On Spectrum Generating Algebra of the Heun Operator PriyasriKara,RiteshK.Singhb,AnandaDasguptac,PrasantaK.Panigrahid DepartmentofPhysicalSciences, IndianInstituteofScienceEducationandResearchKolkata,Mohanpur741246,India a [email protected],b [email protected], c [email protected],d [email protected] April6,2016 6 1 0 2 Abstract r p A TheHeunoperatorhasbeencast,intermsoftheelementsofanunderlyingsu(1,1)algebra, 5 underspecificparametricconditions,forthepurposeofspectrumgeneration. Theseelements aredifferentialoperatorsofdegrees±1/2and0.Itisfoundthattheregularsingularitiesat0and ] ∞ofthegeneralHeunequationmustbeelementaryundertherequiredparametricconditions. h Thespectrumgenerationhasbeendemonstratedthroughasetofexamples. p - t n Differential equations, in general, do not admit exact solutions. Several techniques have been de- a u vised, to obtain partial, or in certain cases, the complete eigenspace of the differential operator q under study [1]. In this regard, the well known factorization method, which at a fundamental [ levelreducestheorderofthedifferentialoperatorinvolved,hasfoundextensiveapplications[2]. 2 The harmonic oscillator spectrum generating Heisenberg algebra, the Darboux transformation, v 1 factorization of the hypergeometric equation by Schro¨dinger [3] and subsequent study of a large 0 classofequationsbyInfeldandHull[2]serveasclassicexamples. Supersymmetricquantumme- 6 chanics(SUSYQM)usesfactorization,torevealunexpectedinterconnectionsamongstapparently 3 0 differentquantalproblems[4,5,6]. . 1 0 Grouptheoreticalstructuresalsohaveplayedsignificantroleinunravellingthesymmetryofdif- 6 ferentialoperators[1]. Apartialalgebraizationofthewholesolutionspacehasbeenachieved[7, 1 8,9]forawideclassofquantalquasi-exactlysolvable(QES)Hamiltoniansofoneormoredimen- : v sions, which are essentially hermitian differential operators of the second order. These Hamil- i X tonians, or their appropriately similarity transformed versions, often take the form of some well r known differential operators, e.g., Heun, Lame´ etc. Hence, the study of the underlying symme- a tries of these operators is of immense importance and several attempts have been made to this end. For example, the Lame´ equation has been expressed as a bilinear in both su(2) [10] and su(1,1) [11] generators. The latter case relates the non-unitary representations of su(1,1), to the eigenfunction of the periodic Lame´ potential. Closed form solutions for confluent hypergeomet- ricandhypergeometricequationshavebeenfound[12],usingaconnectionbetweenthespaceof monomials and the solution space of the above equations [13, 14], which reveals the underlying su(1,1)anddeformedsl(2)structures,respectively[15]. Recently, the Heun operator has been cast [16] in terms of the elements of cubic deformations of sl(2) algebra. This has led to two known solutions of a Heun equation, encountered earlier byChristandLee[17]. However,non-lineardeformationsofsl(2),areassociatedwithnon-trivial 1 representationtheory[18]. Castingthedifferentialoperatorintermsoflinearsl(2)algebraenables one to exploit