differential constraints.
The key idea of the method of differential constraint was proposed by Yanenko in [22]. The general formulation of this method requires that the original system of partial differential equations and the additional differential constraints satisfy some conditions of compatibility. The problem of finding all differential constraints compatible with certain equations can be more complicated than the investigation of solving the original equations. Hence, it is better to content oneself with finding constraints in some fixed classes, and these classes must be chosen using additional considerations.
In practice, differential constraints related to separation of variables [23] and invariant subspaces [24] are powerful to seek for reductions of different types of evolution equations. Olver and Rosenau [25, 26], Olver [27], Kaptsov [28], and Levi and Winternitz [29] showed that many reductions including Bluman and Cole’s nonclassical method [6], partial invariance method [30], Clarkson and Kruskal’s direct method [31], and Galaktionov’s higher-order direct method [3] can be understood within the framework of differential constraints.
B-determining equations [32] and linear determining equations [33, 34] were both proposed to construct differential constraints. The method of linear determining equations with free parameters is much easier to illustrate than the one of B-determining equations with undetermined functions. The linear determining equations are more general than the classical determining equations for Lie’s generators [30].
In this paper, we will consider the general third-order CLBS with the characteristic(1)η=u3+gt,x,u,u1,u2of the nonlinear reaction-diffusion equations(2)ut=ukuxx+Quby using the linear determining equations. Equation (2) is widely used as mathematical models of heat conduction and diffusion, filtration of gases and fluids in porous media, process of chemical kinetics, and biological process. It is noted that uj=∂ju/∂xj(j=1,2,3) in (1).
There are numerous works aimed to study symmetry classifications and reductions of (2). CLBS was also used to consider the variant forms of (2). Three types of CLBS were applied to study such equations. The CLBS with the characteristic(3)η=fuxtwas presented to study functionally separable solutions (4)fu=ϕt+ψx of (2) in [13]. The CLBS with the characteristic(5)η=u2+Huu12+Guu1+Fuwas proposed to study the variant forms of (2) in [11, 12], which can give symmetry interpretation for first-order sign-invariants of the considered equations. The CLBS with the characteristic(6)η=fun+a1xfun-1+a2xfun-2+⋯+anxfuwas proposed to study the various forms of (2) in [15, 16]. The corresponding linearized form(7)σ=un+a1xun-1+a2xun-2+⋯+anxuof (6) is related to the invariant subspace (8)W=Lf1x,f2x,…,fnx,which is the right solution space of the linear ODE σ=0.
The forms of CLBS in [11–16] are all determined in advance. Thus the corresponding results are of course restricted to this ansatz. The third-order differential constraints (1) of the nonlinear diffusion equation ut=(ukux)x are considered in [33]. The differential constraints(9)η=un+gt,x,u,u1,u2,…,un-1=0with n=2,3 of (2) are discussed in [34]. However, the study there was worked out for only several special cases. Here we will extend the work there to develop a complete classification for third-order CLBS with the form (1) of (2) by applying the method of linear determining equations. New types of CLBS which are not included in the form of (3), (5), and (6) will be derived and therefore a huge number of new exact solutions of (2) will be constructed due to the compatibility of the governing equations and corresponding resulting differential constraints.
2. General Statement
Consider a nonlinear evolution equation(10)ut=Ft,x,u,u1,u2,…,ukwith Lie–Bäcklund vector field (LBVF)(11)V=η∂∂u+Dxη∂∂u1+Dtη∂∂ut+Dx2η∂∂u2+DxDtη∂∂u1t+⋯,where η=η(x,t,u,u1,ut,utt,u1t,…) is the characteristic of LBVF (11) and the total differentiation operators, respectively, denote(12)Dt=∂∂t+ut∂∂u+utt∂∂ut+u1t∂∂u1+⋯,Dx=∂∂x+u1∂∂u+u1t∂∂ut+u2∂∂u1+⋯.
Definition 1 (see [5]).
The evolutionary vector field (11) is said to be a Lie–Bäcklund symmetry of the evolution equation (10) if the condition (13)Vut-FM=0holds, where M denotes the set of all differential consequences of (10).
Definition 2 (see [1, 2]).
