Curvature homogeneity of type (1, 3) in pseudo-riemannian manifolds

Transcription

1 Curvature homogeneity of type (1, 3) in pseudo-riemannian manifolds Cullen McDonald August, 013 Abstract We construct two new families of pseudo-riemannian manifolds which are curvature homegeneous of type (1, 3) The first family given has signature (k, k + 1) and is curvature homogeneous of type (1, 3) but not curvature homogeneous The second family given has signature (1, ) and is curvature homogeneous of type (1, 3) of all orders but not locally homogeneous, showing there is no finite Singer number for this type of curvature homogeneity 1 Introduction Let (M, g) be a pseudo-riemannian manifold, and be its Levi-Civita connection We begin with some preliminaries and definitions 11 Curvature Homogeneous Manifolds Let X denote the ring of smooth vector fields on M Let T M denote the tangent bundle of M, and T M its dual The Riemann curvature tensor R of type (1,3), that is, R T M (T M) 3 is then given by R(X, Y )Z = X Y Z Y X Z [X,Y ] Z, (1) where X, Y, Z X The Riemann curvature tensor R is a tensor of type (0, 4), that is, R 4 (T M), and is given by R(X, Y, Z, W ) = g(r(x, Y )Z, W ) () Let p M, and T p M denote the tangent space of M at p Singer [8] defines a pseudo-riemannian manifold to be curvature homogeneous (CH) if for any points p, q M, there exists a linear isometry F : T p M T q M such that the pullback F satisfies F g q = g p and F R q = R p Let r R denote the r-th covariant derivative of R A pseudo-riemannian manifold is curvature homogeneous up to order n (CH n ) if there exists a linear isometry F which also satisfies F r R q = r R p for all r n 1

2 Singer proved the following theorem in the Riemannian setting, which motivated the study of curvature homogeneity [8] Theorem 11 Let M be an m-dimensional manifold There exists an integer k m so that if M is curvature homogeneous of order k m, M is locally homogeneous Later, Podesta and Spiro proved an analogous theorem for the pseudo- Riemannian case [7] Theorem 1 Let (M, g) be a pseudo-riemannian manifold of signature (p, q) There exists an integer k p,q so that if (M, g) is curvature homogeneous of order k p,q, (M, g) is locally homogeneous A generalization of this is given by Kowalski and Van zurová in [5] and [6] A pseudo-riemannian manifold is curvature homogeneous of type (1,3) (CH(1,3)) if for any pair p, q M, there exists a linear homothety φ : T p M T q M such that the pullback φ satisfies φ R q = R p We say that the manifold is curvature homogeneous of type (1,3) of order n (CH n (1, 3)) if for every r n there exists a linear homothety φ r whose pullback satisfies φ r r R q = r R p There has been quite a bit of work done on curvature homogeneous manifolds of higher signatures The cases most closely related to this paper include the ones which are of balanced signature researched by Dunn, Gilkey, and Nik cević in [3] and [], as well as those which are nearly balanced researched by Dunn in [1] CH(1,3) manifolds have been researched only in the Riemannian case by Kowalski and Van zurová in [5] and [6] We will make use of the following results in [6]: Theorem 13 Let (M, g) be a pseudo-riemannian manifold of dimension m, and suppose that for every 0 r n there exists and with an orthonormal moving frame {E 1, E m } on M, such that for fixed p M and any q M, r R(E i, E j, E k, E l ; E s1,, E sr ) q = Θ r (q) r R(E i, E j, E k, E l ; E s1,, E sr ) p where Θ r is a smooth positive function Then (M, g) is CH n (1, 3) 1 Model Spaces Let V be a finite dimensional real vector space A bilinear symmetric function σ : V R is an inner product on V if for every non-zero vector v, there exists some vector w such that σ(v, w) 0 We say a multilinear function R : V 4 R is an algebraic curvature tensor if R(x, y, z, w) = R(y, x, z, w) = R(z, w, x, y), R(x, y, z, w) + R(z, x, y, w) + R(y, z, x, w) = 0 We call the collection (V, σ, R) a model space The collection (V, R) is called a weak model space For example, for any p M, (T p M, g p, R p ) is a model