available representation theory. Linear su(1,1) symmetry of the Heun operator is present in the literature [7]. In the present work, we use a different representation of su(1,1) generators. Under certain parametric conditions, this reveals a new symmtery of the equation and leads to a number of solutions, that are unavailable from the methods in [7]. It is found that this new symmetry exists if the singularities at z = 0 and z = ∞ are elementary1. Under these conditions, an exact map is established between the Heun solution space and the unitary and non-unitaryrepresentationsofsu(1,1)algebra. TheHeunequationisgivenby[20] d2y(z) (cid:18)γ δ ε (cid:19) dy(z) αβz−q Hy(z) = + + + + y(z) = 0, (1) dz2 z z−1 z−a dz z(z−1)(z−a) with regular singularities at z = 0, 1, a((cid:54)= 0,1)and ∞; the exponents at these singularities being (0, 1−γ),(0, 1−δ),(0, 1−ε)and(α, β),respectively. Here,qplaystheroleoftheeigenparameter. In this work, we study Heun equations with only real parameters. The equation, being a second orderlinearhomogeneousequationwith4regularsingularities,satisfiestheFuchsiancondition: γ +δ+ε = α+β +1, (2) allowingeliminationofεinfavoroftheothers. Eq.(1)canbewrittenas: (cid:20) d2 d (cid:21) Hy(z) = f (z) +f (z) +f (z) y(z) = 0, (3) 1 dz2 2 dz 3 where,f (z) = a z3+a z2+a z, f (z) = a z2+a z+a and f (z) = a z+a . Theparameters 1 0 1 2 2 3 4 5 3 6 7 a ∈ R, fori = 0,...,7aregivenby, i a =1 a = 1+α+β a = αβ (4a) 0 3 6 a =−(a+1) a = −[aγ +aδ−δ+α+β +1] a = −q (4b) 1 4 7 a =a a = aγ. (4c) 2 5 Evidently,theHeunequationconsistsofdifferentialoperatorsofdegrees2 +1, 0and−1,whichare denotedbyP ,F(P )andP ,respectively. Eq.(3)canthenberewrittenas[16], + 0 − Hy(z) = [P +F(P )+P ]y(z) = 0 (5) + 0 − where, d2 d d d2 d P = a z3 +a z2 +a z, P = z −j, P = a z +a (6) + 0 dz2 3 dz 6 0 dz − 2 dz2 5dz and F(P ) = a P2+((2j −1)a +a )P +(j(j −1)a +ja +a ). (7) 0 1 0 1 4 0 1 4 7 Theaboveoperatorssatisfycubicdeformationofsl(2)algebraamongthemselves,whichhasbeen exploited[16]totractapartoftheeigenspaceofaHeunoperatoranalytically,asmentionedearlier. WiththeaimtocasttheHeunoperatorintermsoflinearsu(1,1)algebra,wedefinetheconstituent operatorsas, d √ d d 2ν E = 2z3/2 +2µ z, H = 2z +µ+ν and E = 2z1/2 + √ ; (8) + − dz dz dz z 1Anyregularsingularitycanbecharacterizedbytwoexponentsρ1 andρ2,whicharethetworootsoftheindicial equation. The cases with |ρ1 −ρ2| = 1/2 are known as elementary singularity and are of special significance. All regularandirregularsingularitiescanbeobtainedbythecoalescenceoftwoandthreeormoreelementarysingularities, respectively[19]. 2Thedegreed,ofanoperatorOd,isdefinedasthechangeinthepowerofamonomial,whenacteduponbyit,i.e., Odzp ∝zp+d. where µ, ν and λ are parameters. The operators E , H and E , have degrees +1/2, 0 and −1/2, + − respectively. Theyarefoundtosatisfysu(1,1)algebra,withcommutators, [H,E ] = ±E , [E ,E ] = −2H (9) ± ± + − andtheCasimir: 1 C(µ,ν) = (E E +E E )−H2 = −(µ−ν)(µ−ν −1). (10) + − − + 2 The problem in question, determines the parameters of the constituent operators, which deter- mine the Casimir value. The Casimir value, in