The evolutionary vector field (11) is said to be a CLBS of (10) if the following condition(14)Vut-FM∩Lx=0holds, where Lx denotes the set of all differential consequences of equation η=0 with respect to x.
It is noted that we can exclude all derivatives with respect to t on the set of (10) and thus obtain the vector field (11) with η of the form(15)η=ηt,x,u,u1,u2,…,un.
A direct computation of (14) yields (16)DtηM∩Lx=0.On the other hand, we can derive that (17)DtηM∩Lx=DtηMx∩Lx,where Mx denotes the set of all differential consequences of (10) with respect to x. Thus, we have that the condition(18)DtηMx∩Lx=0holds, which is exactly the sufficient condition [28, 35] for which η=0 is a differential constraint of the evolution equation (10). Thus η=0 defines an invariant manifold of the equation (10), which is right corresponding to the symmetry reductions due to the compatibility of η=0 and the considered (10). There is an equivalence relation between CLBS and differential constraint for the evolution equation (10).
Kaptsov [33, 34] proposed a method for finding differential constraints, which uses linear determining equations. The definition of linear determining equation was first given in [33], which is listed as follows.
Definition 3 (see [33, 34]).
A linear determining equation corresponding to differential constraints η(t,x,u,u1,u2,…,un)=0 of the evolution equation (10) is an equation of the form(19)DtηMx=∑i=0n∑j=0ibijDxi-jFun-jDxn-iη,where bij are some constants.
3. CLBS of (2)
As stated in [34], the linear determining equation for the differential constraints η=0 of (2) is right(20)DtηEx=ukDx2η+b1kuk-1u1Dxη+b3kuk-1u2+b2kk-1uk-2u12+b4fuη,where b1,b2,b3,b4∈R and Ex is the differential consequences of (2) with respect to x. It is easy to see that (2) admits LBVF (11) if the condition (20) holds.
Substituting (1) into (20), the left-hand side and the right-hand side are, respectively, simplified as (21)DtηEx=uku5+uk-1ugu2+5ku1u4+ukgu1+4kuk-1u1gu2+10kuk-1u2+10kk-1uk-2u12+Quu3+kk-1k-2uk-3u14+3kuk-1u22+6kk-1uk-2u2u12+u12Quu+u2Qugu2+u1Qu+kk-1uk-2u13+3kuk-1u2u1gu1+uku2+kuk-1u12+Qugu+gt+u13Quuu+3u2u1Quu+kk-1k-2k-3uk-4u15+10kk-1k-2uk-3u2u13+15kk-1uk-2u22u1,ukDx2η+b1kuk-1u1Dxη+b3kuk-1u2+b2kk-1uk-2u12+b4fuη=uku5+uk-1ugu2+b1ku1u4+ukgu2u2u32+b4Qu+uk2u1guu2+2gxu2+2u2gu1u2+gu1+b1kuk-1u1gu2+b2kk-1uk-2u12+b3kuk-1u2u3+uku22gu1u1+2u1u2guu1+u12guu+2u2gxu1+2u1gxu+gxx+u2gu+b1kuk-1u1u2gu1+u1gu+gx+b3kuk-1u2+b2kk-1uk-2u12+b4Qug,which can be regarded as two polynomials with respect to u5,u4, and u3. It is noted that the subscripts here denote the differentiation with respect to the indicated variables.
Considering the coefficients of u4 and u32, we obtain (22)b1=5,gu2u2=0.So g can be represented by (23)g=ht,x,u,u1u2+ft,x,u,u1. Here h and f are both functions of t,x,u, and u1. Collecting the coefficients of u3, we have (24)2ukhu1+b3-10kuk-1u2+2uku1hu+hx+kuk-1u1h+kk-1b2-10uk-2u12+b4-1Qu=0.Thus, we know that h must satisfy(25)2ukhu1+b3-10kuk-1=0,(26)2uku1hu+hx+kuk-1u1h+kk-1b2-10uk-2u12+b4-1Qu=0.Solving (25), we obtain (27)h=-kb3-102uu1+at,x,uand a(t,x,u) is the function of t,x, and u. As a consequence, (26) can be simplified as (28)k5k-kb2+12kb3+b2-b3uk-2u12-2uau+kuk-1au1-2uax+1-b4Qu=0,which implies (29)k5k-kb2+12kb3+b2-b3=0,2uau+kuk-1a=0,-2uax+1-b4Qu=0.