3 space, where g p and R p denotes the metric and curvature at p, respectively, and (T p M, R p ) is a weak model space Consider two model spaces (V, σ, R) and (W, υ, S) These model spaces are isomorphic if there exists an isomorphism of vector spaces H : V W which satisfies σ(x, y) = υ(h(x), H(y)), R(x, y, z, w) = S(H(x), H(y), H(z), H(w)) Two weak model spaces (V, R) (W, S) are isomorphic if there exists an isomorphism of vector spaces G : V W which satisfies R(x, y, z, w) = S(G(x), G(y), G(z), G(w)) Let ϕ be a symmetric bilinear form on V tensor given by R ϕ is an algebraic curvature R ϕ (x, y, z, w) = ϕ(x, w)ϕ(y, z) ϕ(x, z)ϕ(y, w) If there exists such a symmetric bilinear ϱ so that R = ±R ϱ, we say R is a canonical algebraic curvature tensor The following theorem may be found in [4] Theorem 14 Let ϕ be a symmetric bilinear form with rank greater than or equal to 3 If R ϕ = R ϱ, ϕ = ±ϱ It is evident that a space is curvature homogeneous if and only if the model spaces at every pair of points are isomorphic This motivates another kind of curvature homogeneity A pseudo-riemannian manifold is weakly curvature homogeneous if the weak model space at every pair of points are isomorphic 13 Weyl Scalar Invariants Let {x 1,, x m } be local coordinates on a pseudo-riemannian manifold (M, g) Adopt the Einstein convention of summing over repeated indices, and denote j1 jr R( x i1, xi ) xi3 = R i4 i 1i i 3j 1j r xi4 We may construct scalar invariants by contracting all indices with other curvature entries or with the metric tensors g ij and g ij A scalar invariant formed in this method is a Weyl scalar invariant They were defined by Weyl in [9] For our purposes, we will be most interested in the scalar curvature τ, the norm of the Ricci tensor ρ, and the norm of the curvature tensor R These scalars are defined by τ = g ij R k kij, ρ = g i1j1 g ij R k ki 1j 1 R l li j, R = g i1j1 g ij g i3j3 g i4j 4 R i4 i 1i i 3 R j4 j 1j j 3 3

4 14 Geodesics dγ A curve γ : [a, b] R M is a geodesic at t 0 if dγ dt dt t=t 0 = 0 We say γ is a geodesic if it is a geodesic at all t [a, b] If every geodesic in a pseudo- Riemannian manifold extends for infinite time, we say the manifold is complete 15 The Manifold M f,h Let M f,h = (R 4k+1, g f,h ), where R 4k+1 has coordinates given by (x 0, x 1,, x k, y 1, y k ), and the metric g f,h has non-zero entries: g f,h ( x0, x0 ) = e f(x1), g f,h ( xi 1, xi ) = h(x 0 ), 1 i k, g f,h ( xs, ys ) = 1, 1 s k, where f and h are smooth functions We prove the following theorem in Section Theorem 15 Let h and h be non-vanishing (a) τ, ρ, and R all vanish on M f,h (b) M f,h is CH(1,3) if and only if f = f (c) M f,h is CH if and only if h h (d) If M f,h is CH(1,3), then it is not complete if f is non-vanishing 16 The Manifold M f Let M f = (R 3, g f ), where R 3 has coordinates given by (x 0, x 1, y 1 ), and the metric g f has non-zero entries: g f ( x0, x0 ) = e f(x1), g f ( x1, y1 ) = 1, where f is smooth We prove the following theorem in Section 3 Theorem 16 Suppose (f ) f (a) τ, ρ, and R all vanish on M f (b) M f is CH(1,3) of all orders (c) M f is weakly CH if and only if (f ) + f is constant 4

5 The Geometry of M f,h Lemma 1 For M f,h, (a) The non-zero Christoffel symbols are given by: x0 x0 = f e f y1, x0 x1 = f x0 + h y, 1 i k x0 xi = h yi 1, x0 xl 1 = h yl, l k, xi 1 xi = h e f x 0 (b) Up to the usual Z symmetries, the non-zero entries of R are given by: R 1001 = e f ((f ) + f ), R (i 1)00(i) = h, R 1(i)0(i 1) = f h, R 1(l 1)0(l) = f h, R (i 1)(j)(i)(j 1) = (h ), 1 j k, ef R (i 1)(j 1)(i)(j) = (1 δ ij ) (h ) e f, where all indices denote subscripts of a x Proof The proof is evident, using the Christoffel symbols of the second kind, characterized by Γ ijk = g( xi xj, xk ), and may be calculated by Γ ijk = xi g( xj, xk ) + xj ( xi, xk ) xk g( xi, xj ) (3) From here, we may simply use equations (1) and () to calculate the curvature There are two moving frames β 1, β which will be particularly useful Let β 1 = {X 0, X 1,, Y 1,, Y k }, where X 0 = x 0 e f, X i 1 = λ i 1 ( xi 1 h yi ), X i = λ i ( xi h yi 1 ), Y s = 1 λ s ys, 5