turn, determines the representations available for thealgebraassociatedwiththeproblem. We now proceed to construct the differential operators P ,P and F(P ), from the constituent + − 0 su(1,1) generators and identify the conditions, under which this is possible. The constituent op- erators are of degrees ±1/2 and 0, whereas, the degrees of operators P and F(P ) are ±1 and ± 0 0, respectively. Hence, it is clear that the complete Heun operator will comprise of linear and quadraticformsoftheconstituentoperators3. Assumingtheforms, P = c E E and P = c E E (11) + + + + − − − − wherec areconstants,weobtain, ± a a 0 0 a = 4c , a = 2c (3+4µ) = (3+4µ), a = 2c µ(1+2µ) = µ(1+2µ), (12) 0 + 3 + 6 + 2 2 a 2 a = 4c , a = 2c (1+4ν) = (1+4ν), 2c ν(2ν −1) = 0. (13) 2 − 5 − − 2 UsingEqs.(12)and(4a),onecansolveforµ,α,β: 1 |α−β| = , (14) 2 implyingthesingularityatz = ∞iselementary. Similarly,usingEqs.(13)and(4c),oneisleftwith justtwovaluesofν: ν = 0 =⇒ a /a = γ = 1/2 or, ν = 1/2 =⇒ a /a = γ = 3/2. (15) 5 2 5 2 Above solutions of γ imply that the singularity at z = 0 is also elementary. Finally, F(P ) can be 0 writtenintheform F(P ) = c HH +c H +c , (16) 0 2 1 0 where, a = 4c , a = 2(c +2c (1+µ+ν)) and a = −q = c +(µ+ν)(c +c (µ+ν)). (17) 1 2 4 1 2 7 0 1 2 Thus, a Heun operator with elementary singularities at z = 0 and z = ∞, i.e., with parametric conditionsEqs.(14)and(15),becomes, H = c E E +c E E +c HH +c H +c . (18) + + + − − − 2 1 0 The choices of E and E fix the values of µ and ν, respectively, which in turn determines the + − Casimir, C = C(µ,ν). Finding the solution to Heun equation, then boils down to identifying a suitable representation space V of su(1,1) and finding a y ∈ V , such that, Hy = 0. For su(1,1) C C algebra, there are five different classes of representation spaces [21] determined by the values of C andh,theeigenvalueofoperatorH. TheseareshowninFig.1andarelistedbelow. 3Cubicandquarticcombinations,suchasE+E+H,E+E+HH,E+E+E+E− (andsimilartermswithE+ andE− interchanged),arealsoofthedesireddegreesandthereforecouldbeusedintheconstructionoftheHeunoperators. However,theycontaindifferentialoperatorsofhigherorders,notpresentintheHeunequation.Thus,thecoefficientsof thoseundesiredtermsmustbeputtozero,whichrevealednomoreparametricfreedomthanthechosencombinations. (Quadruplet) (Triplet) (Doublet) (Singlet) (PD/ND/PS) −15/4 −2 −3/4 0 1/4 (PD/ND) (PS) Casimir (Real line) (PD/ND/CS) Figure 1: Casimir value for su(1,1) and corresponding representations are listed: positive discrete (PD), negativediscrete(ND),principalseries(PS),complementaryseries(CS)andsomefinitedimensionalones (opencircles). Principalseries(PS):Infinitedimensionalspace,C ≥ 1/4,h ∈ (−∞,∞)and(C,h) (cid:54)= (1/4,1/2), Complementaryseries(CS):Infinitedimensionalspace,C ∈ (0,1/4),h ∈ (−∞,∞),h (cid:54)= 1/2, Positivediscrete(PD):Infinitedimensionalspace,C ≤ 1/4,h ≥ 0, Negativediscrete(ND):Infinitedimensionalspace,C ≤ 1/4,h ≤ 0, (cid:104) (cid:105) Finitedimensions: Finitedimensionalspace,C = −(n2−1),h ∈ −(n−1), (n−1) withn ∈ N. 4 2 2 For a given value of Casimir C, one of the available