5.4. Producción estadística de la AGE
5.4.4. Estadísticas de síntesis y no sectoriales
Recall that a bilinear form on a real-valued,k-dimensional vector space V is a binary mapα:V ×V →R, such that:
α(u+v, w) =α(u, w) +α(v, w)
α(u, v+w) =α(u, v) +α(u, w) α(r·u, v) =r·α(u, v) =α(u, r·v)
for everyu, w, v∈V andr∈R. We will denote the space of all bilinear forms onV byT2(V∗).14The spaceT2(V∗) can be turned into a vector space using
the operations of pointwise addition and scalar multiplication of functions.15
The dimension of the spaceT2(V∗) is equal tok2, with a natural basis given
by the collection{ǫi⊗ǫj |i, j≤n},where{ǫi}ni=1 forms a basis for the dual
spaceV∗.16
In order to eventually define what it means for a smooth manifoldM to be a spacetime, we will need to discuss the bilinear forms built from elements T M. In this case, we can use the fibresT2(T∗
pM) and apply the Bundle chart
lemma to equip the set
T2(T∗M) := G
p∈M
T2(T∗
pM)
with a vector bundle structure. The exact details of this construction are not too important for our purposes, though we will remark that the local trivialisations ofT2(T∗M) are given by:
Φ(p, α) =Φ(p, rijdϕi⊗dϕj) = (p, r11, ..., r1n, ..., rn1, ..., rnn),
where we here we denote by dϕi the dual-basis element associated to the
directional derivative ∂/∂ϕi. Local charts of the bundle T2(T∗M) are then
given by:
π−1(U)7→U×
Rn2 7→ϕ(U)×Rn2
in the obvious way, and the resulting spaceT2(T∗M) becomes a vector bundle
of rankn2overM.17
Given a smooth mapf between smooth manifolds M and N, there is a natural sense in which (some of) the bilinear forms onN can be transferred toM. Suppose we have some element of the form (f(p), α) inT2(T∗
f(p)N). We
can then define the objectf∗αinT2(T∗
pM) by:
f∗α(v, w) =α(dfp(v), dfp(w))
for allv, w∈TpM. Observe that sincef is smooth, by Proposition A.10.4 the
images of v and w under the differential dfp are indeed elements ofTf(p)N,
so this definition makes sense. Moreover, the formf∗αinherits its bilinearity
fromα, so (p, f∗α) lies in the fibreT2(T∗
pM). We will refer to the objectf∗α
as the (pointwise) pullback ofαunder f.18
14This space is also commonly denoted asV∗
⊗V∗
, orT(0,2)V, orT20(V). 15That is, we define (α+β)(v, w) :=α(v, w) +β(v, w) and (r·α)(v, w) =r·α(v, w). 16The dual space for a vector space V is the set of all linear mapsα:V →R. If
{ei} is a basis forV, then the collection {ǫi} of maps defined byǫi(ej) = 1 iff
i=j andǫi(ej) = 0 iffi6=jforms a basis forV ∗
.
17When viewed as a smooth manifold,T2(T∗
M) is of dimensionn+n2.
18There is a sense in which fields of bilinear forms (and in general, covariant tensor
There is also a sense in which the pointwise pullback has a restricted dual. In the case where M and N have the same dimension and f is a smooth embedding (which by Lemma A.12 means that M and f(M) are diffeomor- phic), we can also push bilinear forms of M forward into N. Given some (p, β)∈T2(T∗
pM), we can define the mapf∗β to act as:
f∗β(v, w) =β(v′, w′), wherev=dfp(v′) andw=dfp(w′).
Observe that wheneverf is a diffeomorphism, the pointwise differentialdfpis
a bijection (see Prop. A.10.4), and since f(M) is a submanifold of the same dimension as N, we can use the maps as in the proof of Lemma A.12 to conclude that every element of the tangent spaceTf(p)N is of the formdfp(v)
for some vectorv in TpM, and thus f∗β is a well-defined bilinear form that
lives in the fibre T2(T∗
f(p)N). We have the following facts about pointwise
pullbacks and pushforwards.
Lemma A.19.Let f : M →N andg :N →P be smooth embeddings, with dim(M) = dim(N) = dim(P), and let α, β and δ be elements of the fibres T2(T∗ pM), T2(Tf(p)∗ N)andT 2(T∗ g◦f(p)P)respectively. 1.(idM)∗α= (idM)∗α=α 2.f∗α= (f−1)∗α 3.f∗β = (f−1) ∗β 4.(f∗◦g∗)δ= (g◦f)∗δ 5.f∗(f ∗α) =α 6.g∗◦f∗(α) = (g◦f)∗α
We saw in Proposition A.18 that wheneverf :M →N is a smooth map, the mapdf (defined fibrewise as the pointwise differential mapsdfp) becomes
a smooth bundle morphism from T M toT N. Similarly, we can also define a map that acts on fibres as the pointwise pullback (pushforward) does. Just as the differential map associated to a diffeomorphism is a bundle isomorphism between tangent bundles, there is an associated result for the bundle of bilinear forms.
Lemma A.20.Iff :M →N is a diffeomorphism, then the bundlesT2(T∗M)
andT2(T∗N)are isomorphic.
We remark that in the proof of the above result, the isomorphism is given by the map ξ: T2(T∗M)→ T2(T∗N) defined by ξ−1(p, α) = (f(p), f
∗α). This
will be important in Section 7.1, when we adjoin bundles of bilinear forms.