1,886
Views
2
CrossRef citations to date
0
Altmetric
Research Articles

Euler–Lagrange equations for variational problems involving the Riesz–Hilfer fractional derivative

ORCID Icon & ORCID Icon
Pages 678-696 | Received 12 Oct 2019, Accepted 16 Apr 2020, Published online: 20 May 2020

Abstract

In this paper, we obtain the Euler-Lagrange equations for different kind of variational problems with the Lagrangian function containing the Riesz-Hilfer fractional derivative. Since the Riesz-Hilfer fractional derivative is a generalization for the Riesz-Riemann-Liouville and the Riesz-Caputo derivative, then our results generalize many recent works in which the Lagrangian function involving the Riesz-Riemann-Liouville or the Riesz-Caputo derivative. We also study the problem in the presence of delay derivatives and establish a version for Noether theorem in the Riesz-Hilfer sense. In order to achieve our aims we derive some formulas to integration by parts for the Riesz-Hilfer fractional derivative. In the last section, examples are given to clarify the possibility of applicability of our results. In order to clarify the significant conclusions of the paper, we refer to our techniques enable to study many different variational problems containing the Riesz-Hilfer derivative.

1. Introduction

There are many applications for differential equations and inclusions of fractional order in various fields [Citation1–4], and many results are obtained on this subject, one can see, e.g. [Citation5–14] and the references therein. Furthermore, the field of calculus of variation have importance in engineering, optimal control theory, and pure and applied mathematics, see for example, [Citation15–18]. The calculus of variations with fractional derivative is initiated with the work of Riewe [Citation19, Citation20]. It deals with variational problems where the Lagrangian function containing fractional derivative or fractional integrals. In order to mention some recent works on fractional variational problems, Agrawal [Citation21] derived a necessary conditions for functionals, containing multiple of left and right Riemann–Liouville fractional derivative (RLFD), to have an extremum. Agrawal [Citation22]. considered a Lagrangian containing the Riesz–Caputo fractional derivative (RCFD). Almeida [Citation23] considered different variational problems involving Caputo fractional derivatives(CFD). Almeida [Citation24] obtained the necessary conditions for a pair function-time to be an optimal solution, when the Lagrangian function involving (RCFD) and the interval of integration is contained in the interval of fractional derivative. Odzijewicz et al. [Citation25]. obtained the Euler–Lagrange equations for functionals containing Caputo and combined Caputo fractional derivatives. Sayevand et al. [Citation26], considered delay fractional variation problems with isoperimetric and holomorphic constraints and involving (CFD). Tavares et al. [Citation27]studied two variational problems involving (RCFD) and a state time delay. Almiada et al. [Citation28] considered functional containing distributed–order fractional derivatives. For more results on fractional variational calculus, we refer to [Citation29–40].

On the other hand, Hilfer [Citation41] introduced the Hilfer fractional derivative(HFD), which includes (RLFD) and (CFD). Agrawal et al. [Citation42] developed the fractional Euler–Lagrange equations for functionals containing (HFD) with three parameter fractionals.

For new and important developments for searching exact and numerical solutions for nonlinear partial differential equations by used a kind of mathematical methods, we refer to [Citation43–57].

In this paper, and in order to generalize many results cited above, we give the notations for the Riesz–Hilfer fractional derivative (RHFD) which includes Riesz–Riemann–Liouville fractional derivative (RRLFD) and the Riesz–Caputo fractional derivatives (RCFD), and hence we obtain the Euler–Lagrange equations for different kind of fractional variational problems, with a Lagrangian containing (RHFD). In Section 2, we derive some formulas to integration by parts for (RHFD). In Section 3, we consider a simple fractional variational problem, then we take the case when the interval of integration of the functional is contained in the interval of fractional derivative. In Section 4, we obtain the Euler–Lagrange equations for isoperimetric problems, in which the function eligible for the extremization of a given definite integral is required to conform with certain restrictions. In Section 5, we study the problem in the presence of delay derivatives. In Section 6, we obtain the conditions that assure a pair of function-time to be an optimal solution. In Section 7, we establish a version for Noether theorem in the Riesz–Hilfer sense is given. Finally, we give examples to clarify the possibility of applicability of our results.

To the best of our knowledge, up to now, no work has reported on fractional variational problems, with a Lagrangian function involving (RHFD). Moreover, since (RHFD) is generalization for (RRLFD) and (RCFD), then our work generalizes many works mentioned above and allow to consider other variational problems, where the Lagrangian function involving (RHFD).

2. Preliminaries and notations

For any natural number m let ACm[a,b]:={f:f has continuous derivatives up toorder m on [a,b] and f(m)L1[a,b]} In the sequel we use the following notations:

  1. (FVP) is the fractional variational problem

  2. (RLFI) is the Riemann–Liouville fractional integral.

  3. (LRLFI) is the left-sided Riemann–Liouville fractional integral.

  4. (RRLFI) is the right-sided Riemann–Liouville fractional integral.

  5. (LRLFD) is the left-sided Riemann–Liouville fractional derivative.

  6. (RRLFD) is the right-sided Riemann–Liouville fractional derivative.

  7. (RFI) is the Riesz fractional integral.

  8. (LHFD) is the left-sided Hilfer fractional derivative.

We recall some concepts on fractional calculus [Citation1, Citation4].

Definition 2.1

The (LRLFI)of order μ>0 for a Lebesgue integrable function h:[a,b]R is given by: (1) aItμh(t):=1Γ(μ)at(ts)μ1h(s)ds,t[a,b],(1) where Γ is the Euler gamma function.

Definition 2.2

The (RRLFI)of order μ>0 for a Lebesgue integrable function h:[a,b]R is defined as: (2) tIbμh(t):=1Γ(μ)tb (st)μ1h(s)ds,t[a,b].(2)

It is known that aItμ(aItβh(t))=aItμ+βh(t), and tIbμ(tIbβh(t))=tIbμ+βh(t), β,μ>0. If μ=0, we set aIt0h(t)=tIb0h(t)=h(t), t[a,b].

Lemma 2.1

[Citation1]

Let μ>0 and hLp[a,b],ηLs[a,b] such that p1,s1 and 1/p+1/s1+μ (p>1,s>1, in the case when 1/p+1/s<1+μ). Then (3) abh(t)(aItμη)(t)dt=abη(t)(tIbμh)(t)dt.(3)

Definition 2.3

The (RFI) of order μ>0 for a function hL1[a,b], is defined as: (4) aRIbμh(t):=12[aItμh(t)+tIbμh(t)],t[a,b].(4)

Notice that

(5) aRIbμh(t)=12Γ(μ)at(ts)μ1h(s)ds+tb(st)μ1h(s)ds,=12Γ(μ)at|ts|μ1h(s)ds+tb|ts|μ1h(s)ds,=12Γ(μ)ab|ts|μ1h(s)ds,t[a,b].(5) In what follows μ denotes to a positive real number and m is the smallest natural number such that μm.

Definition 2.4

Let hL[a,b] such that aItmμhACm[a,b]. The (LRLFD) of order μ for h at t[a,b] is given by (6) aDtμh(t)=dmdtm[aItmμh(t)],if m1<μ<m,dmhdtm(t),if μ=m.=1Γ(mμ)dmdtm[at(ts)mμ1h(s)ds],if m1<μ<m,dmhdtm(t),if μ=m.(6)

Definition 2.5

Let hL1([a,b],E) such that tIbmμhACm([a,b],E). The right-sided Riemann–Liouville fractional derivative of order μ for h at t[a,b] is defined by (7) tDbμh(t):=(1)mdmdtm[tIbmμh(t)],if m1<μ<m,(1)mdmhdtm(t),if μ=m.=(1)mΓ(mμ)dmdtmtb(st)mμ1h(s)ds,if m1<μ<m,(1)mdmhdtm(t),if μ=m.(7)

Definition 2.6

Let hL1[a,b] such that aItmμh,tIbmμhACm[a,b]. The (RRLFD) of order μ for h is given by (8) aRDbμh(t)=12[aDtμh(t)+(1)m tDbμh(t)],t[a,b].(8)

Remark 2.1

If μ=m then for any t[a,b] (9) aRDbmh(t)=12[aDtmh(t)+(1)m tDbmh(t)]=dmhdtm(t).(9)

Definition 2.7

The left-sided Caputo fractional derivative of order μ for hACm[a,b] is given at t[a,b] by (10) aCDtμh(t):=aItmμh(m)(t),if m1<μ<m, dmhdtm(t),if μ=m.=1Γ(mμ)at(ts)mμ1h(m)ds,if m1<μ<m(1)mdmhdtm(t),if μ=m.(10)

Definition 2.8

The right-sided Caputo fractional derivative of order μ for h ACm[a,b]is given at t[a,b] by (11) tCDbμh(t):=tIbmμddtmh(t),if m1<μ<m,dmhdtm(t),if μ=m.=1Γ(mμ)at (ts)mμ1h(m)ds,if m1<μ<m, (1)mdmhdm(t),if μ=m.(11)

Definition 2.9

Let hACm[a,b]. The (RCFD) of order μ>0 for h ACm[a,b] is given by (12) aRCDbμh(t):=12[aCDtμh(t)+(1)m tCDbμh(t)],t[a,b].(12)

Definition 2.10

[Citation41]

The (LHFD) of orderμ(0,1] and type β[0,1] for a function h:[a,b]E is given by (13) aDtμ,βh(t):=aIt(1β)(1μ)D aItβ(1μ)h(t),(13)

where, D=d/dt. Notice if hC1γγ([a,b],E), then aDtμ,βh(t) is well defined on [a,b], where γ=μ+βμβ, C1γγ([a,b],E):={xC1γ([a,b],E):aDtγxC1γ([a,b],E)}, and C1γ([a,b],E):={xC((a,b],E):(ta)1γx(t)C([a,b],E).

