Tensors and Differential Forms on Vector Spaces

APPENDIX A TENSORS AND DIFFERENTIAL FORMS ON VECTOR SPACES Since only so much of the vast and growing field of differential forms and differentiable manifolds will be actually used in this survey, we shall attempt to briefly review how the calculus of exterior differential forms on vector spaces can serve as a replacement for the more conventional vector calculus and then introduce only the most elementary notions regarding more topologically general differentiable manifolds, which will mostly be used as the basis for the discussion of Lie groups, in the following appendix. Since exterior differential forms are special kinds of tensor fields – namely, completely-antisymmetric covariant ones – and tensors are important to physics, in their own right, we shall first review the basic notions concerning tensors and multilinear algebra. Presumably, the reader is familiar with linear algebra as it is usually taught to physicists, but for the “basis-free” approach to linear and multilinear algebra (which we shall not always adhere to fanatically), it would also help to have some familiarity with the more “abstract-algebraic” approach to linear algebra, such as one might learn from Hoffman and Kunze [1], for instance. 1. Tensor algebra. – A tensor algebra is a type of algebra in which multiplication takes the form of the tensor product. a. Tensor product. – Although the tensor product of vector spaces can be given a rigorous definition in a more abstract-algebraic context (See Greub [2], for instance), for the purposes of actual calculations with tensors and tensor fields, it is usually sufficient to say that if V and W are vector spaces of dimensions n and m, respectively, then the tensor product V ⊗ W will be a vector space of dimension nm whose elements are finite linear combinations of elements of the form v ⊗ w, where v is a vector in V and w is a vector in W. The tensor product ⊗ then takes the form of a bilinear map V × W → V ⊗ W, (v, w) ֏ v ⊗ w. In particular, that kind of product is not closed, since the tensor product of two vectors will belong to a different vector space. Bilinearity means that the map is linear in each factor individually, but not collectively linear. Hence: (α v + β v′) ⊗ w = α v ⊗ w + β v′ ⊗ w, v ⊗ (α w + β w′) = α v ⊗ w + β v ⊗ w′. (1.1) (1.2) One can also see that this means that the tensor product is right and left distributive over vector addition. Because of this bilinearity, if {ei , i = 1, …, n} is a basis for V and {fa , a = 1, …, m} is a basis for W then {ei ⊗ fa , i = 1, …, n, a = 1, …, m} will constitute a basis for V ⊗ W. <ul style="display: flex;"><li style="flex:1">396 </li><li style="flex:1">Appendix A – Tensors, differential forms on vector spaces </li></ul>Hence, if t is an element of V ⊗ W then it can be expressed as a linear combination of the basis elements in the form: <ul style="display: flex;"><li style="flex:1">n</li><li style="flex:1">m</li></ul>t = tia ei ⊗ fa ≡ tia e ⊗f . (1.3) ∑∑ <ul style="display: flex;"><li style="flex:1">i</li><li style="flex:1">a</li></ul>i=1 a=1 The numbers tia are referred to as the components of t with respect to the chosen basis on V ⊗ W. Most of the literature of theoretical physics, even up to the present era, deals exclusively with the components of tensors, although if brevity be the soul of wit then one can easily see that dealing with the intrinsic objects, such as v, w, and t, can add a certain clarity and conciseness to one’s mathematical expressions, even if it does involve a small investment of abstraction in the process. One can also use the bilinearity of the tensor product to express the tensor product v ⊗ w in terms of components. Suppose that v = vi ei and w = wa fa (1). One will then have: v ⊗ w = v i wa ei ⊗ fa ; (1.4) i.e., the components of v ⊗ w with respect to the chosen basis will be v i wa. We now see that there are two distinct