Solving the linear ordinary differential equation about a, we get a=C(t,x)u-k/2. The simplification of the above equation about Q gives (30)1-b4Qu-2uk/2Cx=0. We know that Cx=d1 must hold and then the above equations about Q become (31)1-b4Qu-2d1uk/2=0.Solving this ordinary differential equation, we know that there are three subcases for further discussion, which are listed as
(1) (32)b4=1,d1=0;
(2) (33)k=-2,Q=2d1lnu1-b4+d3;
(3) (34)Q=2d1u1+k/21-b41+k/2+d3,b4≠1,k≠-2.
For these three different cases, a further identical vanishing of (20) will finally determine the form of g in (1). The detailed procedure is omitted here. We just present the corresponding results in Table 1, where di(i=1,2,3,4,5) are arbitrary constants.
In this section, we construct exact solutions of (2) due to the compatibility of differential constraints η=0 and the governing equation. The equivalence relation between CLBSs and differential constraints for the evolution equation (10) is first rigorously proved in [36], where it is also stated that finding of the whole set of CLBSs of an evolution equation is a no-go problem. In fact, the general forms CLBSs η of (2) are completely determined by solving the linear system of determining equations about η, which are listed in Table 1. Solving η=0, we can obtain u as a function of x with t-dependent undetermined functions. Substituting the form of u into (2) will yield that the t-dependent coefficients in (2) satisfy three-dimensional dynamical system. Here we give several examples to illustrate the reduction procedure.
Example 1.
Equation (35)ut=ukuxx+d1u+d2u1-k+k+1d3k2u1+kadmits the CLBS (36)η=u3+3k-1uu1u2+k-1k-2u2u13+d3u1.
The corresponding solutions are given as follows.
For d3>0, (37)ux,t=αt+βtsind3x+γtcosd3x1/k,
where α(t),β(t), and γ(t) satisfy three-dimensional dynamical system (38)α′=k+1d3kα2+d3kβ2+d3kγ2+kd1α+kd2,β′=k+2d3kαβ+kd1β,γ′=k+2d3kαγ+kd1γ.
For d3<0, (39)ux,t=αt+βtsinh-d3x+γtcosh-d3x1/k,
where α(t),β(t), and γ(t) satisfy three-dimensional dynamical system (40)α′=k+1d3kα2-d3kβ2+d3kγ2+kd1α+kd2,β′=k+2d3kαβ+kd1β,γ′=k+2d3kαγ+kd1γ.
For d3=0, (41)ux,t=αt+βtx+γtx21/k,
where α(t),β(t), and γ(t) satisfy three-dimensional dynamical system (42)α′=1kβ2+2αγ+kd1α+kd2,β′=2k+2kβγ+kd1β,γ′=2k+2kγ2+kd1γ.
Example 2.
Equation(43)ut=u-4/3uxx+d1u+d2u5/3admits the CLBS (44)η=u3-5uu1u2+409u2u13.
The corresponding solutions of (43) are given by (45)ux,t=αt+βtx+γtx2-3/2,where α(t),β(t), and γ(t) satisfy three-dimensional dynamical system (46)α′=2α2γ-12αβ2-23d1α-23d2,β′=-12β3+2αβγ-23d1β,γ′=2αγ2-12β2γ-23d1γ.
Example 3.
Equation(47)ut=u-1uxx+d1u2+d2uadmits the CLBS (48)η=u3-6uu1u2+6u2u13.
The corresponding solutions are given by (49)ux,t=-βtαt-x+c12,where α(t),β(t), and γ(t) are listed as follows.
For d2≠0, (50)αt=2d1lnc2d2expd2t-2d24c22+d1expd2td23c2+c3c2d2-2exp-d2t,βt=c2expd1t-2d2.
For d2=0, (51)αt=12d1ln2t+c2+c32t+c22,βt=2t+c2.
Example 4.
Equation(52)ut=u-1uxxadmits the CLBS (53)η=u3-4uu1u2+3u2u13+d1u2u1.
The corresponding solutions are listed as follows.