Definition 2.11

[Citation41]

The right-sided Hilfer fractional derivative of order μ(0,1] and type β[0,1] for a function h:[a,b]E is given by (14) tDbμ,βh(t):=tIb(1β)(1μ)(D)tIbβ(1μ)h(t).(14)

Definition 2.12

The (RHFD) for hC1([a,b],E) is given by (15) aRHDbμ,βh(t):=12[aDtμ,βh(t)tDbμ,βh(t)],t[a,b].(15)

Remark 2.2

From the above definitions it follows that:

(i) If β=0, then aDtμ,βh(t)=aCDtμh(t), tDbμ,βh(t)=tCDbμh(t) and aRHDbμ,βh(t)=aRCDbμh(t).

(ii) If β=1, then aDtμ,βh(t)=aDtμh(t), tDbμ,βh(t)=tDbμh(t) and aRHDbμ,βh(t)=aRDbμh(t).

In the following lemmas we derive formulas to integration by parts for (RHFD)

Lemma 2.2

If μ(0,1], β[0,1] and h,η:[a,b]R are continuously differentiable, then (16) abh(t) aRHDbμ,βη(t)dt=ab η(t) aRHDbμ,1βh(t)dt+12 tIb(1β)(1μ)h(t)aItβ(1μ)η(t)|ab+12 aIt(1β)(1μ)h(t)tIbβ(1μ)η(t)|ab.(16)

Proof.

Let q=(1β)(1μ). Using Lemma 2.1, then ordinary integration by parts and then again by Lemma 2.1, it yields (17) abh(t)aDtμ,βη(t)dt=abh(t)aItq(DaItβ(1μ)η(t))dt=ab(DaItβ(1μ)η(t))tIbqh(t)dt=tIbqh(t).aItβ(1μ)η(t)|abab(aItβ(1μ)η(t))(DtIbqh(t))dt=tIbqh(t).aItβ(1μ)η(t)|ababη(t)tIbβ(1μ)DtIbqh(t)dt=tIbqh(t).aItβ(1μ)η(t)|ab+abη(t)tIbβ(1μ)(D)tIbqh(t)dt=tIbqh(t).aItβ(1μ)η(t)|ab+abη(t)tDbμ,βh(t)dt.(17) Similarly, one can obtain (18) ab h(t) tDbμ,βη(t)dt=abh(t) tIbq(D tIbβ(1μ)η(t))dt=ab(D tIbβ(1μ)η(t))aItqh(t)dt=aItqh(t).tIbβ(1μ)η(t)|abab( tIbβ(1μ)η(t))(D aItqh(t))dt=aItqh(t).tIbβ(1μ)η(t)|ababη(t)aItβ(1μ)DaItqh(t)dt=aItqh(t).tIbβ(1μ)η(t)|ab+abη(t)aItβ(1μ)(D)aItqh(t)dt=aItqh(t).tIbβ(1μ)η(t)|ab+abη(t)aDtμ,βh(t)dt.(18) From the definition of aRHDbμ,βη(t), (Equation17) and (Equation18) it yields (Equation16).

Corollary 2.1

(1) If we put β=0 in (Equation16), we obtain relation (Equation20) in [Citation22] and the integration rule by parts formula in [Citation24]. In fact (19) abh(t) aRCDbμη(t)dt=abh(t)aRHDbμ,0η(t)dt=abη(t) aRHDbμ,1h(t)dt+12η(t) tIb1μh(t)|ab+12η(t) aIt1μh(t)|ab=ab η(t) aRDb1μh(t)dt+η(t)aRIb1μh(t)|ab.(19) (2) If we put β=1 in (Equation16) we obtain relation (Equation21) in [Citation22]. In fact (20) abh(t) aRDbμη(t)dt=abh(t)aRHDbμ,1η(t)dt=abη(t) aRHDbμ,0h(t)dt+12h(t)aIt1μη(t)|ab+12h(t)tIb1μη(t)|ab=abη(t)aRCDbμh(t)dt+h(t)aRIb1μη(t)|ab.(20)

We need to the following lemma in the third section.

Lemma 2.3

If μ(0,1],β[0,1] and h,η:[a,b]R are continuously differentiable and r(a,b), then (21) arh(t) aRHDbμ,βη(t)dt=arη(t) aRHDbμ,1βh(t)dt+12arη(t)(tDrμ,1βtDbμ,1β)h(t))dt+tIrqh(t).aItβ(1μ)η(t)|arrbη(t)(aDtμ,1βrDtμ,1β)h(t)dt+aItqh(t).tIbβ(1μ)η(t)|abrItqh(t).tIbβ(1μ)η(t)|rb.(21) and (22) rbh(t)aRHDbμ,βη(t)dt=rbη(t)aRHDbμ,1βη(t)dt+12[rbη(t)(aDtμ,1βrDtμ,1β)h(t))dt+rItqh(t)tIbβ(1μ)η(t)|rb+ar η(t)(tDbμ,1βtDrμ,1β)h(t)dt+tIbqh(t)aItβ(1μ)η(t)|abtIrqh(t)aItβ(1μ)η(t)|ar].(22)

Proof.

Let q=(1μ)(1β). According to (Equation17) we have (23) ar h(t) aDtμ,βη(t)dt=ar η(t) tDrμ,1βh(t)dt+tIrqh(t).aItβ(1μ)η(t)|ar=arη(t)tDbμ,1βh(t)dt+arη(t)(tDrμ,1βh(t)tDbμ,β1h(t))dt+tIrqh(t).aItβ(1μ)η(t)|ar.(23) Also, relation (Equation18) leads to (24) ar h(t)tDbμ,βη(t)dt=abh(t)tDbμ,βη(t)dtrbh(t)tDbμ,βη(t)dt=ab η(t)aDtμ,1βh(t)dtaItqh(t).tIbβ(1μ)η(t)|abrb η(t)rDtμ,1βh(t)dt+rItqh(t).tIbβ(1μ)η(t)|rb=abη(t)aDtμ,1βh(t)dtrbη(t)aDtμ,1βh(t)dt+rbη(t)aDtμ,1βh(t)dtrb η(t)rDtμ,1βh(t)dtaItqh(t).tIbβ(1μ)η(t)|ab+tItqh(t).tIbβ(1μ)η(t)|rb=abη(t)aDtμ,1βh(t)dt+rbη(t)(aDtμ,1βrDtμ,1β)h(t)dtaItqh(t).tIbβ(1μ)η(t)|ab+rItqh(t).tIbβ(1μ)η(t)|rb.(24) It follows from Equations (Equation23) and (Equation24) that (25) arh(t) aRHDbμ,βη(t)dt=12ar h(t) aDtμ,βη(t)dtar h(t) tDbμ,βη(t)dt=arη(t) aRHDbμ,1βη(t)dt+12arη(t)(tDrμ,1βh(t)tDbμ,β1h(t))dt+12tIbqh(t).aItβ(1μ)η(t)|ab12rbη(t)(aDtμ,1βrDtμ,1β)h(t)dt+aItqh(t).tIbβ(1μ)η(t)|abrItqh(t).tIbβ(1μ)η(t)|rb,(25) and this proves (Equation21). Similarly we prove the validity of (Equation22). In view of (Equation17) and (Equation18) we get (26) rbh(t) tDbμ,βη(t)dt=rb η(t) rDtμ,1βh(t)dttIbβ(1μ)η(t).rItqh(t)|rb=rbη(t)aDtμ,1βh(t)dt+rbη(t)(rDtμ,β1aDtμ,β1)h(t))dttIbβ(1μ)η(t).rItqh(t)|rb,(26) and (27) rb h(t)aDtμ,βη(t)dt=abh(t) aDtμ,βη(t)dtarh(t)aDtμ,βη(t)dt=abη(t)tDbμ,1βh(t)dt+tIbqh(t).aItβ(1μ)η(t)|abarη(t)tDrμ,1βh(t)dttIrqh(t).aItβ(1μ)η(t)|ar=ab η(t)tDbμ,1βh(t)dt+tIbqh(t).aItβ(1μ)η(t)|abar η(t)tDbμ,1βh(t)dt+ar η(t)(tDbμ,1βtDrμ,1β)h(t)dttIrqh(t).aItβ(1μ)η(t)|ar=rbη(t)tDbμ,1βh(t)dt+ar η(t)(tDbμ,1βtDrμ,1β)h(t)dt+tIbqh(t).aItβ(1μ)η(t)|abtIrqh(t).aItβ(1μ)η(t)|ar.(27) It follows from (Equation26) and (Equation27) that (28) rbh(t)aRHDbμ,βη(t)dt=rbη(t)aRHDbμ,1βη(t)dt+12[rbη(t)(aDtμ,1βrDtμ,1β)h(t))dt+tIbβ(1μ)η(t).rItqh(t)|rb+ar η(t)(tDbμ,1βtDrμ,1β)h(t)dt+tIbqh(t).aItβ(1μ)η(t)|abtIrqh(t).aItβ(1μ)η(t)|ar].(28) So, (Equation22) is true.

