Fuzzy Complex Quadrics and Spheres

DIAS-STP-03-09 Fuzzy Complex Quadrics and Spheres Brian P. Dolan,a),b)∗Denjoe O’Connorb)†and P. Preˇsnajderc)‡ a)Dept. of Mathematical Physics, NUI, Maynooth, Ireland b)School of Theoretical Physics, Dublin Institute for Advanced Studies, 10 Burlington Rd., Dublin 8, Ireland c)Dept. of Theoretical Physics, Comenius University, Mlynsk´a dolina, SK-84248 Bratislava, Slovakia January 17, 2004 Abstract A matrix algebra is constructed which consists of the necessary degrees of freedom for a finite approximation to the algebra of functions on the family of orthogonal Grassmannians of real dimension 2N, known as complex quadrics. These matrix algebras contain the relevant degrees of freedom for describing truncations of harmonic expansions of functions on N-spheres. An In¨on¨u-Wigner contraction of the quadric gives the co-tangent bundle to the commutative sphere in the continuum limit. It is shown how the degrees of freedom for the sphere can be projected out of a finite dimensional functional integral, using second-order Casimirs, giving a well-defined procedure for construction functional integrals over fuzzy spheres of any dimension. ∗[email protected] †[email protected] ‡[email protected] 1 1 Introduction Non-commutative geometry has slowly been increasing in importance in physics over the last 20 years and has recently received a strong impetus through work in string theory. An important concept in the non-commutative programme is that of a “fuzzy” space — this is perhaps more correctly described as a finite, non-commuting, matrix approximation to the algebra of functions on a continuous manifold which can reproduce the commutative algebra in the limit of the matrices becoming infinite in size. It has been suggested that fuzzy spaces could provide a regularisation technique for numerical calculations in quantum field theory which would be an alternative to lattice gauge theory, [1]-[4]. The prototypical example of a fuzzy space is the fuzzy two-sphere, [5], but there are many more examples, and indeed much of the work on generalised coherent states in quantum mechanics [6], when restricted to compact groups, can be related to fuzzy spaces. The geometry of fuzzy CPNand fuzzy unitary Grassmannians has been examined in the literature in some detail [7]-[10]. Clearly, from the point of view of a regularisation technique in field theory as well as for the theory of D-branes in string theory, it would be desirable to have an explicit construction of the fuzzy sphere SN Fin dimensions other than N= 2. Unfortunately, despite the elegant simplicity of the fuzzy two-sphere S2 F, higher dimensional spheres are not so amenable to a fuzzy description. To our knowledge there is no closed finite matrix approximation to the algebra of functions on a sphere for dimensions greater than two, though the problem was tackled in [11] and [12] (non-commutative spheres in the continuum were analysed in [13]). It appears that, while one can represent truncated harmonic expansions of functions on spheres by square matrices, the product of two such matrices takes one out of the required space and a projection back into the space of functions is necessary after every multiplication — this renders the product non-associative. In other words a star product cannot be defined on the fuzzy sphere SN Ffor N > 2 (this is related to the fact that, for N6= 2, SNcannot be obtained as the co-adjoint orbit of a compact group). Nevertheless the construction in [11] does associate a square matrix with the truncation of a harmonic expansion on SNand so does, in a sense, constitute a fuzzy sphere, even though there is no associative product. For S4 Falternative descriptions are possible, [14] [15]. The construction in [15] is in terms on CP3 Fand uses the fact that the continuum CP3is an S2bundle over S4[15]. This approach has the advantage that it is designed to be used in a functional integral and the technique was extended to S3 F and S1 Fin [16]. The method of [11] would be particularly cumbersome to implement in functional integrals over S3 Fand S1 F, or indeed any odd sphere. 