6 where λ s is a smooth, non-vanishing function on M f,h Let β = { X 0, X 1,, Ȳ1,, Ȳk}, where X 0 = X 0, X s = 1 (X s + Y s ), Ȳ s = 1 (X s Y s ) Note that β is an orthonormal frame In the frame β, the curvature entries are given by: R 1001 = λ 1 ((f ) + f ), (4) R (i 1)00(i) = λ i 1λ i R 1(i)0(i 1) = λ 1λ i 1 λ i h R 11(l 1)0(l) = λ 1λ l 1 λ s R (i 1)(j)(i)(j 1) = λ i 1λ i λ j 1 λ j 4 R (i 1)(j 1)(i)(j) = λ i 1λ i λ j 1 λ j 4, ef (5) f h e f, (6) f h, ef l 1, (7) where each subscript may be the subscript of either a X or Ȳ (h ), (8) ef (1 δ ij ) (h ), (9) ef 1 A few Weyl Scalar Invariants on M f,h Proof of Theorem 15 (a) We note that it is simpler to calculate the listed invariants in the orthonormal frame, as then we need only consider terms which have g ss or g ss Then one must take care to calculate the entries of the (1, 3) tensor This may be done quite readily from the entries of the (0, 4) tensor, as the metric is reduced to an orthonormal one It is then immediate that τ, ρ, and R vanish, as any curvature entry which uses an X will be cancelled out by the entry with a corresponding Ȳ in that spot, as g(x s, X s ) = 1 and g(y s, Y s ) = 1 To demonstrate this, we will look specifically at τ 6

7 Using our orthonormal frame, τ is given by: τ = s,r g ss g rr R rssr = g( X 0, X 0 )R 1001 (g( X 1, X 1 ) + g(ȳ1, Ȳ1)) + g( X 0, X 0 )R 0110 (g( X 1, X 1 ) + g(ȳ1, Ȳ1)) + i k R (i)(i 1)(i 1)(i) g( X i 1, X i 1 )(g( X i, X i ) + g(ȳi, Ȳi)) + i k R (i)(i 1)(i 1)(i) g(ȳi 1, Ȳi 1)(g( X i, X i ) + g(ȳi, Ȳi)) + i k R (i 1)(i)(i)(i 1) g( X i, X i )(g( X i 1, X i 1 ) + g(ȳi 1, Ȳi 1)) + i k R (i 1)(i)(i)(i 1) g(ȳi, Ȳi)(g( X i 1, X i 1 ) + g(ȳi 1, Ȳi 1)) Note that every entry has a term g(x s, X s ) + g(y s, Y s ), which is equal to 0 Therefore, the scalar curvature vanishes, and one may use a similar set up to show ρ and R also vanish Curvature Homogeneity on M f,h Proof of Theorem 15 (b) Case I: f 0 We operate in the orthonormal frame β Suppose that each of these entries is non-zero, and that all are equal to some function a Note that because Equation (8) holds even if i = j, Equation (9) in conjunction with Equation (8) give us that λ i 1 λ i = λ j 1 λ j for all i, j Then without loss of generality, we may proceed using only λ 1 and λ Dividing the square of Equation (5) by Equation (8) gives us ( ) h a = h e f (10) Note that because h, h 0, a is both well-defined and positive Dividing Equation (6) by Equation (8) yields f h e f 1 = λ (h ) e f, which implies f e f λ = h However, we may also attain λ by dividing Equation (4) by Equation (6) After simplification, we obtain λ = e f ((f ) + f ) f h 7