Remark 2.3

If we add Equations (Equation21) and (Equation22), then we obtain (Equation16).This means that the obtained results in Lemmas 2.2 and 2.3 are compatible.

We need to the following basic Lemma [Citation18].

Lemma 2.4

If G:[a,b]R is continuous function and ab G(t)η(t)dt=0, for every choice of the continuously differentiable function η for which η(a)=η(b)=0, we conclude that G(t)=0, for any t[a,b].

3. Euler–Lagrange equation for a simple fractional variational problem involving the Riesz–Hilfer derivative

Let μ(0,1] and β[0,1].

Theorem 3.1

Assume that the first and second partial derivatives of a Lagrangian function L:[a,b]×R3n R with respect to all its arguments are continuous. Consider a functional of the form (29) J[q1,,qn]=abL(t,q1(t),,qn(t),q1(t),,qn(t),aRHDbμ,βq1(t)),,aRHDbμ,βqn(t))dt,(29) defined on the set of functions q1,,qn which are continuously differentiable and such that aRHDbμ,βqi(i=1,2,,n) is continuous in [a,b] and that (30) qi(a)=qiaandqi(b)=qib,i=1,2,,n.(30) Then a necessary condition for the functional (Equation29) attains an extremum at qi, i=1,2,,n is that qi satisfy the following Euler–Lagrange equations: (31) Lqi(t)ddtLqi(t)aRHDbμ,1βL aRHDbμ,βqi(t)=0,t[a,b].(31) If β=1, then the functions qi should be satisfy the transversality condition: (32) L aRDbμqi(t)t=a=L aRDbμqi(t)t=b=0.(32)

Proof.

It is known that the necessary condition for qi, i=1,2,,n to be extremum, is given by (33) J(qi+ϵηi)ϵϵ=0=0,i=1,2,,n,(33) where ηi are arbitrary continuously differentiable functions for which (34) ηi(a)=ηi(b)=0.(34) That is

(35) 0=abi=1nLqidqidϵ+Lqidqidϵ+L aRHDbμ,βqid(aRHDbμ,βqi)dϵdt=i=1nabLqiηi(t)+Lqiηi(t)+L aRHDbμ,βqi(aRHDbμ,βηi)(t)dt.(35) Put q=(1μ)(1β) and using Lemma 2.2, it follows that (36) abL aRHDbμ,βqi(aRHDbμ,βηi)(t)dt=ab ηi(t) aRHDbμ,1βzi(t)dt+12 tIbqzi(t).aItβ(1μ)η(t)|ab+12 aItqzi(t).tIbβ(1μ)η(t)|ab,(36) where zi(t)=(L/ aRHDbμ,βqi)(t). Notice that if β=0, then by (34) we get (37) 12 tIb1μzi(t)(t).η(t)|ab+12 aIt1μzi(t)(t).η(t)|ab=12 tIb1μzi(t)(t).η(t)|ab+12 aIt1μzi(t)(t).η(t)|ab=0.(37) If β(0,1), then q0β(1μ), and consequently, the continuity of L/( aRHDbμ,βqi) and η(t) implies (38) tIbqzi(t).aItβ(1μ)η(t)|ab+aItqzi(t).tIbβ(1μ)η(t)|ab=0.(38) If β=1, then Equation (Equation32) leads to (39) tIbqzi(t).aItβ(1μ)η(t)|ab+aItqzi(t).tIbβ(1μ)η(t)|ab=zi(t)aIt1μη(t)|ab+zi(t).tIb1μη(t)|ab=zi(b)aIt1μη(b)zi(a).tIb1μη(a)=0.(39) Then, by (Equation37)–(Equation39), Equation (Equation36) becomes (40) abL aRHDbμ,βqi(aRHDbμ,βηi)(t)dt=ab ηi(t) aRHDbμ,1βzi(t)dt.(40) Moreover, Equations (Equation34) leads to (41) abLqiηi(t)dt=Lqiηi(t)|ababddtLqi(t)dt=abddtLqi(t)dt.(41) It yields from (Equation35),(Equation40) and (Equation41) that (42) 0=i=1nabLqi(t)ddtLqi(t)aRHDbμ,1β×L aRHDbμ,βqi(t)ηi(t)dt,(42) and this should be true for all admissible functions ηi. Since the functions ηi are independent, then we can choose for any fixed i, ηk(t)=0,t[a,b],ki. Hence Lemma 2.4, gives us (43) Lq(t)ddtLqi(t)aRHDbμ,βL aRHDbμ,βqi(t)=0,i=1,2,,n,(43) and the proof is completed.

Remark 3.1

If β=0in the previous theorem and i = 1 then we obtain Theorem 1 in [Citation22].

Now we extend Theorem 3.1 when the interval of integration is contained in the interval of fractional derivative.

Theorem 3.2

Assume that the Lagrangian function L:[a,b]×R3n R be as in Theorem 3.1. Consider the functional (44) J[q1,,qn]=ABL(t,q1(t),,qn(t),aRHDbμ,βq1(t)),,aRHDbμ,βqn(t))dt,(44) defined on the set of functions q1,,qn which are continuously differentiable and have continuous Riesz–Hilfer derivatives of order μ and of type β in [a,b], where [A,B][a,b]. Then necessary conditions for the functions qi, i=1,2,,n, which satisfy (Equation30) to be an extremum of the functional given by (Equation31) are that qi, i=1,2,,n, satisfy the following Euler–Lagrange equations: (45) LqiddtLqiARHDBμ,1βL aRHDbμ,βqi(t)=0,for any t[A,B],(45) (46) tDBμ,1βL aRDbμqi(t)tDAμ,1βL aRHDbμ,βqi(t)=0,forany t[a,A],(46) (47) ADtμ,1βL aRDbμqi(t)BDtμ,1βL aRHDbμ,βqi(t)=0,for any t[B,b].(47) If β=1, then qi, i=1,2,,n, should be satisfy the transversality condition (48) L aRHDbμ,βqit=A=L aRHDbμ,βqit=B=0.(48)

Proof.

As above, the necessary condition for qi,i=1,2,,n to be extremum, is given by (49) J(qi+ϵηi)ϵϵ=0=0,i=1,2,,n,(49) where ηi, i=1,2,,n, are arbitrary continuously differentiable functions for which (50) ηi(a)=ηi(b)=ηi(A)=ηi(B)=0,i=1,2,,n.(50) That is

