Variational methods for fractional q-Sturm–Liouville Problems Zeinab S.I. Mansour 6 1 Department of Mathematics, Faculty of Science, King Saud University,Riyadh 0 2 n a J 5 Abstract 2 Inthis paper,weformulatearegularq-fractionalSturm–Liouvilleproblem(qF- ] SLP) which includes the left-sided Riemann–Liouville and the right-sided Ca- A puto q-fractional derivatives of the same order α, α (0,1). We introduce C the essential q-fractional variational analysis needed in∈ proving the existence . of a countable set of real eigenvalues and associated orthogonal eigenfunctions h t for the regular qFSLP when α > 1/2 asscociated with the boundary condition a y(0) = y(a) = 0. A criteria for the first eigenvalue is proved. Examples are m included. These results are a generalization of the integer regular q-Sturm– [ Liouville problem introduced by Annaby and Mansour in[1]. 1 Keywords: Left and right sided Riemann–Liouville and Caputo q-derivatives, v 8 eigenvalues and eigenfunctions, q-fractional variational calculus. 9 2000 MSC: 39A13, 26A33, 49R05. 4 1 0 . 2 0 1. Introduction 6 1 In the joint paper of Sturm and Liouville [2], they studied the problem : v i d dy X p +r(x)y(x)=λwy(x), x [a,b], (1.1) − dx dx ∈ r (cid:18) (cid:19) a with certainboundary conditions at a and b. Here, the functions p, w are posi- tive on[a,b] and r is a realvalued function on [a,b]. They provedthe existence ofnon-zerosolutions(eigenfunctions)onlyforspecialvaluesofthe parameterλ which is called eigenvalues. For a comprehensive study for the contribution of SturmandLiouvilletothetheory,see[3]. Recently,manymathematicianswere interestedinafractionalversionof (1.1),i.e. whenthe derivativeisreplacedby a fractional derivative like Riemann–Liouville derivative or Caputo derivative, see [4–9]. Iterative methods, variational method, and the fixed point theory Email address: [email protected], [email protected] (ZeinabS.I.Mansour) Preprint submitted toElsevier February 5, 2016 are three different approaches used in proving the existence and uniqueness of solutionsofSturm–Liouvilleproblems,c.f.[3,10,11]. Thecalculusofvariations has recently developed to calculate extremum of functional contains fractional derivatives,whichiscalledfractionalcalculusofvariations,seeforexample[12– 19]. In [4], Klimek et al. applied the methods of fractional variational calculus toprovetheexistenceofacountablesetoforthogonalsolutionsandcorrespond- ingeigenvalues. In[1]AnnabyandMansourintroducedaq-versionof (1.1),i.e., when the derivative is replaced by Jackson q-derivative. Their results are ap- plied and developed in different aspects, for example, see [20–25]. Throughout this paper q is a positive number less than 1. The set of non negative integers is denoted by N , and the set of positive integers is denoted by N. For t>0, 0 A := tqn : n N , A∗ :=A 0 , q,t { ∈ 0} q,t