2 In this paper we generalise the construction in [16] to fuzzy spheres of any dimension. The analysis relies on properties of the orthogonal Grassmannian, SO(N+ 2)/[SO(N)×SO(2)], which is a co-adjoint orbit of dimension 2N. This orthogonal Grassmannian can also be obtained as the complex quadratic zaza= 0 in CPN+1, where zaare natural complex co-ordinates, [17]. The notation in Kobayashi and Nomizu is QN∼ =SO(N+ 2)/[SO(N)×SO(2)] and we shall use this as a shorthand. In¨on¨u-Wigner contraction of SO(N+ 2) to the Euclidean group of RN+1 relates QNto the co-tangent bundle T∗SN. In a sense, elaborated on in section 2, QNcan be thought of as the compactified co-tangent bundle for a non-commutative sphere. There is a finite dimensional, non-commutative, matrix algebra approximation to the algebra of functions QNwhich contains harmonic expansions on SN(described in section 3). While it is still the case that the resulting matrix algebra takes one out of the space of harmonic expansions on SNupon multiplication (so one does not have a closed matrix approximation to the algebra of functions on SNfor N6= 2) it is nevertheless very easy to suppress unwanted modes in a functional integral over a fuzzy complex quadric, QN, in a manner which lends itself naturally to numerical computation for field theory on SN F, as described in section 4. 2 The deformed co-tangent bundle In this section it is shown how the co-tangent bundle T∗SNcan be deformed to a version related to a non-commutative sphere and compactified to complex quadric, QN. The construction starts with Cartesian co-ordinates Xain RN+1, where a= 1, . . . , N + 1. Consider the sphere SN∼ =SO(N+ 1)/SO(N) of radius R defined by XaXa=R2. The isometry group is SO(N+ 1) with algebra [Lab, Lcd] = i(δbcLad +δadLbc −δacLbd −δbdLac),(1) where Lab =−Lba (there is no distinction between upper and lower Euclidean indices in RN+1,Xa=Xa). This can be extended to a representation of the Euclidean group, EN+1 acting on RN+1, [Lab, Lcd] = i(δbcLad +δadLbc −δacLbd −δbdLac) (2) [Lab, Xc] = i(δbcXa−δacXb) (3) [Xa, Xb]=0.(4) An explicit realisation of this algebra in the continuum is Lab =i Xa ∂ ∂Xb −Xb ∂ ∂Xa!(5) 3 acting on functions and Xabeing commutative multiplication by the coordinates. The algebra will now be deformed, essentially using the inverse of In¨on¨uWigner contraction. Let Xa:= µLa,N+2, with µreal, and replace the commutator (4) above with [Xa, Xb] = −iµ2Lab,(6) leaving the other commutators unchanged. The Euclidean group EN+1 is thus deformed to SO(N+ 2) [LAB, LCD] = i(δBCLAD +δADLBC −δACLBD −δBDLAC),(7) where A, B, C, D = 1, . . . , N + 2 and La,N+2 =−LN+2,a. In terms of the quadratic Casimirs1CN+1 2= (1/2)LabLab and CN+2 2= (1/2)LABLAB we see that XaXa=µ2CN+2 2−CN+1 2.