(51) 0=ABi=1nLqidqidϵ+Lqidqidϵ+L aRDbμqid(aRHDbμ,βqi)dϵdt=i=1nABLqiηi(t)+Lqiηi(t)+L aRDbμqi aRHDbμ,β ηi(t)dt=i=1nABLqiηi(t)dt+ABLqiηi(t)dt+12ABLaRHDbμ,βqi(aDtμ,βηi)(t)dt12ABLtRHDbμ,βqi (tDbμ,βηi)(t)d=i=1nABLqiddtLqiηi(t)dt+i=1nLqi(ηi(A)ηi(B))+12aBL aRHDbμ,βqi(aDtμ,βηi)(t)dtaAL aRHDbμ,βqi(aDtμ,βηi)(t)d12AbL aRDbμqi(tDbμ,βηi)(t)dtBbL aRHDbμ,βqi(tDbμ,βηi)(t)d.(51) Utilizing the rule of integration by parts (Equation21) and (Equation22) and taking into account the assumptions ηi(A)=ηi(B)=0, i=1,2,,n, we get (52) dJdϵ=i=1nABLqiddtLqiηi(t)dt+12aBηi(t)tDBμ,1βL aRHDbμ,βqidt+12 tIB(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aB12aAηi(t)tDAμ,1βL aRHDbμ,βqidt12 tIA(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aA12Abηi(t)ADtμ,1βL aRHDbμ,βqidt12 AIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Ab+12Bbηi(t)BDtμ,1βL aRHDbμ,βqidt+12 BIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Bb.(52) Notice that if β=0, then equation (Equation40) leads to (53) tIB(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aBtIA(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aAAIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Ab+BIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Bb=tIB1μL aRHDbμ,βqiη(t)|aBtIA1μL aRHDbμ,βqi.η(t)|aAAIt1μL aRHDbμ,βqi.η(t)|Ab+BIt1μ×L aRHDbμ,βqi.η(t)|Bb=0.(53) If β(0,1), then (1β)(1μ)0β(1μ), and consequently, the continuity of L/( aRHDbμ,βqi) and η(t) implies to (54) 0= tIB(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aB=tIA(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aA=AIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Ab=BIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Bb.(54) If β=1, then equation (Equation48) implies that (55) tIB(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aBtIA(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)η(t)|aAAIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Ab+BIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)η(t)|Bb=L aRHDbμ,βqi.aIt1μη(t)|aBL aRHDbμ,βqi.aIt1μη(t)|aAL aRHDbμ,βqi.tIb1μη(t)|Ab+L aRHDbμ,βqi tIb1μη(t)|Bb=L aRHDbμ,βqi.aIt1μη(t)|t=BL aRHDbμ,βqi.aIt1μη(t)|t=A+L aRHDbμ,βqi.tIb1μη(t)|t=AL aRHDbμ,βqi.tIb1μη(t)|t=B=0.(55) Using equations (Equation53)–(Equation55), equation (Equation52) becomes (56) dJdϵ=i=1nABLqiddtLqiηi(t)dt+12aBηi(t)tDBμ,1βL aRHDbμ,βqidt12aAηi(t)tDAμ,1βL aRHDbμ,βqidt12Abηi(t)ADtμ,1βL aRHDbμ,βqi)dt+12Bbηi(t)BDtμ,1βL aRHDbμ,βqid=i=1nABLqiddtLqiηi(t)dt+12aAηi(t)tDBμ,1βL aRHDbμqidt+12ABηi(t)tDBμ,1βL aRHDbμ,βqidt12aAηi(t)tDAμ,1βL aRDbμqidt12ABηi(t)ADtμ,1βL aRHDbμ,βqi)dt12Bbηi(t)ADtμ,1βL aRHDbμ,βqi)dt+12Bbηi(t)BDtμ,1βL aRHDbμ,βqidt.(56) and hence

(57) 0=i=1i=nABLqiddtLqiARHDBμ,1β×L aRHDbμ,βqi(t)ηi(t)dt+12aA tDBμ,1βL aRHDbμqi(t)tDAμ,1βL aRHDbμ,βqi(t)ηi(t)dt+12BbBDtμ,1βL aRHDbμqi(t)ADtμ,1β×L aRHDbμ,βqi(t)ηi(t)dt.(57) Since the functions ηi are independent, then for appropriate choices of ηi, we derive, by applying Lemma 2.4, the necessary conditions (Equation45)–(Equation47).

Remark 3.2

If we put β=0 in the previous theorem, then we obtain Theorem 3 in [Citation24].

4. Isoperimetric problem

In this section we consider a problem in which the eligible function for the extremization of a given definite integral is required to satisfy certain restrictions that are added to the usual conditions.

Theorem 4.1

Assume that the first and second partial derivatives of a Lagrangian function L:[a,b]×R2n R and a function z:[a,b]×RnR with respect to all its arguments are continuous. Consider the functionals (58) J[q1,q2,,qn]=abL(t,q1(t),,qn(t),aRHDbμ,βq1(t)),,aRHDbμ,βqn(t))dt.(58) and (59) I[q1,q2,,qn] =abz(t,q1(t),,qn(t))dt.(59) If aRHDbμ,βqi(t) (i=1,2,,n) is continuous on [a,b]. Then, necessary conditions for J[q1,q2,,qn] to have an extremum at qi, i=1,2,,n, which satisfies the boundary conditions (Equation30), such that (60) I[q1,q2,,qn] =abz(t,q1(t),,qn(t))dt=l,(60) are that qi satisfy the following Euler–Lagrange equations (61) LqiaRHDbμ,1βL aRDbμqi+λzqi=0,i=1,2,,n.(61) If β=1, then qi should satisfy the transversality conditions (62) tIb(1β)(1μ)L aRHDbμ,βqi(t)t=a=0,(62) and (63) aIt(1β)(1μ)L aRHDbμ,βqi(t)t=b=0,(63) where λ is the Lagrange's multiplier whose value can be determine by the conditions on L and z.

Proof.

To derive the necessary conditions let (64) qi(t)=qi(t)+ϵ1ηi(t)+ϵ2ζi(t),i=1,2,,n,(64) where ηi, and ζi i=1,2,,n, are arbitrary continuously differentiable functions for which (65) ηi(a)=ηi(b)=ζi(a)=ζi(b)=0,i=1,2,,n.(65) Inserting (Equation64) in (Equation59) and (Equation60), respectively, we get (66) J[ϵ1,ϵ2]=abL(t,q1(t)+ϵ1η1(t)+ϵ2ζ1(t),,qn(t)+ϵ1ηn(t)+ϵ2ζn(t),aRHDbμ,βq1(t)+ϵ1aRHDbμ,βη1(t)+ϵ2aRHDbμ,βζ1(t),,aRHDbμ,βqn(t)+ϵ1aRHDbμ,βηn(t))+ϵ2aRHDbμ,βζn(t))dt,(66) and (67) I[ϵ1,ϵ2]=abg(t,q1(t)+ϵ1η1(t)+ϵ2ζ1(t),,qn(t)+ϵ1ηn(t)+ϵ2ζn(t))dt=l.(67) Clearly, the parameters ϵ1 and ϵ2 are not independent because I[ϵ1,ϵ2]=l.Since qi, i=1,2,,n are assumed to be the actual extermizing functions, we have J[ϵ1,ϵ2] is extremum with respect to ϵ1 and ϵ2 which satisfy (Equation65), when ϵ1=ϵ2=0. According to the method of Lagrange multipliers we introduce (68) J[ϵ1,ϵ2]=J[ϵ1,ϵ2]+λI[ϵ1,ϵ2],(68) where λ is the Lagrange's multiplier. Then, according to the method of Lagrange multipliers we must have (69) Jϵ1=Jϵ2=0whenϵ1=ϵ2=0.(69) It follows, by applying the rule of integration by parts (Equation16),

(70) 0=Jϵ1=i=1nabLqiηi+L aRHDbμ,βqi(aRHDbμ,βηi)dt+λabi=1i=nabgqiηidt=i=1nabLqiaRHDbμ,1βLaRHDbμ,βqi+λgqiηidt+12 tIb(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)ηi(t)|ab+12 aIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)ηi(t)|ab],(70) and

(71) 0=Jϵ2=i=1nabLqiςi(t)+L aRDbμqi(aRHDbμ,βςi)(t)dt+λabi=1i=ngqiςi(t)dt=i=1nabLqiaRHDbμ,1βL aRHDbμ,βqi+λgqiςidt+12 tIb(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)ςi(t)|ab+12 aIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)ςi(t)|ab].(71) As in the proof of Theorem 3.1, one can show that if β[0,1), then by (Equation66) one obtains (72) tIb(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)ηi(t)|ab=aIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)ηi(t)|ab]= tIb(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)ςi(t)|ab=aIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)ςi(t)|ab=0.(72) Notice that equations (Equation62) and (Equation63) imply the validity of (Equation72) when β=1. Since setting ϵ1=ϵ2=0 is equivalent to replacing qi and aRHDbμ,βqi by qi and aRHDbμ,βqi. Consequently from (Equation57)–(Equation59), we get

(73) 0=Jϵ1=i=1nabLqiηi+L aRHDbμ,βqi(aRHDbμ,βηi)dt+λabi=1i=nabgqiηidt=i=1nabLqiaRHDbμ,1βLaRHDbμ,βqi+λgqiηidt,(73) and

(74) 0=Jϵ2=i=1nabLqiςi(t)+L aRDbμqi(aRHDbμ,βςi)(t)dt+λabi=1i=ngqiςi(t)dt=i=1nabLqiaRHDbμ,1βL aRHDbμ,βqi+λgqiςidt.(74)

Since the functions ηi and ςi are independent, the proof is finished.

5. Fractional variational problem with delay

We study the case when there is a delay on the system. Let τ(0,ba) and consider the functional (75) T[q1,,qn]=abL(t,q1(t),,qn(t),aRH×Dbμ,βq1(t),,aRHDbμ,βqn(t),q1(t),,qn(t),q1(tτ),,qn(tτ),q1(tτ),,qn(tr))dt,(75) where qi :[a,b]R, i=1,2,,n are continuously differentiable and aRHDbμ,βqi(t)(i=1,2,,n) is continuous in [a,b] and L:[a,b]×R5n R is a Lagrangian function.