q,t∪{ } and := tqn : n N . q,t 0 A {± ∈ } When t = 1, we simply use A , A∗, and to denote A , A∗ , and , q q Aq q,1 q,1 Aq,1 respectively. We follow [26] for the definitions and notations of the q-shifted factorial, the q-gamma and q-beta functions, the basic hypergeometric series, andJacksonq-difference operatorandintegrals. A setA is calleda q-geometric set if qx A whenever x A. Let X be a q-geometric set containing zero. A ∈ ∈ function f defined on X is called q-regular at zero if lim f(xqn)=f(0) for allx X. n→∞ ∈ Let C(X) denote the space of all q-regular at zero functions defined on X with values in R. C(X) associated with the norm function f =sup f(xqn) : x X, n N , 0 k k {| | ∈ ∈ } is a normed space. The q-integration by parts rule [27] is b b f(x)D g(x)=f(x)g(x)b + D f(x)g(qx)d x, a,b X, (1.2) q |a q q ∈ Za Za and f, g are q-regular at zero functions. For p>0, and Y is A or A∗ , the space Lp(Y) is the normed space of all q,t q,t q functions defined on Y such that t 1/p f := f(u)pd u < . k kp | | q ∞ (cid:18)Z0 (cid:19) If p=2, then L2(Y) associated with the inner product q t f,g := f(u)g(u)d u (1.3) q h i Z0 2 is a Hilbert space. By a weighted L2(Y,w) space is the space of all functions f q defined on Y such that t f(u)2w(u)d u< , q | | ∞ Z0 where w is a positive function defined onY. L2(Y,w) associatedwith the inner q product t f,g := f(u)g(u)w(u)d u q h i Z0 is a Hilbert space. The space of all q-absolutely functions on A∗ is denoted by q,t C (A∗ )anddefinedasthespaceofallq-regularatzerofunctionsf satisfying A q q,t ∞ f(uqj) f(uqj+1) K for allu A∗ , | − |≤ ∈ q,t j=0 X and K is a constantdepending on the function f, c.f. [27, Definition 4.3.1]. I.e. ∗ ∗ C (A ) C (A ). A q q,t ⊆ q q,t The space C(n)(A∗ ) (n N) is the space of all functions defined on X such that fA, Dqf, ..q.,,tDn−1∈f are q-regular at zero and Dn−1f C (A∗ ), q q q ∈ A q q,t c.f. [27, Definition 4.3.2]. Also it is proved in [27, Theorem 4.6] that a function f C(n)(A∗ ) if and only if there exists a function φ L1(A∗ ) such that ∈A q q,t ∈ q q,t n−1 Dkf(0) xn−1 x f(x)= Γ (qk+1)xk+ Γ (n) (qu/x;q)n−1φ(u)dqu, x∈A∗q,t. k=0 q q Z0 X In particular, f C(A∗ ) if and only if f is q-regular at zero such that D f L1(A∗ ). ∈It Ais wortqh,t noting that in [27], all the definitions and results q ∈ q q,t we have just mentioned are defined and proved for functions defined on the interval [0,a] instead of A∗ . In [28], Mansour studied the problem q,t Dα p(x)cDα y(x)+(r(x) λw (x))y(x)=0, x A∗ , (1.4) q,a− q,0+ − α ∈ q,a where p(x) = 0 and w > 0 for all x A∗ , p,r, w are real valued functions defined in A6∗ and theαassociated bou∈ndaqr,ya conditioαns are q,a c y(0)+c I1−αpcDα y (0)=0, (1.5) 1 2 q,a− q,0+ h i a d y(a)+d I1−αpcDα y ( )=0, (1.6) 1 2 q,a− q,0+ q h i withc2+c2 =0andd2+d2 =0. itisprovedthattheeigenvaluesarerealandthe 1 2 6 1 2 6 eigenfunctions