(8) While it may be tempting to think of Xain (8) as co-ordinates on a fuzzy sphere we must be careful: a general irreducible representation of SO(N+2) will decompose into a sum of different irreducible representations of SO(N+ 1), with different values of CN+1 2, so the right-hand side of (8) will not be central. However XaXais central in the fundamental spinor representation of Spin(N+ 2). To see this consider the even and odd cases separately: •Even N: choose one chirality of spinor with 2N/2components. Under Spin(N+ 2) →Spin(N+ 1) this reduces uniquely to the single spinor representation of Spin(N+ 1), which has the same dimension. In this case the right-hand side of (8) is a multiple of the identity. •Odd N: in this case the 2N/2dimensional spinor representation of Spin(N+ 2) decomposes into two spinor representations of opposite chirality under SO(N+ 2) →SO(N+ 1), both of dimension 2(N/2)−1. But the second order Casimir CN+1 2has the same value on the two chiralities, it does not distinguish between them, [18]. So again the right-hand side of (8) is a multiple of the identity. The fundamental spinor representations of Spin(k) have quadratic Casimir Ck 2=k(k−1)/4 for both even and odd k, see [18] for example. For a fundamental spinor representation of Spin(N+ 2) equation (8) therefore gives XaXa=µ2(N+ 1) 21,(9) 1Relative to the standard conventions for SU(n) our normalisation here is such that CSU(2) 2=1 2C3 2and CSU(4) 2=1 2C6 2. 4 where 1is the identity operator, and we might interpret qN+1 2µas the radius of a fuzzy sphere. However, unlike S2 F, the matrix algebra generated by Xa will not close in general even when XaXais central. Also one cannot get higher dimensional representations of SN Fby taking tensor products. This reflects the fact that there is no closed finite matrix approximation to SN Ffor N6= 2. Let us now investigate the geometry of the space generated by the adjoint action of Spin(N+ 2) on a fiducial direction XN+1 (the “north pole” of SN F). Clearly [Lαβ, XN+1] = 0 for α, β = 1, . . . , N, so Spin(N) leaves XN+1 invariant. Also LN+1,N+2 commutes with XN+1, since they are just multiples of one another, so the SO(2) generated by LN+1,N+2 also leaves Xainvariant. The upshot of this is that the manifold that D(g)∈Spin(N+ 2) generates with the action D−1(g)XN+1D(g) (10) is the orthogonal Grassmannian QN, which has dimension 2N. For N≤4 these spaces have the following structures: •Q1∼ =SO(3)/SO(2) ∼ =S2; •Q2∼ =SO(4)/[SO(2) ×SO(2)] ∼ =S2×S2(S2×S2was used in a consideration of fuzzy S3/Z2in [19]); •Q3∼ =SO(5)/[SO(3) ×SO(2)] ∼ =CP3/Z2(this identification is described in [16]); •Q4∼ =SO(6)/[SO(4) ×SO(2)] ∼ =SU(4)/[S(U(2) ×U(2))] (matrix approximations to this space were described in [10]). We propose to identify QNwith a ‘compactified’ co-tangent bundle for a non-commutative sphere. This reduces to the usual T∗SNunder In¨on¨uWigner contraction µ→0: this limit performs the dual function of rendering the X’s in (6) commutative while at the same time de-compactifying SO(N+ 2) to the Euclidean group EN+1. 3 Fuzzy Complex Quadrics, QN F Any unitary irreducible representation Tof a simple compact Lie group is finite dimensional and the extension to its enveloping algebra is a finite dimensional matrix algebra. This matrix algebra provides a fuzzy approximation to the algebra of functions on the co-adjoint orbit of any Lie algebra element in T. 5 We shall refer to a finite matrix approximation to QNas a fuzzy complex quadric and denote it by QN F. Being a co-adjoint orbit QNis a symplectic manifold whose algebra of functions can be approximated by closed finite dimensional matrix algebras. The harmonic expansion of a function on QN requires all representations of SO(N+ 2) which contain the trivial representation under SO(N)×SO(2). A spinor representation of SO(N+ 2) never contains a singlet of SO(N)×SO(2) for either even or odd N, so we restrict to vectorial representations. The representation space of SO(N+ 2) is then the space of rank-ntensors, TA1···An. Irreducible representations can be characterised by their symmetries under interchange of their indices and are traceless in any two indices. They can be represented by Young tableau with nboxes which reflect their permutation symmetries. Any tensor that is anti-symmetric in three or more of its indices must vanish when restricted to a representation of SO(2) and hence cannot contribute to the harmonic expansion of functions on QN. Thus we can restrict our attention to tensors that are anti-symmetric in pairs of indices only. If TA1···Anhas mpairs of anti-symmetric indices and n−2msymmetric indices then the corresponding Young tableau is m z }| { ·· ·· n−2m z }| { ·· .