Theorem 5.1

Assume that the first and second partial derivatives of a Lagrangian function L:[a,b]×R5n R with respect to all of its arguments are continuous and ϕi:[ar,0]R, i=1,2,,n are continuous functions. Then a necessary condition for the functional  (Equation75) subject to boundary conditions (76) qi(t)=ϕi(t), t[ar,0]andqi(b)=qb, i=1,2,,n(76) achieves an extremum at qi ,i=1,2,,n, is that qi satisfy following Euler–Lagrange equations (77) Lqi(t)ddtLqi(t)aRHDbμ,1βL aRHDbμ,βqi(t)+Lqi(tr)(t+r)ddtLqi(tr)(t+r)=0,(77) for t[a,br], (78) Lqi(t)ddtLqi(t)aRHDbμ,1βqi(t)=0,for t[br,b],(78) If β=1, then qi should be verify the transversality condition (79) LaRHDbμ,βqit=b=LaRHDbμ,βqit=a=0.(79)

Proof.

We follow the approach discussed in the proof of Theorem 3.1, the necessary condition for qi ,i=1,2,,n to be extremum, is given by (80) J(qi+ϵηi)ϵϵ=0=0,i=1,2,,n,(80) where ηi are arbitrary continuously differentiable functions for which (81) ηi(t)=0 ifartarandηi(b)=0,i=1,2,,n.(81) Then

(82) i=1nabLqi(t)ηi(t)+Lqi(t)ηi(t)+LaRHDbμ,βqi(t) aRHDbμ,βηi(t)+Lqi(tr)(t)ηi(tr)+Lqi(tr)(t)ηi(tr)dt=0.(82) In the fourth and fifth term making the change of variables for tr and taking into account that ηi(t)=0, for t[ar,a], we obtain

(83) abLqi(tr)(t)ηi(tr)+Lqi(tr)(t)ηi(tr)dt=abrLqi(tr)(t+r)ηi(s)+Lqi(tr)(t+r)ηi(t)dt.(83)

It follows from (Equation82) and (Equation83) that (84) i=1nabrLqiηi(t)+Lqi(t)ηi(t)+LaRHDbμ,βqi(aRHDbμ,βηi)(t)+Lqi(tr)(t+r)ηi(t)+Lqi(tr)(t+r)ηi(t)dt+i=1nbrbLqiηi(t)+L aRHDbμ,βqi(aRHDbμ,βηi)(t)+Lqi(t)ηi(t)dt=0.(84) Using the usual rule integrating by parts and Equations (Equation21), (Equation22), equation (Equation84) becomes

(85) i=1nabrLqiddtLqiaRHDbμ,1βLaRHDbμ,βqi(t)×ηi(t)dt+Lqi(t)ηi(t)|abr+12abr×η(t)(tDbrμ,1βtDbμ,1β)LaRHDbμ,βqidt12brbη(t)(aDtμ,1βbrDtμ,1β)LaRHDbμ,βqidt+tIbr(1β)(1μ)LaRHDbμ,βqi.aItβ(1μ)η(t)|abr+aIt(1β)(1μ)LaRHDbμ,βqi.tIbβ(1μ)η(t)|abbrIt(1β)(1μ)LaRHDbμ,βqi.tIbβ(1μ)η(t)|brb+abrLqi(tr)(t+r)ddtLqi(tr)(t+r)×ηi(t)dt+Lqi(tr)(t+r)ηi(t)|abr]+i=1nbrbLqiddtLqiaRHDbμ,1βLaRHDbμ,βqi(t)ηi(t)dt+Lqi(t)ηi(t)|brb+12×brbη(t)(aDtμ,1βbrDtμ,1β)LaRHDbμ,βqidt+abr η(t)(tDbμ,1βtDbrμ,1β)LaRHDbμ,βqidt+brIt(1β)(1μ)LaRHDbμ,βqitIbβ(1μ)η(t)|brb+tIb(1β)(1μ)LaRHDbμ,βqi.aItβ(1μ)η(t)|abtIbr(1β)(1μ)LaRHDbμ,βqi.aItβ(1μ)η(t)|abr.(85) This equation reduces to

(86) 0=i=1nabrLqiddtLqiaRHDbμ,1βLaRHDbμ,βqi(t)×ηi(t)dt+Lqi(t)ηi(t)|abr+abrLqi(tr)(t+r)ddtLqi(tr)(t+r)×ηi(t)dt+Lqi(tr)(t+r)ηi(t)|abr+i=1nbrbLqiddtLqiaRHDbμ,1βLaRHDbμ,βqi(t)×ηi(t)dt+Lqi(t)ηi(t)|brb+aIt(1β)(1μ)LaRHDbμ,βqi.tIbβ(1μ)ηi(t)|ab+tIb(1β)(1μ)LaRHDbμ,βqi.aItβ(1μ)ηi(t)|ab.(86) If β[0,1), then by (Equation81) (87) aIt(1β)(1μ)LaRHDbμ,βqi.tIbβ(1μ)ηi(t)|ab=tIb(1β)(1μ)LaRHDbμ,βqi.aItβ(1μ)ηi(t)|ab=0,(87) and if β=1, then by condition (Equation79), equation (Equation87) still true. Therefore, (Equation86) becomes

(88) 0=i=1nabrLqiddtLqiaRHDbμ,1βLaRHDbμ,βqi(t)×ηi(t)dt+Lqi(t)ηi(t)|abr+abrLqi(tr)(t+r)ddtLqi(tr)(t+r)×ηi(t)dt+Lqi(tr)(t+r)ηi(t)|abr+i=1nbrbLqiddtLqiaRHDbμ,1βLaRHDbμ,βqi(t)×ηi(t)dt+Lqi(t)ηi(t)|brb.(88) If ηi(t)=0 on [br,b] and free on (a,br), for any i=1,2,,n, it follows from (Equation80) and (Equation88) the validity of (Equation77). If ηi(t)=0 on [a,br] and free on (br,b], for any i=1,2,,n it follows from (Equation80) and (Equation88) the validity of (Equation78).

6. Optimal time problem

In this section, we find the necessary conditions for a variational problem to have a extremum on an optimal time.

Theorem 6.1

Suppose that the first and second partial derivatives of a Lagrangian function L:[a,)×R3 R with respect to all its arguments are continuos. Consider a functional of the form (89) J[q,T]=aTL(t,q(t),aRHDbμ,βq(t))dt,(89) defined on the set of pairs (q,T),where T[a,b) and q is continuously differentiable, aRHDbμ,βqi(t)(i=1,2,,n) is continuous in [a,) and satisfy the boundary condition q(a)=qa. Then necessary conditions for the functional  (Equation89) achieves an extremum at a pair (q,S) are (90) L(T,q(S)+aRHDbμ,βq(S))=0,(90) (91) LqiaRHDbμ,1βL aRHDbμ,βq+12(tDSμ,1βtDbμ,1β)×L aRHDbμ,βq=0,(91) t[a,S], (92) (aDtμ,1βSDtμ,1β)L aRHDbμ,βq=0,t[S,b].(92) If β=0, then the following transversality condition should be hold (93) aIt1μL aRCDbμq|t=b= SIt1μz(t)|t=b.(93) If β=1, then the following transversality condition should be hold (94) L aRHDbμ,βq(t)|t=a=L aRHDbμ,βq(t)|t=S=0.(94)

Proof.

Let ϵ>0 and define a family of curves q(t)=q(t)+ϵυ(t), where ν is an arbitrary continuously differentiable functions for which ν(a)=0. Let T be a positive real number. Then the function (95) J[q+ϵν,S+ϵT]=aS+ϵTL(t,q(t)+ϵν(t),aRHDbμ,βq(t)+ϵaRHDbμ,βν(t))dt,(95) depends on ε only. Since J admits an extremum at (q,S) then the necessary condition for which J[ϵ] achieves a minimum, is (96) dJdϵ(0)=0.(96) Applying Lieibniz integral rule we get

(97) dJ[ϵ]dϵ=△T L(S+ϵT,q(S+ϵT)+ϵν(S+ϵT),aRHDbμ,βq(S+ϵT)+ϵaRHDbμ,βν(S+ϵT))+aS+ϵTLqν(t)+L aRHDbμ,βq(aRHDbμ,βν)(t)dt.(97)