For d1>0, (54)ux,t=-4d1c12exp-c12/d1x+γt-exp-2c12/d1x+γt+4d13exp-c12/d1x+γtβt-4d14αt,
where the unknown functions about t are, respectively, given as (55)αt=c121+tan2c1t+c2,βt=-c1d1tanc1t+c2,γt=c3-12d1c12ln1+tan2c1t+c2.
For d1<0, (56)ux,t=4d1c12exp--c12/d1x+γt-exp-2-c12/d1x+γt+4d13exp--c12/d1x+γtβt-4d14αt.
The unknown functions in the solutions are, respectively, given as (57)αt=c12tanh2c1t+c2-1,βt=c1d1tanhc1t+c2,γt=c3+12-d1c12lntanh2c1t+c2-1.
Exact solutions of the fast diffusion equation (52) are investigated from the symmetry point of view in many literatures. The widest and most arranged collection of such solutions is given in [37]. The second-order CLBS and corresponding symmetry reductions for (52) are referred to [12]. The solutions constructed here are not equivalent to the ones in [12, 37].
5. Conclusions
In this paper, the linear determining equations with parameters are applied to construct CLBSs of the nonlinear reaction-diffusion equation (2). The method generalizes the determining equations used in the search for admissible Lie operators. The procedure for the derivation of exact solutions to the evolution equation (10) is sketched as below. Firstly, finding solutions of the linear determining equations (20), we will obtain the explicit form of CLBSs (1) for (2). Solving the obtained differential constraints, we obtain u including three arbitrary functions depending on t. Consequently, substituting the form of u into (10) will lead to three-dimensional dynamical system for the undetermined functions about t.
We are mainly concerned with third-order CLBSs of (2) in the present paper. The corresponding exact solutions are constructed due to the resulting symmetry reductions. Naturally, a question will arise: how complicated can the CLBS admitted by the evolution equation (10) be? In other words, what is the vast value of n in the CLBS (9) of (10)? For linear case (7), the question is answered by Svirshchevskii in [38] due to the dimension estimate of invariant subspace admitted by the nonlinear operator F, where F[u]=F(x,t,u,u1,…,uk). Does the resulting relation n≤2k+1 there is still effective for the nonlinear case? This will be involved in our future study.
Competing Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This work was supported by Key Scientific and Technological Project of Higher Education of Henan Province (no. 16A110032).
ZhdanovR. Z.Conditional Lie-Backlund symmetry and reduction of evolution equations199528133841385010.1088/0305-4470/28/13/0272-s2.0-21844486280FokasA. S.LiuQ. M.Nonlinear interaction of traveling waves of nonintegrable equations199472213293329610.1103/PhysRevLett.72.3293ZBL0973.355022-s2.0-12044249804GalaktionovV. A.MatrosovV. M.On new exact blow-up solutions for nonlinear heat conduction equations with source and applications199035863874ZBL0735.350742-s2.0-84972500466GalaktionovV. A.Invariant subspaces and new explicit solutions to evolution equations with quadratic nonlinearities1995125222524610.1017/s03082105000280182-s2.0-84971166861BlumanG. W.KumeiS.1989New York, NY, USASpringerBlumanG. W.ColeJ. D.The general similarity solution of the heat equation19961810251042FushchychW. I.ShtelenW. M.SerovN. I.1989Kiev, UkraineNaukova DumkaTranslated into English by Kluwer Academic Publishers, Dordrecht, The Netherlands, 1993IbragimovN. K.1985Boston, Mass, USAReidelLiuQ. M.Exact interaction of solitary waves for certain nonintegrable equations199637132434510.1063/1.531393LiuQ. M.Exact solutions to nonlinear equations with quadratic nonlinearity200134245083508810.1088/0305-4470/34/24/306ZBL1051.350362-s2.0-0035933687QuC. Z.Group classification and generalized conditional symmetry reduction of the nonlinear diffusion-convection equation with a nonlinear source199799210713610.1111/1467-9590.000582-s2.0-0001290663QuC. Z.Exact solutions to nonlinear diffusion equations obtained by a generalized conditional