8 Both expressions must be valid, so we get that f satisfies f = (f ) Now that we have λ, we may use it in conjuction with Equations (5) and (10) to find h e f λ 1 = h f For all other λ i, they need only satisfy λ i 1 λ i = h (h ) to make all non-vanishing curvature entries equal to ±a at every point For some fixed p M f,h, we may define a function ϕ p by We have then that ϕ p (q) = a(q) a(p) R ijkl (q) = ϕ p (q)r ijkl (p) for all q M f,h As ϕ p is smooth and positive and we are on an orthonormal frame, we have M f,h is CH(1,3) by Theorem 13 Case II: f = 0 If f = 0, we have non-zero curvature entries given in β by R (i 1)00(i) = λ i 1λ i R (i 1)(j)(i)(j 1) = λ i 1λ i λ j 1 λ j 4 R (i 1)(j 1)(i)(j) = λ i 1λ i λ j 1 λ j 4 h, (11) ef (h ), (1) ef (1 δ ij ) (h ) (13) ef Assume all non-zero entries are equal to some function a We may use a similar trick from the first part to get that λ 1 λ = λ i 1 λ i We divide the square of Equation (11) by Equation (1) and get once again, ( ) h a(p) = h e f Now we have that all non-zero entries are this function, provided λ i 1 λ i = h (h ) We see that as we are still on the orthonormal frame β, we may define ϕ in an identical manner so M f,h is CH(1,3) 8

9 Consider the symmetric bilinear form φ p on T p M, which has non-zero entries given the following on the basis β 1 φ p (X 0, X 0 ) = 1, φ p (X 0, X 1 ) = ɛ(f), φ p (X 0, Y 0 ) = ɛ(f), φ p (X i 1, X i ) = 1, φ p (X i 1, Y i ) = 1, φ p (Y i 1, Y i ) = 1, where ɛ(f) = 0 if f = 0, and is otherwise 1 After a few calculations, it is verified that the canonical algebraic curvature tensor R a(p)φ p is the Riemann curvature tensor at p Proof of Theorem 15 (c) If we have a scalar invariant of the 0-model which varies from point to point, the space will not be curvature homogeneous Consider the trace of the form a(p)φ p This is not invariant to the space, as the curvature tensor is also defined by a(p)φ p But by Theorem 14, these are the only canonical algebraic curvature tensors possible Therefore, a quantity invariant to both forms is invariant to the space The square of the trace is such a quantity We calculate the trace by looking at the self-adjoint operator a(p)φ p associated to a(p)φ p in the orthonormal basis β This is given by: a 1 ɛ(f) ɛ(f) ɛ(f) ɛ(f) It is easy to see then that the trace of ± a(p)φ p = a(p) This invariant is then equal to a(p), which is constant for all p if and only if h = c 0 h, where c 0 is a constant Conversely, if a(p) is constant for all p, then R ijkl (q) = R ijkl (p), so (M, g) is CH 9

10 3 Incompleteness of M f,h Proof of Theorem 15 (d) If M f,h is CH(1,3) and f 0, then f is given by f = 1 log(x 1 + c 1 ) + c, where c 1 and c are arbitrary constants M f,h is defined only for x 1 > 1 c 1 Consider the curve γ(t) = (0, 1 c 1(1 t), 0,, 0) Differentiating, we get and it follows from Lemma 1 that dγ dt = c 1 x1, dγ dγ dt dt = (c 1) x1 x1 = 0 γ is therefore a geodesic It is not defined for t = 1, so the space is not complete 3 The Geometry of M f Lemma 31 For M f, (a) The non-zero Christoffel symbols are given by: x0 x0 = f e f y1, x0 x1 = f x1 (b) Let = (f ) +f Up to the usual Z symmetries, the non-zero curvature entry of R is given by: R 1010 = e f, where all indices are of some x (c) Up to the usual Z symmetries, the non-zero entry of r R is given by: r R 1010;11 = e f (r) Proof As in Lemma 1, we may use Equation 3 to calculate part (a), and Equations 1 and to calculate part (b) For part (c), we note that the non-zero curvature entry can only be differentiated by x1, as differentiating by y1 will yield zero outright by (a), and differentiating by x0 must be zero because of the symmetries The derivative of the exponential portion will cancel out when the x0 entries are differentiated by x1 For this manifold, the derivative of the zero entries of R and r R is always zero 10