From (Equation97) and (Equation96) one obtains (98) 0=△T L(S,q(S),aRHDbμ,βq(S))+aSLqν(t)+L aRHDbμ,βq(aRHDbμ,βν)(t)dt.(98) Since setting ε equal to zero is equivalent to replacing qi and aRDbμqi by qi and aRDbμqi, the last equation becomes (99) 0=△T L(S,q(S),aRDbμ,βq(S))+aSLqiν(t)+L aRHDbμ,βq(aRHDbμ,βν(t))dt.(99) Now, to simplify the notations we put z(t)=L/( aRHDbμ,βq). Then, applying relation (Equation20) to get (100) aSz(t)aRHDbμ,βν(t)dt=aS ν(t) aRHDbμ,1βz(t)dt+12aSν(t)(tDSμ,1βtDbμ,1β)z(t))dtSbν(t)(aDtμ,1βSDtμ,1β)z(t)dt+tIS(1β)(1μ)z(t).aItβ(1μ)ν(t)|aS+aIt(1β)(1μ)z(t).tIbβ(1μ)ν(t)|abSIt(1β)(1μ)z(t).tIbβ(1μ)ν(t)|Sb.(100) It follows from (Equation99) and (Equation100) that (101) T L(S,q(S),aRDbμq(S))+aS ν(t)Lqi aRHDbμ,1βz(t)dt+12[aSν(t)(tDSμ,1βtDbμ,1β)z(t))dtSbν(t)(aDtμ,1βSDtμ,1β)z(t)dt+tIS(1β)(1μ)z(t).aItβ(1μ)ν(t)|aS+aIt(1β)(1μ)z(t).tIbβ(1μ)ν(t)|ab SIt(1β)(1μ)z(t).tIbβ(1μ)ν(t)|Sb]=0.(101) Remark that If β(0,1), then by the continuity of ν and z=L/( aRHDbμ,βqi) one obtains (102) tIS(1β)(1μ)z(t).aItβ(1μ)ν(t)|aS=aIt(1β)(1μ)z(t).tIbβ(1μ)ν(t)|ab=SIt(1β)(1μ)z(t).tIbβ(1μ)ν(t)|Sb=0.(102) If β=0, then from the assumption ν(a)=0 and (Equation93), we get (103) tIS(1β)(1μ)z(t).aItβ(1μ)ν(t)|aS=tIS1μz(t).ν(t)|aS=0,(103) (104) aIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)ν(t)|ab=aIt1μL aRCDbμqi.ν(t)|ab=aIt1μL aRCDbμqi.ν(t)|t=b=0,(104) and (105) SIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)ν(t)|Sb=SIt1μL aRCDbμqi.ν(t)|Sb=SIt1μL aRCDbμqi.ν(t)|t=b=0.(105) If β=1, it follows from (Equation94) that (106) tIS(1β)(1μ)L aRHDbμ,βqi.aItβ(1μ)ν(t)|aS=L aRDbμqi aIt1μν(t)|t=S=0,(106) (107) aIt(1β)(1μ)L aRHDbμ,βqi.tIbβ(1μ)νi(t)|ab=L aRHDbμ,1qi.tIb1μν(t)|ab=L aRDbμqi.tIb1μν(t)|t=a=0,(107) and (108) SIt(1β)(1μ)L_aRHDbμ,βqi.tIbβ(1μ)ν(t)|Sb=L aRHDbμ,1qi.tIb1μν(t)|Sb=L aRDbμqi.tIb1μν(t)|t=S=0.(108) Using equations (Equation102)–(Equation108), equation (Equation101) becomes (109) T L(S,q(S),aRDbμq(S))+aS ν(t)Lqi aRHDbμ,1βz(t)dt+12aSν(t)(tDSμ,1βtDbμ,1β)z(t)dtSbν(t)(aDtμ,1βSDtμ,1β)z(t)dt=0.(109) Since T is arbitrary, then if we choose, in (Equation109), η(t)=0 on [a,b] we get (Equation90). If ν(t)=0 on [S,b] and free on [a,S), one obtains (Equation91). If ν(t)=0 on [a,S] and free on (S,b], we arrive to (Equation92).

Remark 6.1

If we put β=0 in the previous theorem, then we obtain Theorem 8 in [Citation24].

7. The Noether theorem in the sense of Riesz–Hilfer

In this section, we give a version of Noether theorem for Riesz–Hilfer derivative. Firstly, we give the following lemma.

Lemma 7.1

Assume that the functional (110) J[q]=abL(t,q(t),aRHDbμ,βq(t))dt,(110) satisfies the condition (111) cdL(t,q(t),aRHDbμ,βq(t))dt=cdL(t,q~(t),aRHDbμ,βq~(t))dt,(111) where (112) q~(t)=ϵϕ(t,q(t))+o(ϵ)(112) Then (113) Lq(t).ϕ(t,q(t))+LaRHDbμ,βq(aRHDbμ,βϕ(t,q(t)))=0,t[a,b].(113)

Proof.

Since (Equation111) is valid for any subinterval of [a,b], if follows that (114) L(t,q(t),aRHDbμ,βq(t))L(t,q~(t),aRHDbμ,βq~(t))=0,t[a,b].(114) By differentiating equation (Equation114) with respect to ε and then putting ϵ=0 we get (115) Lq~.ϕ(t,q(t))+LaRHDbμ,βq~(aRHDbμ,βϕ(t,q(t)))=0,t[a,b].(115) Since setting ϵ=0 is equivalent to q=q~, hence equation (Equation115) leads to (Equation113).

In the following, we give a version of Noether theorem for Riesz–Hilfer derivative.

Theorem 7.1

If the functional (Equation110) satisfies equation (Equation111), where q~(t) is given by (Equation112), then for any t[a,b] (116) aRHDbμ,1βL aRHDbμ,βqi(t).ϕ(t,q(t))+LaRHDbμ,βq(aRHDbμ,βϕ(t,q(t)))=0.(116)

Proof.

According to Theorem 3.1, the function q should satisfy the Euler–Lagrange equation: (117) Lq(t)=aRHDbμ,1βL aRHDbμ,βq(t),t[a,b].(117) Replacing (Equation117) in (Equation113) we obtain (Equation116).

8. Applications

In this section we give many examples to illustrate our obtained results.

Example 8.1

Consider the Lagrangian function (118) L(t,q(t),q(t),(0RHD1μ,0q)(t)=12 [q(t)]2+(t2t)(0RHD1μ,0q)(t),(118) where t[0,1].Let (119) J[q]=01 L(t,q(t),q(t),(0RHD1μ,0q)(t)))dt.(119) We will find qsuch that J[q] is an extremum and satisfy the boundary condition: (120) q(0)=q(1)=12Γ(4μ)1Γ(5μ).(120) By applying Theorem 3.1, the function q should satisfy the following Euler–Lagrange equation: (121) ddtLq(t)+0RHD1μ,1L 0RHD1μ,0q(t)=0,for any t[0,1].(121) Then (122) d2d2q(t)+0RHD1μ,1(t2t)=0,forany t[0,1],(122) and this is equivalent to (123) d2d2q(t)+0RD1μ(t2t)=0,forany t[0,1].(123) Now according to property (2.1), p. 71, in [Citation1] we get (124) 0RD1μ(t2t)=12[0Dtμ(t2t)tD1μ(t2t)]=12[0Dtμ(t2t)tD1μ((1t)2(1t))]=12[0Dtμt20DtμttD1μ((1t)2+tD1μ(1t)]=122Γ(3μ)t2μ1Γ(2μ)t1μ2Γ(3μ)(1t)2μ+1Γ(2μ)(1t)1μ.(124) It yields from (Equation123) and (Equation124) that for any t[0,1], (125) d2d2q(t)=t2μΓ(3μ)t1μ2Γ(2μ)(1t)2μΓ(3μ)+(1t)1μ2Γ(2μ).(125) By integrating twice both side of this equation we get (126) q(t)=t4μΓ(5μ)t3μ2Γ(4μ)(1t)4μΓ(5μ)+(1t)3μ2Γ(4μ).+c1t+c2.(126) Using the boundary condition q(0)=1/2Γ(4μ)1/Γ(5μ), it follows that c2=0. Then (127) q(t)=t4μΓ(5μ)t3μ2Γ(4μ)(1t)4μΓ(5μ)+(1t)3μ2Γ(4μ)+c1t.(127) By applying the boundary condition, q(1)=1/Γ(5μ)1/2Γ(4μ) we get c1=0. Therefore, (128) q(t)=t4μΓ(5μ)t3μ2Γ(4μ)(1t)4μΓ(5μ)+(1t)3μ2Γ(4μ).(128) The graph of the function q is clarified in Figure (a) for different values of μ.

Figure 1. (a) The graph of the solution function given by (Equation128) for different values of μ (b) The graph of the solution function given by (Equation131), (Equation136) and (Equation137) for different values of μ.

Figure 1. (a) The graph of the solution function given by (Equation128(128) q∗(t)=t4−μΓ(5−μ)−t3−μ2Γ(4−μ)−(1−t)4−μΓ(5−μ)+(1−t)3−μ2Γ(4−μ).(128) ) for different values of μ (b) The graph of the solution function given by (Equation131(131) p(s)=ϕ(s)=34Γ(6−μ)−14Γ(4−μ)−16s,for s∈34,1,(131) ), (Equation136(136) p(s)=3(s−1)4−μ2Γ(5−μ)−3(s−1)5−μ4Γ(6−μ)−(s−1)3−μ2Γ(4−μ)+3(2−s)5−μ4Γ(6−μ)−(2−s)3−μ4Γ(4−μ)−s36−(s−1)1Γ(5−μ)344−μ−1Γ(6−μ)345−μ−23Γ(4−μ)343−μ+1Γ(6−μ)145−μ−13Γ(4−μ)143−μ+2(s−1)9743,s∈1,74+2(s−1)9743,s∈1,74,(136) ) and (Equation137(137) p(s)=123Γ(3)Γ(5−μ)(s−1)4−μ−Γ(4)Γ(6−μ)(s−1)5−μ−2Γ(2)Γ(4−μ)(s−1)3−μ+Γ(4)Γ(6−μ)(2−s)5−μ−Γ(2)Γ(4−μ)(2−s)3−μ−s36+2(s−2)3Γ(3)Γ(5−μ)344−μ−Γ(4)Γ(6−μ)×345−μ−2Γ(2)Γ(4−μ)343−μ+Γ(4)Γ(6−μ)145−μ−Γ(2)Γ(4−μ)143−μ−(s−2)47433,s∈74,2.(137) ) for different values of μ.