associated to different eigenvalues are orthogonal in the Hilbert space L2(A∗ ,w ). A sufficient condition on the parameter λ to guarantee q q,a α 3 the existence and uniqueness of the solution is introduced by using the fixed point theorem, also a condition is imposed on the domain of the problem in order to prove the existence and uniqueness of solution for any λ. This paper is organized as follows. Section 2 is on the q-fractional operators and their properties which we need in the sequel. Cardoso [29] introduced basic Fourier seriesfor functions defined ona q-lineargridofthe form qn : n N 0 . 0 {± ∈ }∪{ } In Section 3, we reformulate Cardoso’s results for functions defined on a q- linear grid of the form aqn : n N 0 . In Section 4, we introduce a 0 {± ∈ }∪{ } fractional q-analogue for Euler–Lagrange equations for functionals defined in termsofJacksonq-integrationandthe integrandcontainsthe left sidedCaputo fractional q-derivative. We also introduce a fractional q-isoperimetric problem. InSection5,weusethevariationalq-calculusdevelopedinSection4toprovethe existenceofacountablenumberofeigenvaluesandorthogonaleigenfunctionsfor the fractional q-Sturm–Liouville problem with the boundary condition y(0) = y(a) = 0. We also define the Rayleigh quotient and prove a criteria for the smallest eigenvalue. 2. Fractional q-Calculus Thissectionincludesthedefinitionsandpropertiesoftheleftsidedandright sided Riemann–Liouville q-fractional operators which we need in our investiga- tions. The left sided Riemann–Liouville q-fractional operator is defined by xα−1 x Iqα,a+f(x)= Γ (α) (qt/x;q)α−1f(t)dqt. (2.1) q Za This definition is introduced by Agarwal in [30] when a = 0 and by Rajkovi´c et.al [31] for a=0. The right sided Riemann–Liouville q-fractionaloperatorby 6 1 b Iqα,b−f(x)= Γ (α) tα−1(qx/t;q)α−1f(t)dqt, (2.2) q Zqx see [28]. The left sided Riemann–Liouville q-fractional operator satisfies the semigroup property Iα Iβ f(x)=Iα+βf(x). q,a+ q,a+ q,a+ The case a=0 is proved in [30] and the case a>0 is proved in [31]. The right sided Riemann–Liouville q-fractional operator satisfies the semi- group property [28] Iα Iβ f(x)=Iα+βf(x), x A∗ , (2.3) q,b− q,b− q,b− ∈ q,b for any function defined on A and for any values of α and β. q,b Forα>0andpαq=m,theleftandrightsideRiemann–Liouvillefractional q-derivatives of order α are defined by m 1 Dqα,a+f(x):=DqmIqm,a−+αf(x), Dqα,b−f(x):= −q Dqm−1Iqm,b−−αf(x), (cid:18) (cid:19) 4 theleftandrightsidedCaputofractionalq-derivativesoforderαaredefinedby m 1 cDqα,a+f(x):=Iqm,a−+αDqmf(x), cDqα,b− := −q Iqm,b−−αDqm−1f(x). (cid:18) (cid:19) see [28]. From now on, we shall consider left sided Riemann–Liouville and Caputo fractional q-derivatives when the lower point a = 0 and right sided Riemann–Liouville and Caputo fractional q-derivatives when b = a. According to [27, pp. 124,148], Dα f(x) exists