11 We construct two particularly useful moving frames on our manifold b 1 and b Let b 1 be given by: X 0 = x 0 e f, X 1 = λ x1, Y 1 = 1 λ y 1, where λ is a smooth, non-vanishing function on M f b is an orthonormal basis given by: X 0 = X 0, X 1 = X 1 + Y 1, Ȳ 1 = X 1 Y 1 In b, the curvture entries are given by: r R 1010;11 = λr+ r (r), where the index 1 may be an index of either X 1 or Ȳ1 Proof of Theorem 16 (a) We operate in the orthonormal frame b As in the case of M f,h, each entry with a g( X 1, X 1 ) will be cancelled out by an entry with a g(ȳ1, Ȳ1) We calculate τ to demonstrate this In an orthonormal frame, we have τ = s,r g ss g rr R rssr = g( X 0, X 0 )R 1001 (g( X 1, X 1 ) + g(ȳ1, Ȳ1)) + g( X 0, X 0 )R 0110 (g( X 1, X 1 ) + g(ȳ1, Ȳ1)) Each term has g( X 1, X 1 ) + g(ȳ1, Ȳ1) in it, which is zero Therefore, τ vanishes on M f In a similar method, one may show ρ and R also vanish Proof of Theorem 16 (b) We use the frame b Fix some point p M, and define a function ω : M f R defined by ω(q) = r R 1010;11 (q) r R 1010;11 (p) At any point q M f, the non-zero curvature entries are given by ω(q) r R 1010;11 (p) We may do this independently for each r By Theorem 13, M f is CH r (1, 3) Lemma 3 M f is weakly CH 1 11

12 Proof Consider the frame b 3 = { X 0, X 1, Ỹ1}, where X 0 = X 0 X 1 = µx 1 Ỹ 1 = Y 1, where µ = ( ) 1 In this frame, then, every non-zero curvature entry is given by ±1 The space is then weakly curvature homogeneous Consider all bases which preserve this weak 1-model Represented in b 3, these are bases obtained under the action of the matrix A = ±d 1 b 0 0 d 0, s t c where d, c are non-zero Proof of Theorem 16 (c) Suppose M f is weakly CH Then the function R 1010;11 ( R 1010;1 ) must be constant in all frames of the form Ab 3, as these are the frames where R 1010;1 is constant, and if the space is weakly CH n, some frame must make all curvature entries constant Thus this quantity may be expressed as R 1010;11 ( R 1010;1 ) = d d = This can only be constant if is constant 4 Further Work The work done on this paper motivate a few new questions Can the involved metric be generalized further and preserve CH(1,3)? Do all Weyl scalar invariants vanish on these manifolds? In general, the manifolds given are not complete non-extendible? Are these manifolds Is it possible to have a space which is CH(1,3) of order 1, but not curvature homogeneous? Does WCH n imply CH n+1 (1, 3)? 1

13 5 Acknowledgements The author would like to thank C Dunn for his valuable insights throughout the entire research process This research was jointly funded by NSF grant DMS , and by California State University, San Bernardino References [1] C Dunn, A new family of curvature homogeneous pseudo-riemannian manifolds, Rocky Mountain Journal of Mathematics Vol 39, No 5 (009), pp [] C Dunn and P Gilkey, Curvature homogeneous manifolds which are not locally homogeneous, Complex, Contact and Symmetric Manifolds Vol 34 (005), pp [3] C Dunn, P Gilkey, and S Nik cević, Curvature homogeneous signature (,) manifolds, Differential Geometry and its Applications, Proceedings of the 9th International Conference (004), pp 9-44 [4] P Gilkey, The Geometry of Curvature Homogeneous Pseudo-Riemannian Manifolds, Imperial College Press (007) [5] O Kowalski and A Van zurová, On curvature homogeneous spaces of type (1,3), Mathematische Nachrichten Vol 84, No (011), pp [6] O Kowalski and A Van zurová, On a generalization of curvature homogeneous spaces, Results in Mathematics Vol 63, No 1- (013), pp [7] F Podesta and A Spiro, Introduzione ai Gruppi di Transformazioni, Volume of the Preprint Series of the Mathematics Department V Volterra of the University of Ancona, Via delle Brecce Bianche, Ancona, Italy, (1996) [8] I Singer, Infinitesimally homogeneous spaces, Communications on Pure and Applied Mathematics Vol 13, No 4 (1960), pp [9] H Weyl, The Classical Groups Princeton University Press, Princeton (1946) 13