Example 8.2

Let r=14,a=1,b=2, and consider the functional (129) F[p]=abL(s,p(s), aRHDbμ,0p(s),p(sr),p(sr))ds,(129) where for s[1,2] (130) L(s,p(s),bRHDbμ,0p(s),p(s),p(sr),p(sr))=sp(s)+(6s2s311s+6)aRHDbμ,0p(s)12[p(s)]2+s14p(sr)12ps142.(130) We find a function p such that F[p] is an extremum and satisfies the boundary conditions: (131) p(s)=ϕ(s)=34Γ(6μ)14Γ(4μ)16s,for s34,1,(131) and (132) p74=0,p(2)=3Γ(5μ)32Γ(6μ)1Γ(4μ)43.(132) By applying Theorem 5.1 the function pmust satisfy the following conditions: (133) 2p(s)+2s=1RHD2μ,1(6s2s311s+6),(133) for s1,74 and (134) p(s)+s=1RHD2μ,1(6s2s311s+6),(134) for s74,2. Observe that, according to property (2.1) in [Citation1] it yields (135) 1RHD2μ,1(6s2s311s+6)=1RD2μ(6s2s311s+6)=12[1Dsμ(6s2s311s+6)sD2μ(6s2s311s+6)]=12[1Dsμ(3(s1)2(s1)32(s1))]+12[sD2μ((2s)3(2s))]=3(s1)2μΓ(3μ)3(s1)3μ2Γ(4μ)(s1)1μΓ(2μ)+3(2s)3μ2Γ(4μ)(2s)1μ2Γ(2μ).(135) Inserting the expression (Equation135) into (Equation133) and (Equation134) and taking into account the boundary conditions (Equation131) and (Equation122), we get after some manipulations (136) p(s)=3(s1)4μ2Γ(5μ)3(s1)5μ4Γ(6μ)(s1)3μ2Γ(4μ)+3(2s)5μ4Γ(6μ)(2s)3μ4Γ(4μ)s36(s1)1Γ(5μ)344μ1Γ(6μ)345μ23Γ(4μ)343μ+1Γ(6μ)145μ13Γ(4μ)143μ+2(s1)9743,s1,74+2(s1)9743,s1,74,(136) and (137) p(s)=123Γ(3)Γ(5μ)(s1)4μΓ(4)Γ(6μ)(s1)5μ2Γ(2)Γ(4μ)(s1)3μ+Γ(4)Γ(6μ)(2s)5μΓ(2)Γ(4μ)(2s)3μs36+2(s2)3Γ(3)Γ(5μ)344μΓ(4)Γ(6μ)×345μ2Γ(2)Γ(4μ)343μ+Γ(4)Γ(6μ)145μΓ(2)Γ(4μ)143μ(s2)47433,s74,2.(137) The graph of the function p is clarified in Figure (b) for different values of μ.

Remark 8.1

According to Remark 2.2, if β=0, then aRHDbμ,βq(t)=aRCDbμq(t);t[a,b], and hence, when μ=1, we get aRHDbμ,βq(t)=dq/d;t[a,b]. So, the figures (a) and (b) emphasize that when μ approach to the value one, the solution function curve approaches to the solution function curve if the Riesz–Hilfer derivative is replaced by the first drivative.

Example 8.3

Fractional Lagrangian for RLC.

In this example we consider a simple loop current that is described by a fractional Lagrangian. We assume that this single loop circuit involves a capacitor C, a resistor R and an inductor χ. The fractional Lagrangian for this loop take the form (138) L=χ2dqdt212cq2(t)+iγR2 bRHDbμ,βq(t),(138) where, q is the charge. According to Theorem 3.1, Euler–Lagrange equation corresponding to (138) is (139) qc+χd2qd2+bRHDbμ,βiγR2(t)=0.(139)

Example 8.4

Let r(0,ba), h:[a,b]R be such that the function z(t)=aRHDbμ,βh(t) is continuously differentiable. Consider the functional (140) J[q]=ab12q2(t)+h(t)aRHDbμ,1q(t)+12[q(tr)]2dt,(140) with the boundary conditions (141) q(t)=ϕ(t), t[ar,a]andq(b)=qb.(141) According to Theorem 5.1, the functional J has an extremum at a function q if q satisfies: (142) q(t)aRCDbμh(t)q(tr)=0,for t[a,br],(142) and (143) q(t)aRCDbμh(t)=0,for t[br,b].(143)

Remark 8.2

Like in many papers, see for example [Citation21], the numerical methods are more suitable to find the solutions of FVP.

9. Results and discussion

As mentioned earlier, variational problems and fractional calculus have many applications in different branches in engineering and mathematics, moreover, the Riesz–Hilfer fractional derivative (RHFD) is a generalization for the Riesz–Riemann–Liouville and the Riesz–Caputo derivative. The results that we obtained are to find Euler–Lagrange equations for various of fractional variational problems with the Lagrangian function containing (RHFD), and hence our results generalize many recent papers in the literature, for example, [Citation22–25, Citation36]. On other hand, our technique allows to generalize some works, such as the obtained results in [Citation26] to the case when the functional involving (RHFD). As we mentioned in Corollary 2.1, relations (Equation20) and (Equation21) in [Citation22] are particular cases of Lemma 2.1, and if β=0 and i = 1 in Theorem 3.1, then we obtain Theorem 3.1, in [Citation22]. Moreover, If we put β=0 in both Theorems 3.2 and 6.1, we obtain Theorems 4.1 and 7.1, respectively, in [Citation24].

10. Conclusion

Euler–Lagrange equations for different kind of fractional variational problems with the Lagrangian function containing the Riesz–Hilfer fractional derivative are obtained. Since the Riesz–Hilfer fractional derivative is a generalization for the Riesz–Riemann–Liouville and the Riesz–Caputo derivative, then our results generalize many recent works in which the Lagrangian function involving the Riesz–Riemann–Liouville or the Riesz–Caputo derivative. Fractional variational problem in the presence of delay derivatives is considered. Moreover, a version for Noether theorem in the Riesz–Hilfer sense is established. Necessary conditions for a pair function-time to be an optimal solution to the problem are investigated. Examples are given to illustrate the applicability of the obtained results. Furthermore, our obtained results generalize some existing results such as Theorem 1 in [Citation22] and Theorems 3 and 6 in [Citation24]. Also, the technique used in the present paper enable to extend the results in [Citation26, Citation28, Citation31] when the treated problems in these works involving Riesz–Hilfer fractional derivative. Moreover, this work, may be, encourages to study partial differential equations containing Riesz–Hilfer fractional derivative.

Acknowledgments

The authors acknowledge the Deanship of Scientific Research at King Faisal University for financial support under the Research Group Support Track. (Grant No. 1811003).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The authors acknowledge the Deanship of Scientific Research at King Faisal University for financial support under the Research Group Support Track [grant number 1811003].