if q,0+ f L1(A∗ ) such that Im−αf C(m)(A∗ ), ∈ q q,a q,0+ ∈A q q,a and cDα f exists if q,a+ f C(m)(A∗ ). ∈A q q,a The following proposition is proved in [28] Proposition 2.1. Let α (0,1). ∈ (i) If f L1(A∗ ) such that Iα f C (A∗ ) then ∈ q q,a q,0+ ∈A q q,a Iα f(0) cDα Iα f(x)=f(x) q,0+ x−α. (2.4) q,0+ q,0+ − Γ (1 α) q − Moreover, if f is bounded on A∗ then q,a cDα Iα f(x)=f(x). (2.5) q,0+ q,0+ (ii) For any function f defined on A∗ q,a a−α a cDqα,a−Iqα,a−f(x)=f(x)− Γ (1 α)(qx/a;q)−α Iqα,a−f (q). (2.6) q − (cid:16) (cid:17) (iii) If f L1(A ) then ∈ q q,a Dα Iα f(x)=f(x). (2.7) q,0+ q,0+ (iv) For any function f defined on A∗ q,a Dα Iα f(x)=f(x). (2.8) q,a− q,a− (v) If f C (A∗ )) then ∈A q q,a Iα cDα f(x)=f(x) f(0). (2.9) q,0+ q,0+ − (vi) If f is a function defined on A∗ then q,a aα−1 a Iqα,a−Dqα,a−f(x)=f(x)− Γ (α)(qx/a;q)α−1 Iq1,−a−αf (q). (2.10) q (cid:16) (cid:17) (v) If f is defined on [0,a] such that D f is continuous on [0,a] then q cDα f(x)=Dα [f(x) f(0)]. (2.11) q,0+ q,0+ − 5 Set X =A or A∗ . Then q,a q,a C(X) L2(X) L1(X). ⊆ q ⊆ q Moreover,if f C(X) then ∈ f √a f a f . k k1 ≤ k k2 ≤ k k We have also the following inequalities: 1. If f C(A∗ ) then Iα f C(A∗ ) and ∈ q,a q,0+ ∈ q,a aα Iα f f . (2.12) q,0+ ≤ Γ (α+1)k k q (cid:13) (cid:13) (cid:13) (cid:13) 2. If f L1(X) then Iα (cid:13)f L1((cid:13)X) and ∈ q q,0+ ∈ q aα(1 q)α Iα f M f , M := − . (2.13) q,0+ 1 ≤ α,1k k1 α,1 (1 qα)(q;q)∞ (cid:13) (cid:13) − (cid:13) (cid:13) 3. If f L(cid:13)2(X) t(cid:13)hen Iα f L2(X) and ∈ q q,0+ ∈ q Iα f M f , (2.14) q,0+ 2 ≤ α,2k k2 (cid:13) (cid:13) where (cid:13) (cid:13) (cid:13) (cid:13) aα (1 q) 1 1/2 M := − (qξ;q)2 d ξ . α,2 Γq(α)s(1−q2α)(cid:18)Z0 α−1 q (cid:19) 4. If α> 1 and f L2(X) then Iα f C(X) and 2 ∈ q q,0+ ∈ Iα f M f , M := aα−21 1(qξ;q)2 d ξ 1/2. (2.15) q,0+ ≤ αk k α Γ (α) α−1 q (cid:13) (cid:13) q (cid:18)Z0 (cid:19) 5. Since(cid:13)(cid:13) f (cid:13)(cid:13) √aff , we cfonclude that if f C(X) then Iα f L2(X) k k2 ≤ k k ∈ q,0+ ∈ q and Iα f K f , K :=√aM . (2.16) q,0+ 2 ≤ αk k α α,2 6. If f ∈C(A∗q,a) th(cid:13)(cid:13)(cid:13)en Iqα,a−(cid:13)(cid:13)(cid:13)f ∈C(A∗q,a) and aα(1 q)α Iqα,a−f ≤cα,0kfk, cα,0 := (1 qα)−(q;q)∞. (cid:13) (cid:13) − (cid:13) (cid:13) 7. If f ∈L1q(X)(cid:13)then Iqα(cid:13),a−f ∈L1q(X) and (1 q)αaα − f , ifα<1, (1 qα)(q;q)∞ k k1 Iα f − (cid:13) q,a− (cid:13)1 ≤ (1−q)α−1aα−1 f , ifα 1. (cid:13)(cid:13) (cid:13)(cid:13) (q;q)∞ k k1 ≥ 6 8. If α= 1 and f L2(X) then Iα f L1(X) and 6 2 ∈ q q,a− ∈ q (1−q)α−12aα f , ifα< 1, 1 q2α−1(q;q)∞ k k2 2 (cid:13)(cid:13)(cid:13)Iqα,a−f(cid:13)(cid:13)(cid:13)2 ≤ (q;q)∞p (−(11−qq2)αα−a1α)(1 q2α)kfk2, ifα> 12. The following lemmas are introdupced a−nd proved−in [28] Lemma 2.2. Let α>0. If (a) f L1(X) and g is a bounded function on A , ∈ q q,a or (b) α= 1 and f, g are L2(X) functions 6 2 q then a a g(x)Iα f(x)d x= f(x)Iα g(x)d x. (2.17) q,0+ q