types of elements in V ⊗ W, namely, decomposable elements which have the form v ⊗ w, and indecomposable elements, which have the more general form of finite linear combinations of decomposable elements, such as in (1.3). The fact that not all elements are decomposable is due to the fact that linear combinations of decomposable elements do not have to be decomposable. In the case of exterior algebra, which we shall discuss shortly, the decomposable elements will sometimes define quadric hypersurfaces in the tensor product space, rather than vector subspaces. b. Contravariant tensors. – A common situation in tensor algebra (as well as in physics) is when the vector space W is the vector space V. One can then refer to the elements of V ⊗ V as second-rank contravariant tensors (over V). The term “second- rank” refers to the fact that there are two copies of V in the tensor product. Hence, a basis can be defined by {ei ⊗ ej , i, j = 1, …, n} and components will look like tij : t = tij ei ⊗ ej . (1.5) The term “contravariant” refers to the way that the components transform under a change of basis. In particular, if: e =ej Aj (1.6) iiis a change of linear basis in V (so Aj is an invertible matrix) then the components of v iand w (which were v i and w i with respect to ei) will now be: (1) From now on, we shall invoke the “summation convention,” which is often attributed to Einstein, namely, whenever a superscript agrees with a subscript, one sums over all defined values of the index in question. In the occasional situations where it is necessary to refer to components with doubled indices, such as the diagonal elements of matrices, the convention will usually be rescinded explicitly. <ul style="display: flex;"><li style="flex:1">§ 1. – Tensor algebra. </li><li style="flex:1">397 </li></ul><ul style="display: flex;"><li style="flex:1">i</li><li style="flex:1">j</li><li style="flex:1">i</li><li style="flex:1">j</li></ul>ii<ul style="display: flex;"><li style="flex:1">ɶ</li><li style="flex:1">ɶ</li></ul><ul style="display: flex;"><li style="flex:1">v = Aj v , </li><li style="flex:1">w = Aj w </li></ul>(1.7) iiɶwith respect to e . The notation Aj refers to the inverse of the matrix Aj , so this type of itransformation is referred to as contravariant. From the bilinearity of the tensor product, the resulting transformation of the components of v ⊗ w will be: <ul style="display: flex;"><li style="flex:1">i</li><li style="flex:1">j</li><li style="flex:1">i</li></ul>k<ul style="display: flex;"><li style="flex:1">j</li><li style="flex:1">k</li><li style="flex:1">l</li></ul>ɶ ɶ v w = A A v w , (1.8) land more generally (also due to the bilinearity of the tensor product), the components tij of t, as in (1.5), will transform to: ij i k<ul style="display: flex;"><li style="flex:1">j</li><li style="flex:1">kl </li></ul>ɶ ɶ t = A A t . (1.9) lHence, one can say that they are doubly-contravariant. Since the tensor product is also associative: (a ⊗ b) ⊗ c = a ⊗ (b ⊗ c), (1.10) one can define higher tensor products V ⊗ … ⊗ V of a finite number of copies of V, and the elements of ⊗kV = V ⊗ … ⊗ V when there are – say – k copies of V are then referred to as rank-k contravariant tensors over V. Hence, a basis for ⊗kV can be given by all tensor products of the form e ⊗ … ⊗ e , the components of a general element t ∈ ⊗kV i1 ik i ⋯ik 1<ul style="display: flex;"><li style="flex:1">will take the form t </li><li style="flex:1">, and they will transform to: </li></ul><ul style="display: flex;"><li style="flex:1">i1⋯ik </li><li style="flex:1">i1 </li><li style="flex:1">ik j1⋯ jk </li></ul><ul style="display: flex;"><li style="flex:1">ɶ</li><li style="flex:1">ɶ</li></ul><ul style="display: flex;"><li style="flex:1">t</li><li style="flex:1">= Aj ⋯Aj t </li></ul>(1.11) 1kunder a change of basis on V. Clearly, the dimension of ⊗kV will be nk. One can also form tensors of mixed rank over V by forming finite