(11) When SO(N+ 2) is restricted to SO(N)×SO(2) the anti-symmetric pairs all contain singlets of SO(N)×SO(2) (when both indices in a pair are SO(2) indices, for example). Also if nis even the n−2msymmetric indices can contain singlets of SO(N)×SO(2), but not when nis odd. Thus the harmonic expansion of a function on QNrequires all SO(N+ 2) tensor representations of the form (11) with neven. The dimension of these representation can be determined, using the relevant formulae in [18] for example. With n= 2l they are, for N≥3, dN(2l, m) = (2l+N−1)(2l+ 1 −2m)(4l+N−2m)(N−2+2m) ×(2l+N−2−m)!(N−3 + m)! N!(N−2)!(2l−m+ 1)!m!(12) where m≤l.2If the harmonic expansion of a function on QNis truncated at lmax =Lthe total number of degrees of freedom is L X l=0 l X m=0 dN(2l, m) = "(2L+N)(L+N−1)! L!N!#2 = [dN(L, 0)]2.(13) 2In the notation of [18] the representations (11) have highest weights (r1, . . . , r[N/2]+1) = (n−m, m, 0, . . . , 0), where [N/2] is the integer part of N/2. 6 The fact that this is a perfect square reflects the fact that the degrees of freedom in a truncated harmonic expansion can be arranged into a square matrix of size dN(L, 0)×dN(L, 0). This can be represented in terms of Young tableau by L z }| { ·· × L z }| { ·· = 1 ⊕ ⊕ ⊕ · · · ⊕ L z }| { ·· ·· ⊕ · · · ⊕ 2L z }| { ·· (14) which relates the matrix structure, on the left-hand side, to the harmonic expansion, on the right-hand side. Matrix multiplication then gives a closed associative, but non-commutative, algebra which reproduces the commutative algebra of functions on QNas L→ ∞. At the level of functions the non-commutative product at finite Lcan be realised as a ∗-product for the fuzzy complex quadric, QN F. 4 Functional Integrals on Fuzzy Spheres, SN F There is no closed finite matrix approximation for the truncated algebra of functions on SNknown, except for the special case N= 2 which was first described in [5]. There does exist a matrix approximation to functions on SN, but in general matrix multiplication does not correspond to the algebra of functions and the latter can only be recovered by projecting back onto a function on SNafter matrix multiplication [11]. This results in a nonassociative algebra when N6= 2 which, by a slight abuse of language, is nevertheless still referred to as a “fuzzy” sphere, SN F. In the construction presented here a similar projection can be performed, since the truncated harmonic expansion of a function on SNis buried in QN F. To see this note that functions on SNcan be expanded in symmetric tensor representations of SO(N+ 1). A truncation at level lmax requires using all symmetric tensors of SO(N+ 1), Ta1···al, with 0 ≤l≤lmax. These are all contained in one symmetric representation lmax z }| { ·· (15) of SO(N+ 2) under SO(N+ 2) →SO(N+ 1). Setting lmax = 2Lwe see that the last representation on the right-hand side of (14) contains all the SO(N+1) representations necessary for the harmonic expansion of a function on SNup to angular momentum 2L. We can thus obtain the fuzzy sphere SN Fby projecting the irreducible representation 2L z }| { ·· out from the matrix algebra of QN Fin (14). 7 This can be achieved in a functional integral over QN Fin the same manner as in [16]. Let Φ be a matrix in the algebra of QN Ffor a given L. The SO(N+ 2) invariant Laplacian on QN Fis L2 (N+2)Φ = −(1/2)[LAB,[LAB,Φ]].