representation space V can be chosen. The C elements of this space are |C,h(cid:105) ∝ zp, with p = (h−µ−ν)/2. We represent the solution y as a linearcombinationofzp fromV andfindtheco-efficients. DuetothequadraticdependenceofH C onE andE ,wenoticethat,thesolutionspacesplitsasV = V ⊕V —theevenandoddstates + − C e o oftherepresentationspace. TheprocessoffindingsolutionsinV andV isdemonstratedbelow, e o throughasetofexamples. Example 1: We choose γ = 1/2, δ = ε = −1/2, αβ = 1/2. This leads to |α−β| = 1/2, ensuring thesingularitiesat0and∞tobeelementary. TheHeunequationtakestheform d2y 1 (cid:18)1 1 1 (cid:19) dy z−2q + − − + y = 0, (19) dz2 2 z z−1 z−a dz 2z(z−1)(z−a) withµ = −1,ν = 0i.e.,C = −2. Thisallowsthepositivediscreteandnegativediscreterepresen- tationspacesalongwithatripletrepresentationwithh ∈ {−1,0,+1}orp ∈ {0,1/2,1}. Choosing thetripletspace,V correspondstop ∈ {0,1}andV top ∈ {1/2},leadingtothesolutions e o √ √ y ∈ V = z+ a, witheigenvalue q = + a/2, (20) 1 e √ √ y ∈ V = z− a, witheigenvalue q = − a/2 (21) 2 e √ and y ∈ V = z, witheigenvalue q = (a+1)/4. (22) 3 o The positive discrete and negative discrete representations also yield solutions, not explicitly shownforthisexample. Example 2: We choose γ = 3/2, δ = ε = −1/2, αβ = 0, leading to |α − β| = 1/2. The Heun equationisgivenby, d2y 1 (cid:18)3 1 1 (cid:19) dy q + − − + y = 0 (23) dz2 2 z z−1 z−a dz z(z−1)(z−a) andcorrespondstoµ = −1/2,ν = 1/2withC = −2, asinthepreviouscase. Here, forthetriplet space,oneobtainsp ∈ {−1/2,0,1/2},withp ∈ {0}beingV . Thesolutionsare: o √ (cid:112) √ y ∈ V = z+ a/z, witheigenvalue q = −(a+1)/4+ a/2, (24) 1 e √ (cid:112) √ y ∈ V = z− a/z, witheigenvalue q = −(a+1)/4− a/2 (25) 2 e and y ∈ V = const., witheigenvalue q = 0. (26) 3 o Heretoo,wehavepositivediscreteandnegativediscreterepresentations,likethepreviousexam- ple. Example3: Next,weanalyzetheLame´ equation,whichisaspecialcaseoftheHeunequation: d2y 1 (cid:18)1 1 1 (cid:19) dy q+ρ(ρ+1)z/4 + + + − y(z) = 0. (27) dz2 2 z z−1 z−a dz z(z−1)(z−a) Here, the singularities at z = 0, 1 and a are elementary and choosing ρ(ρ + 1) = 0, makes the singularity at z = ∞ also elementary. With this choice, we have µ = 0 = ν, i.e., C = 0, giving us thepossibilitytochoosefromsinglet,positivediscreteandnegativediscreterepresentations. The singlet solution is, y = const., with q = 0. The positive discrete space splits into the even (V+) 0 e andtheodd(V+)subspacesandthecorrespondingsolutions,forachoseneigenvalueq ∈ R,are: o (cid:34)z q+a+1 (cid:16)z(cid:17)2 2q2+10q(a+1)+8(cid:0)a2+1(cid:1)+7a (cid:16)z(cid:17)3 (cid:35) y ∈ V+ = 1+2q + + +... (28) 1 e a 3 a 45 a √ (cid:34) 4q+a+1 (cid:16)z(cid:17) 16q2+40q(a+1)+9(cid:0)a2+1(cid:1)+6a (cid:16)z(cid:17)2 (cid:35) y ∈ V+ = z 1+ + +... . (29) 2 o 6 a 120 a Similarly,forthenegativediscretespacetoo,wehavetheeven(V−)andtheodd(V−)subspaces e o andthecorrespondingsolutions,forachoseneigenvalueq ∈ R,are: (cid:34)1 q+a+1 (cid:18) 1 (cid:19) 2q2+10q(a+1)+8(cid:0)a2+1(cid:1)+7a (cid:18) 1 (cid:19) (cid:35) y ∈ V− = 1+2q + + +...