linear combinations of elements in various ⊗kV’s for different values of k. For instance, one might form the linear combination a + b ⊗ c. Such expressions cannot generally be simplified further, unless some further structure is imposed upon ⊗kV, which will usually be another algebra, for us. A tensor that does not have mixed type is referred to as homogeneous, although in most cases, that will be tacitly implied. The direct sum ⊗0 V ⊕ ⊗1V ⊕ … ⊕ ⊗kV ⊕ … of all the vector spaces ⊗kV as k varies from 0 (⊗0V ≡ R) to infinity will denoted by simply ⊗*V. It will be referred to as the contravariant tensor algebra over V, and will clearly be infinite-dimensional. One sees that the tensor product then makes the vector space ⊗*V into an algebra, since it defines a bilinear binary product on the vector space. It will also be associative and possess a unity element (namely, 1) as an algebra. In addition, since the tensor product of a homogeneous tensor of rank k with one of rank l will be a homogeneous tensor of rank k + l, one refers to the algebra ⊗*V as a graded algebra. <ul style="display: flex;"><li style="flex:1">398 </li><li style="flex:1">Appendix A – Tensors, differential forms on vector spaces </li></ul>c. Covariant tensors. – The dual space V* to the vector space V (viz., the vector space of all linear functionals on V) is itself a vector space, so one can still define tensors of any rank over it. Hence, ⊗kV ≡ ⊗kV* = V* ⊗ … ⊗ V* (k copies) is a vector space of idimension nk, and if {θ , i = 1, …, n} is a basis for V* then a basis for ⊗kV can be defined ik ki1 by θ ⊗ … ⊗ θ , and a general element t ∈⊗ V will then take the form: ik i1 t = ti ⋯i θ ⊗ … ⊗ θ . (1.12) 1kThe scalar components ti ⋯i are then the components of a rank-k covariant tensor over V. 1kUnder a change of basis on V* : <ul style="display: flex;"><li style="flex:1">i</li><li style="flex:1">j</li></ul>iɶθ = Aj θ , (1.13) (1.14) the components ti ⋯i will transform to: 1kt= Aj1 ⋯ Ajk tj ⋯ j .i1⋯ik i1 ik 1kiOf particular interest is the case in which the basis θ for V* is reciprocal to the basis ei for V. In that case, one will have, by definition: 1 i = j 0 i ≠ j. iθ (ej) = δ i = (1.15) jδ ij is then the Kronecker delta symbol. In order for a reciprocal basis to go to a reciprocal basis under a change of basis ei , such as in (1.6), one must have: lδ ij = θi (ej ) = Bi θ (ek Akj ) = Bi Akj θl (ek ) = Bi Akj δkl = Bki Akj . <ul style="display: flex;"><li style="flex:1">l</li><li style="flex:1">l</li><li style="flex:1">l</li></ul>iiiɶHence, Bj can only be Aj . One then says that the basis θ transforms contragrediently to the basis ei , and in fact, so do the components of tensors over V*. One can also form the direct sum ⊗*V ≡ ⊗*V* = ⊗0V ⊕ ⊗1V ⊕ … ⊕ ⊗kV ⊕ … over all k and call it the algebra of covariant tensors over V. It will once more be an infinitedimensional graded associative algebra with unity. d. Tensors mixed variance. – One can take the tensor product ⊗kV ≡ (⊗kV) ⊗ (⊗ V) lland obtain a vector space of dimension nk+l whose elements are finite linear combinations of (homogeneous) elements of the form αi1 ⊗ … ⊗ αik ⊗ v j ⊗ … ⊗ v j , in which the α’s 1lare linear functionals on V, and the v’s are vectors in V. Such an element will then be a tensor that is k-times covariant and l-times contravariant. Hence, a basis for ⊗kV can be lik i1 defined by the tensor products θ ⊗ … ⊗ θ ⊗ ei ⊗ … ⊗ ei , and a general element t 1l∈⊗kV will take the component form: l<ul style="display: flex;"><li style="flex:1">§ 1. – Tensor algebra. </li><li style="flex:1">399 </li></ul>j1⋯ jl ii⋯ik ik i1 t = t θ ⊗ … ⊗ θ ⊗ ei ⊗ … ⊗ ei . (1.16) 1lUnder a change of basis (1.6) for V and the contragredient change of the reciprocal basis for V*, the components t j1⋯ jl will become: ii⋯ik j1⋯ jl ii⋯ik r i1 rk ik

Tensors and Differential Forms on Vector Spaces

Details

Download

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

Support