symmetry19996232833022-s2.0-6544263763QuC.ZhangS.LiuR.Separation of variables and exact solutions to quasilinear diffusion equations with nonlinear source20001441-29712310.1016/s0167-2789(00)00069-52-s2.0-0039592610QuC. Z.JiL. N.WangL. Z.Conditional lie Bäcklund symmetries and sign-invariants to quasi-linear diffusion equations2007119435539110.1111/j.1467-9590.2007.00388.x2-s2.0-34948834672JiL.QuC.Conditional Lie-Bäcklund symmetries and invariant subspaces to non-linear diffusion equations201176461063210.1093/imamat/hxq069ZBL1230.350102-s2.0-80051675165JiL.QuC.Conditional Lie-Bäcklund symmetries and invariant subspaces to nonlinear diffusion equations with convection and source2013131326630110.1111/sapm.120102-s2.0-84885178216Basarab-HorwathP.ZhdanovR. Z.Initial-value problems for evolutionary partial differential equations and higher-order conditional symmetries200142137638910.1063/1.13301992-s2.0-0035582578ZhdanovR. Z.AndreitsevA. Y.Non-classical reductions of initial-value problems for a class of nonlinear evolution equations200033325763578110.1088/0305-4470/33/32/312ZBL0965.350272-s2.0-0034682683ZhdanovR. Z.Higher conditional symmetry and reduction of initial value problems2002281172710.1023/A:1014962601569ZBL1001.350042-s2.0-0036537869SergyeyevA.Constructing conditionally integrable evolution systems in (1 + 1) dimensions: a generalization of invariant modules approach200235357653766010.1088/0305-4470/35/35/3092-s2.0-0037031664JiL. N.QuC. Z.ShenS.Conditional Lie-Bäcklund symmetry of evolution system and application for reaction-diffusion system2014133111814910.1111/sapm.120422-s2.0-84904052848YanenkoN. N.Theory of consistency and methods of integrating systems of nonlinear partial differential equationsProceedings of the 4th All-Union Mathematical CongressLeningrad, Russia2472591964 (Russian)PucciE.SaccomandiG.Evolution equations, invariant surface conditions and functional separation of variables20001391-2284710.1016/s0167-2789(99)00224-92-s2.0-0347252623GalaktionovV. A.SvirshchevskiS.2007London, UKChapman & HallOlverP. J.RosenauP.The construction of special solutions to partial differential equations1986114310711210.1016/0375-9601(86)90534-7ZBL0937.355012-s2.0-46149142625OlverP. J.RosenauP.Group invariant solutions of differential equations198747226327810.1137/01470182-s2.0-0023327684OlverP. J.Direct reduction and differential constraints1994444192250952310.1098/rspa.1994.00352-s2.0-0028768164KaptsovO. V.Invariant sets of evolution equations199219875376110.1016/0362-546x(92)90219-52-s2.0-0001369885LeviD.WinternitzP.Non-classical symmetry reduction: example of the boussinesq equation198922152915292410.1088/0305-4470/22/15/0102-s2.0-0001678923OvsiannikovL. V.1982New York, NY, USAAcademic PressClarksonP. A.KruskalM. D.New similarity reductions of the Boussinesq equation198930102201221310.1063/1.528613ZBL0698.351372-s2.0-18944400520KaptsovO. V.B-determining equations: applications to nonlinear partial differential equations19956326528610.1017/s09567925000018322-s2.0-84974249785KaptsovO. V.Linear determining equations for differential constraints199818911-121839185410.1070/sm1998v189n12abeh0003682-s2.0-0032243643KaptsovO. V.VerevkinI. V.Differential constraints and exact solutions of nonlinear diffusion equations20033651401141410.1088/0305-4470/36/5/3152-s2.0-0037423468AndreevV. K.KaptsovO. V.PukhnachovV. V.RodionovA. A.1998Dordrecht, NetherlandsSpringer10.1007/978-94-017-0745-9KunzingerM.PopovychR. O.Generalized conditional symmetries of evolution equations2011379144446010.1016/j.jmaa.2011.01.027ZBL1211.350212-s2.0-79952196372PopovychR. O.VaneevaO. O.IvanovaN. M.Potential nonclassical symmetries and solutions of fast diffusion equation20073622-316617310.1016/j.physleta.2006.10.0152-s2.0-33846194257SvirshchevskiiS. R.Nonlinear differential operators of first and second order possessing invariant linear spaces of maximal dimension199510521346135310.1007/BF020709302-s2.0-34249758474