q,a− q Z0 Z0 Lemma 2.3. Let α (0,1). ∈ (a) If g L1(A∗ ) such that I1−αg C (A∗ ), and Dif C(A∗ ) (i = ∈ q q,a q ∈ A q q,a q ∈ q,a 0,1) then a x a a f(x)Dα g(x)d x= f( )I1−αg(x) + g(x)cDα f(x)d x. Z0 q,0+ q − q q,0+ (cid:12)x=0 Z0 q,a− (2.q18) (cid:12) (cid:12) (b) If f C (A∗ ), and g is a bounded function on A∗ such that Dα g ∈A q q,a q,a q,a− ∈ L1(A∗ ) then q q,a a x a a g(x)cDα f(x)d x= I1−αg ( )f(x) + f(x)Dα g(x)d x. Z0 q,0+ q (cid:16) q,a− (cid:17) q (cid:12)x=0 Z0 q,a− (2.1q9) (cid:12) (cid:12) 3. Basic Fourier series on q-Linear grid and some properties The purpose of this section is to reformulate Cardoso’s results of Fourier seriesexpansionsforfunctionsdefinedontheq-lineargrid := qn, n N to q 0 functions defined on q-linear grids :=:= aqn, n AN , {a > 0. ∈Card}oso q,a 0 A {± ∈ } in[29]defined the spaceof all q-linearHo¨lder functions onthe q-lineargrid . q A We generalize his definition for functions defined on a q-linear grid of the form , a>0. q,a A Definition 3.1. Afunctionf definedon ,a>0,iscalledaq-linearHo¨lder q,a A of order λ if there exists a constant M >0 such that f( aqn−1) f( aqn) Mqnλ,for alln N. ± − ± ≤ ∈ (cid:12) (cid:12) (cid:12) (cid:12) 7 Definition 3.2. The q-trigonometricfunctions S (z)andC (z) aredefined for q q z C by, see [29, 32] ∈ S (z) = ∞ ( 1)nqn(n+12)z2n+1 = z φ 0;q3;q,q3/2z2 , q 1 1 − (q;q) 1 q 2n+1 nX=0 − (cid:16) (cid:17) C (z) = ∞ ( 1)nqn(n−12)z2n = φ 0;q;q,q1/2z2 . q 1 1 − (q;q) 2n nX=0 (cid:16) (cid:17) One can verify that w D S (wz) = C (√qz), q,z q q 1 q − w D C (wz) = S (√qz), q,z q q −1 q − wherez Candw Cisafixedparameter. Amodificationoftheorthogonality ∈ ∈ relation given in [32, Theorem 4.1] is Theorem3.3. Letwandw′ berootsofS (z),andµ(w):=(1 q)C (q1/2w)S′(w). q − q q Then a C (q12wx)C (q12w′x)d x = 20a,, ifwifw=6=w′w=′,0, Z−a q a q a q aµ(w), ifw =w′ =0, 6 a qwx qw′x 0, ifw =w′, Sq( a )Sq( a )dqx = aq−1/2µ(w), ifw =6 w′. Z−a (cid:26) Cardosointroducedasufficientconditionforthe uniformconvergenceofthe basic Fourier series ∞ a S (f):= 0 + a C (q1/2w x)+b S (qw x), q k q k k q k 2 k=1 X 1 where a = f(t)d t and for k =1,2,..., 0 −1 q R1 1 1 1 a = f(t)C (q1/2w t)d t, b = f(t)S (qw t)d t, k q k q k q k q µ µ k Z−1 k Z−1 µ =(1 q)C (q1/2w )S′(w ) k − q k q k on the q-linear grid , where w : k N is the set of positive zeros of q k A { ∈ } S (z). Cardoso proved that µ =O(q−2k2) as k for any q (0,1). In the q k →∞ ∈ followingwegiveamodifiedversionofCardoso’sresultforanyfunctiondefined on the q-linear grid , a>0. q,a A Theorem 3.4. If f C( ∗ ) is a q-linear Ho¨lder function of order λ > 1, ∈ Aq,a 2 then the q-Fourier series ∞ a w x w x S (f):= 0 + a C (q1/2 k )+b S (q k ), (3.1) q k q k q 2 a a k=1 X 8 where a = 1 a f(t)d t and for k =1,2,..., 0 a −a q 1R a w t √q a w t a (f)= f(t)C (q1/2 k )d t, b (f)= f(t)S (q k )d t, k q q k q q aµ a aµ a k Z−a k Z−a converges uniformly