(16) Then the action for a scalar field on QN Fcan be written as S[Φ] = 1 dN(L, 0)Tr nΦ†L2 (N+2)Φ + V(Φ)o.(17) with the scalar potential V(Φ†) = V(Φ) assumed bounded below. A functional integral then involves Z=ZDΦe−S[Φ].(18) We focus on the SN Fembedded in QN Fby penalising all the SO(N+ 2) representations in Zexcept the last one on the right-hand of (14). This can be achieved by modifying the kinetic term in the action. The second order Casimir of the representation (11), with n= 2l, is [18] CN+2 2(2l, m) = (2l−m)(2l−m+N) + m(m+N−2).(19) Observe that the completely symmetric tensors with m= 0 have the largest Casimir for any given l, CN+2 2(2l, 0) = 2l(2l+N).(20) Hence the operator −1 2[LAB,[LAB,·]−2L(2L+N) = CN+2 2−2L(2L+N) (21) acting on Φ is negative for all modes in Φ except for the top mode, with l=Land m= 0, on which it vanishes. The SO(N+ 1) invariant Laplacian on SN Fwould be L2 (N+1)Φ = −(1/2)[Lab,[Lab,Φ]].(22) So the action Sh[Φ] = 1 dN(L, 0)Tr nΦ†L2 (N+1)Φ + hΦ†−L2 (N+2) + 2L(2L+N)Φ + V(Φ)o, (23) with h1, will suppress all the unwanted modes in a functional integral and leave the required modes for SN Funaffected. In the limit h→ ∞ all modes, except the ones relevant to SN F, will be suppressed and correlation functions calculated with Z=ZDΦe−S∞[Φ].(24) will be those of the fuzzy sphere, truncated at level 2L. 8 5 Conclusions By constructing fuzzy approximations to complex quadrics, QN F, a prescription for defining functional integrals over finite approximations to spheres has been presented. Although finite matrix approximations to the algebra of functions on N-dimensional spheres are not known for N6= 2, one can construct matrix approximations to the functions, but matrix multiplication then takes one out of the space of functions on the sphere. The construction presented here relies in the fact that the complex quadrics (which are orthogonal Grassmannians QN∼ =SO(N+ 2)/[SO(N)×SO(2)]) are co-adjoint orbits, and hence do have finite matrix approximations to their algebra of functions, QN F. These spaces are related to the co-tangent bundles T∗SN— they are in a sense compactified versions of the co-tangent bundles, compactified at the expense of introducing a non-commutativity on the sphere. The algebra of functions on QN Fcontains the relevant degrees of freedom for a truncated harmonic expansion of a function on SN. The functional integral for a field theory defined on QNcan be regularised in a manner that preserves the isometries and avoids Fermion doubling [20] by defining it over the fuzzy space QN F. By modifying the kinetic term and using the action (23) the degrees of freedom that are not relevant to the underlying sphere can be prevented from contributing to the functional integral and the result is a well defined, finite approximation for the functional integral of a quantum field theory on SN. The correct algebra is ensured by restricting Φ to be matrices of size dN(L, 0) given in (13) and the continuum is recovered in the limit L→ ∞. This construction is well suited to numerical evaluation. References [1] H. Grosse, C. Klimˇc´ık and P. Preˇsnajder, Int. J. Theor. Phys. 35, (1996) 231, [hep-th/9505175]; H.Grosse and A.Strohmaier, Lett. Math. Phys. 48, (1999) 163, [hep-th/9902138] [2] H. Grosse, C. Klimˇc´ık and P. Preˇsnajder, Comm. Math. Phys. 178, (1996) 507; H. Grosse and P. Preˇsnajder, Lett. Math. Phys. 46, (1998) 61 [3] P. Preˇsnajder, J. Math. Phys. 41 (2000) 2789, [hep-th/9912050] [4] A.P. Balachandran and S. Vaidya, Int. J. Mod. Phys. A16, (2001) 17, [hep-th/9910129]; A. P. Balachandran, T. R. Govindarajan and B. Ydri, Mod. Phys. Lett A15 (2000) 1279, [hep-th/9911087]; 9