(30) 3 e z 3 z2 45 z3 1 (cid:34) 4q+a+1 (cid:18)1(cid:19) 16q2+40q(a+1)+9(cid:0)a2+1(cid:1)+6a (cid:18) 1 (cid:19) (cid:35) y ∈ V− = √ 1+ + +... .(31) 4 o z 6 z 120 z2 Thedomainofconvergenceforthepositivediscretespacesolutionsis0 < z < min{1,a},whereas, that for the negative discrete space solutions is max{1,a} < z < ∞. The solutions in infinite dimensional spaces can always be written as a power series in z or 1/z, with an overall factor of zp,p ∈ R,liketheabovefoursolutions. For all the above examples, we have obtained solutions, which are not available from the previ- ously known su(1,1) symmetry of the Heun equation [7]. These solutions are presented here by √ Eqs. (22), (24), (25), (29) and (31). It may be noted that all of them involve powers of z, which reflects the advantage of using a different representation of the su(1,1) generators in the present work. In conclusion, we have found that, the Heun operator can be exactly cast as a quadratic polyno- mial of elements of su(1,1) algebra, if the singularities at z = 0 and z = ∞ are elementary. This allows one to use the representations of su(1,1) to find explicit solutions of Heun equation. The √ finite dimensional representations yield polynomial solutions in z, while positive and negative discreterepresentationsgivepowerseriessolutionsinz and1/z,respectively,moduloanoverall factor of zp. Heun equation is connected with a host of physical problems, related to quantum mechanical and non-linear systems. Hence, the relevance of the algebraically generated solution spacetotheseproblemsneedscarefulconsiderations. Weintendtoreturntosomeoftheseprob- lemsinnearfuture. Acknowledgements: WethankAruneshRoyforusefuldiscussions. References [1] WillardMillerJr.,SymmetryGroupsandtheirApplications,AcademicPress(1972). [2] L.InfeldandT.E.Hull,Rev.Mod.Phys.23,1(1951). [3] E.Schro¨dinger,Proc.Roy.IrishAcad.A47,53(1941). [4] F.Cooper,A.KhareandU.Sukhatme,Phys.Rep.251,267(1995). [5] F. Cooper, A. Khare and U. Sukhatme, Supersymmetry in Quantum Mechanics, World Scientific (2001). [6] A. Gangopadhyaya, J. V. Mallow, C. Rasinariu, Supersymmetic Quantum Mechanics: An Intro- duction,WorldScientific(2010). [7] A.V.Turbiner,Commun.Math.Phys.118,467(1988). [8] M.A.ShifmanandA.V.Turbiner,Commun.Math.Phys.126,347(1989). [9] A. G. Ushveridze, Quasi-exactly Solvable Models in Quantum Mechanics, Inst. of Physics Publ. (1994). [10] Y.Alhassid,F.GurseyandF.Iachello,Phys.Rev.Lett.50,873(1983). [11] H.LiandD.Kusnezov,Phys.Rev.Lett.83,1283(1999). [12] N.Gurappa,P.K.JhaandP.K.Panigrahi,SIGMA3,057(2007). [13] N.GurappaandP.K.Panigrahi,Phys.Rev.B62,1943(2000). [14] N.GurappaandP.K.Panigrahi,J.Phys.A37,L605(2004). [15] T.Shreecharan,P.K.PanigrahiandJ.Banerji,Phys.Rev.A69,012102(2004). [16] A.Roy,A.SenandP.K.Panigrahi,Preprint[arXiv:1304.2225]. [17] N.H.ChristandT.D.Lee,Phys.Rev.D12,1606(1975). [18] M.Rocek,Phys.Lett.B255,554(1991). [19] E.L.Ince,OrdinaryDifferentialEquations,DoverBooksonMathematics(1926). [20] A.Ronveaux(ed.),Heun’sDifferentialEquations,OxfordUniversityPress(1995). [21] Seeforexample,W.Groenevelt,Indag.Math.(N.S.)14,329(2003)andreferencestherein.