to the function f on the q-linear grid . q,a A Proof. The proof is a modification of the proof of [29, Theorem 4.1] and is omitted. Remark 3.5. We replaced the condition f(0+)=f(0−) where f(0+):= lim f(x), f(0−):= lim f(x), x→0+ x→0− in [29, Theorem 4.1] by the weakest condition that f is q-regular at zero. Because he needs this condition only to guarantee that limn→∞f(qn−1/2) = limn→∞f( qn−1/2) and this holds if fis q-regular at zero. See [27, (1.22)] for − a function which is q-regular at zero but not continuous at zero. A modified version of [29, Theorem 3.5] is Theorem 3.6. If there exists c>1 such that a w t a w t f(t)C (√q k )=O(qck)and f(t)S (q k )=O(qck) ask , q q a a →∞ Z−a Z−a then the q-Fourier series (3.1) converges uniformly on . q,a A A modified version of [29, Corollary 4.3] is Corollary 3.7. If f is continuous and piecewise smooth on a neighborhood of theorigin, then thecorresponding q-Fourier series S (f)converges uniformly to q f on the set of points . q,a A Theorem 3.8. If f C( ∗ ) is a q-linear Ho¨lder odd function of order λ> 1 ∈ Aq,a 2 and satisfying f(0)=f(a)=0, then the q-Fourier series ∞ w x k S (f):= c S ( ), q k q a k=1 X where 2 a w t k c (f)=c = f(t)S ( )d t, k k q q a√qµk Z0 a converges uniformly to the function f on the q-linear grid . q,a A Proof. The proof follows from (3.4) by considering the function g(x) := f(qx), x . Since, it is odd, we have a =0 for k =0,1,..., and k ∈A a qw t k b (f)=√qµ g(t)S ( )d t, k k q q a Z−a making the substitution u= qt and using that g is an odd function, we obtain the required result. 9 Definition 3.9. Let (f ) be a sequence of functions in C( ∗ ). We say that n n Aq,a f converges to a function f in q-mean if n a lim f (x) f(x)2d x=0. n q n→∞sZ−a| − | Proposition 3.10. If g C( ∗ ) is an odd function satisfying Dkg (k = ∈ Aq,a q 0,1,2) is continuous and piecewise smooth function in a neighborhood of zero, and satisfying the boundary condition g(0)=g(a)=0 (3.2) then g can be approximated in the q mean by a linear combination − n w x g (x)= c(n)S ( k ) n r q a r= X where at the same time Dkg (k=1,2) converges in q-mean to the Dkg. More- q n q over, the coefficients c(n) need not depend on n and can be written simply as r c . r Proof. We consider the q-sine Fourier transform of D2g. Hence q ∞ qw x D2g(x)= b S ( k )= lim γ (x), x A , (3.3) q k q a n→∞ n ∈ q,a k=1 X where n √q a qw x γ (x)= b S (qw xa), b = D2g(x)S ( k )d x. n k q k k aµ q q a q k=1 k Z0 X Consequently, a lim D2g(x) γ (x) 2 d x=0. n→∞ q − n q Z0 (cid:12) (cid:12) Hence (cid:12) (cid:12) x a(1 q) ∞ b q1/2w x D g(x) D g(0)= D2g(x)d x= − k C ( k )+1 . q − q Z0 q q √q k=1wk (cid:18)− q a (cid:19) X Applying the q-integration by parts rule (1.2) gives a(1 q) a (D g)= − b (D2g). k q − √qwk k q ∞ q1/2w x k I.e. D g(x) D g(0)= a (D g) C ( ) 1 . q q k q q − a − k=1 (cid:18) (cid:19) X 10