References

  • Kilbas AA, Srivastava HM, Trujillo JJ, Theory and applications of fractional differential equations, Amsterdam: Elsevier Science Publishers BV; 2006. (North holland mathematics studies; vol. 204).
  • Miller KS, Ross B. An introduction to the fractional calculus and fractional differential equations. New York: Wiley; 1993.
  • Podlubny I. Fractional differential equations. San Diego (CA): Academic Press; 1999.
  • Smako SG, Kilbas AA, Marichev OI. Fractional integral and derivativies-theory and applications. Longhorne (PA): Gordan and Breach; 1993.
  • Ghomanjani F. A new approach for solving fractional differential algebric equations. J Taibah Univ Sci. 2017;11:1158–1164.
  • Ibrahim AG. Differential equations and inclusions of fractional order with impulse effect in Banach spaces. Bull Malays Math Sci Soc. 2020;43(1):69–109.
  • Moghaddam BP, Mostaghim ZS. A numercial method based on finite difference for solving fractional delay differential equations. J Taibah Univ Sci. 2013;7:120-127.
  • Lukashchuk SY. Approximate conservation laws for fractional differential equations. Commun Nonlinear Sci Numer Simul. 2019;68:147–159.
  • Wang JR, Ibrahim AG, O'Regan D, et al. A general class of noninstantaneous impulsive fractional differential inclusions in Banach spaces. Adv Differ Equ. 2017;2017(1):1–28. DOI:10.1186/s13662-017-1342-8
  • Wang JR, Ibrahim AG, O'Regan D. Hilfer- type fractional differential switched inclusions with noninstantaneous impulsive and nonlocal conditions. Nonlinear Anal Model Control. 2018;23(6):921–941.
  • Wang JR, Ibrahim AG, O'Regan D. Controllability of noninstantaneous impulsive semilinear functional differential inclusions without compactness. Indag Math New Ser. 2018;29:1362–1392.
  • Wang JR, Ibrahim AG, O'Regan D. Controllability of fractional evolutions inclusions with noninstantaneous impulsive. Int J Nonlinear Sci Numer Simul. 2018;19(3-4):321–334.
  • Wang JR, Ibrahim AG, O'Regan D. Global attracting solutions to Hilfer fractional differential inclusions of Sobolev type with noninstantaneous impulses and nonlocal conditions. Nonlinear Anal Model Control. 2019;24(5):775–803.
  • Wang JR, Ibrahim AG, O'Regan D. Nonemptyness and compactness of the solution set for fractional evolution inclusions with non-instantaneous impulses. Electron J Differ Equ. 2019;37:1–17.
  • Bliss GA. Lectures on the calculus of variations. University of Chicago Press; 1963.
  • Gelfand IM, Fomin SV. Calculus of variations. Prenticc-Hail; 1963.
  • Hestenes MR. Calculus of variations and optimal control theory. John Wiley & Sons; 1966.
  • Weinstock R. Calculus of variations with applications to physics and engineering. New-York: Dover publications, Inc.; 1974.
  • Riewe F. Nonconservative Lagrangian and Hamiltonian mechanics. Phys Rev E. 1966;53(2):1890–1899.
  • Riewe F. Mechanics with fractional derivatives. Phys Rev E. 1997;55(3):3581–3592.
  • Agrawal OP. Formulation of Euler–Lagrange equations for fractional variational problems. J Math Anal Appl. 2002;272:368–379.
  • Agrawal OP. Fractional variational calculus in terms of Riesz fractional derivatives. J Phys A Math Theor. 2007;40:6287–6303.
  • Almeida R, Torres DFM. Necessary and sufficient conditions for the fractional calculus of variations with Caputo derivatives. Commun Nonlinear Sci Numer Simul. 2011;16(3):1490–1500.
  • Almeida R. Fractional variational problems with the Riesz–Caputo derivative. Appl Math Lett. 2012;25:142–148.
  • Odzijewicz T, Malinowska AB, Torres DFM. Fractional variational calculus with classical and combined Caputo derivatives. Nonlinear Anal. 2012;75:1507–1515.
  • Sayev K, Rostami MR, Attari HS. A new study on delay fractional variation problems. Int J Comput Math. 2018;95(6-7):1170–1194.
  • Tavares D, Almeida R, Torres DFM. Optimality conditions for fractional variational problems with dependence on a combined Caputo derivative of.variable order. Optimization. 2015;64(6):1381–1391.
  • Almeida R, Morgado ML. The Euler–Lagrange and Legendre equations for functionals involving distributed-order fractional derivatives. Appl Math Comput. 2018;331:394–403.
  • Agrawal OP. Generalized variational problems and Euler–Lagrange equations. Comput Math Appl. 2010;59:1852–1864.
  • Almeida R, Pooseh S, Torres DFM. Computational methods in the fractional calculus of variations. London: Imp. Coll. Press; 2015.
  • Askari H, Ansari A. Fractional calculus of variations with a generalized fractional derivative. Fract Differ Calc. 2016;6:57–72.
  • Atanacković TM, Pilipović S. Variational problems with fractional derivatives: Euler–Lagrange equations. J Phys A Math Theor. 2011;41(9):195–201.
  • Baleanu D. About fractional quantization and fractional variational principles. Commun Nonlinear Sci Numer Simul. 2009;14(6):2520–2523.
  • Baleanu D, Trujillo JI. A new method of finding the fractional Euler–Lagrange and Hamilton equations within Caputo fractional derivatives. Common Nonlinear Sci Numer Simul. 2010;15(5):1111–1115.
  • Herzallah MAE, Baleanu D. Fractional-order Euler–Lagrange equations and formulation of Hamiltonian equations. Nonlinear Dyn. 2009;58(1-2):385–391.
  • Jarad F, Abdeljawad T, Baleanu D. Fractional variational principles with delay within Caputo derivatives. Rep Math Phys. 2010;65(1):17–28.
  • Malinowska AB, Torres DFM. Generalized natural boundary conditions for fractional variational problems in terms of the Caputo derivative. Comput Math Appl. 2010;59(9):3110–3116.
  • Malinowska AB, Torres DFM. Fractional variational calculus in terms of a combined Caputo derivative. In: Proceedings IFAC Workshop on Fractional Derivative and Applications IFAC FDA2010. Badajoz, Spain: University of Extremadura; 2010. p. 18–20.
  • Tavares D, Almeida R, Torres DFM. Constrained fractional variational problems of variable order. IEEE/CAA J Autom Sin. 2017;1(4):65–73.
  • Tavares D, Almeida R, Torres DFM. Combined fractional variational problems of variable order and some computational aspects. J Comput Appl Math. 2018;339:374–388.
  • Hilfer R. Applications of fractional calculus in physics. Singapore: World Scientific; 1999.
  • Agrawal OP, Muslih SI, Baleanu D. Generalized variational calculus in terms of multiparamaters fractional derivatives. Commun Nonlinear Sci Numer Simul. 2010;16:4756–4767.
  • Ali A, Seadawy AR, Lu D. Computational methods and traveling wave solutions for the fourth-order nonlinear Ablowitz-Kaup-Newell-Segur water wave dynamical equation via two methods and its applications. Open Phys. 2018;16:219–226.
  • Iqbal M, Seadawy AR, Lu D. Construction of solitary wave solutions to the nonlinear modified Kortewege-de Vries dynamical equation in unmagnetized plasma via mathematical methods. Mod Phys Lett A. 2018;33(32):1850183.
  • Iqbal M, Seadawy AR, Lu D. Dispersive solitary wave solutions of nonlinear further modified Korteweg–de Vries dynamical equation in an unmagnetized dusty plasma. Mod Phys Lett A. 2018;33(37):1850217.
  • Iqbal M, Seadawy AR, Lu D. Construction of bright–dark solutions and ion-acoustic solitary wave solutions of dynamical system of nonlinear wave propagation. Mod Phys Lett A. 2019;34(37):1950309.
  • Iqbal M, Seadawy AR, Lu D, et al. Construction of a weakly nonlinear dispersion solitary wave solution for the Zakharov–Kuznetsov-modified equal width dynamical equation. Indian J Phys. 2019;21:1–10.
  • Iqbal M, Seadawy AR, Khalil OH, et al. Propagation of long internal waves in density stratified ocean for the (2+1)-dimensional nonlinear Nizhnik-Novikov-Vesselov dynamical equation. Res Phys. 2020;16:102838.
  • Iqbal M, Seadawy AR, Lu D. Applications of nonlinear longitudinal wave equation in a magneto-electro-elastic circular rod and new solitary wave solutions. Mod Phys Lett B. 2019;33(18):1950210.
  • Lu D, Seadawy AA, Iqbal M. Mathematical methods via construction of traveling and solitary wave solutions of three coupled system of nonlinear partial differential equations and their applications. Res Phys. 2018;11:1161–1171.
  • Nasreen N, Seadawy AR, Lu D, et al. Dispersive solitary wave and soliton solutions of the generalized third order nonlinear Schrodinger dynamical equation by modified analytical method. Res Phys. 2019;15:102641.
  • Seadaway AR, Lu D, Yue C. Travelling wave solutions of the generalized nonlinear fifth-order KDV water wave equations and its stability. J Taibah Univ Sci. 2017;11:623–633.
  • Seadawy AR, Iqbal M, Lu D. Nonlinear wave solutions of the kudryashov–Sinelshchikov dynamical equation in mixtures liquid-gas bubbles under the consideration of heat transfer and viscosity. J Taibah Univ Sci. 2019;13:1060–1072.
  • Seadawy AR, Iqbal M, Lu D. Propagation of kink and anti-kink wave solutions for the nonlinear damped modified Korteweg–de Vries equation arising in ion-acoustic wave in an unmagnetized collisional dusty plasma. Phys A. 2020;544:123560.
  • Seadawy AR, Iqbal M, Lu D. Construction of soliton solutions of the modify unstable nonlinear Schrodinger dynamical equation in fiber optics. Indian J Phys. 2019:1–10.
  • Seadawy AR, Lu D, Iqbal M. Application of mathematical methods on the system of dynamical equations for the ion sound and Langmuir waves. Pramana J Phys. 2019;93(1):10.
  • Seadawy AR, Iqbal M, Lu D. Applications of propagation of long-wave with dissipation and dispersion in nonlinear media via solitary wave solutions of generalized Kadomtsev–Petviashvili modified equal width dynamical equation. Comput Math Appl. 2019;78:3620–3632.