whereC, D(Z)3 are as above given by (1.5.107). Note that again by (1.5.83) and (1.5.81) D3(Z) ≤C(n, m, α, t0)×
× Z
Σ0
JN(φ) +
n(n−1)
X2
i=1
JN(Ωiφ) +
n(n−1)
X2
i,j=1
JN(ΩiΩjφ) +
n(n−1)
X2
i,j,k=1
JT(ΩiΩjΩkφ), n
+ Z
Σ0
JZ,1(φ) +
n(n−1)
X2
i=1
JZ,1(Ωiφ) +
n(n−1)
X2
i,j=1
JZ,1(ΩiΩjφ), n .
for some 0< δ < 14, and R > n−2q
8nm
δ , then there is a constant C(n, m, δ, R) such that for r0 < r < R,
|∂tφ(t, r)| ≤ C√ D
τ2−2δ (τ = 1
2(t−R∗)> τ0). (1.6.4) The pointwise bounds are obtained from the energy estimates of Section 1.5 using Sobolev inequalties and elliptic estimates; the former provide the link between pointwise and integral quantities, and the latter allow for the expression of these integral quantities in terms of higher order energies.
Sobolev embedding. By the extension theorem applied to the Sobolev embedding Hs(Rn)⊂L∞(Rn) (s > n2) [27] we have, for r0 < r < R,
|φ(t, r)|2 ≤C(n) Z R∗
r∗0
dr∗ Z
Sn−1
dµγ◦
n−1
( φ2+
|α|≤[n2]+1
X
|α|≥1
∇αφ 2
) rn−1
t=t (1.6.5) where∇ denote the tangential derivatives to the hypersurface Σt, andα denotes a multi- ndex of order n.
Elliptic estimates. Note that for any solution φ of the wave equation T2·φ= ∂2φ
∂r∗2 + 1− 2m rn−2
n−1 r
∂φ
∂r∗ + 1− 2m rn−2
△/
r2γ◦n−1φ .
=L·φ (1.6.6) where the operator
L= 1− 2m rn−2
gij∇i∂j (1.6.7)
is clearly elliptic, (heregt=g|Σt denotes the restriction ofg to the spacelike hypersurfaces Σt, a Riemannian metric on Σt, andi, j = 1, . . . , n). In view of the standard higher order interior elliptic regularity estimate (c.f. [27]),
kφkHm+2(bΣt)≤C
kL·φkHm(bΣt)+kφkL2(bΣt)
Σbt .
= Σt∩ {r0 < r < R}, (1.6.8) we conclude with (1.6.5) that in the case where [n2] + 1 is even,
|φ|2 ≤C(n, m) Z R∗
r∗0
dr∗ Z
Sn−1
dµ◦
γn−1 [n2]+1
X
l=0
Tl·φ2
rn−1; (1.6.9) in general we have:
Lemma 1.54 (Pointwise estimate in terms of higher order energies). Let φ be a solution of the wave equation (1.1.1), and n ≥3. Then there exists a constant C(n, m) such that for all r0 < r < R:
|φ(t, r)|2 ≤C(n, m)
kφk2L2(bΣt)+ Z
Σbt
[n2]
X
l=0
JT(Tl·φ), n
(1.6.10)
Proof of Prop. 1.53. In view of the Lemma 1.54 and the energy decay estimates of Section 1.5 it remains to control the zeroth order term kφkL2(bΣt); we multiply the integrand by (Rr)2 ≥1 and extend the integral to u∗ =τ = 12(t−R∗), v∗ ≥ 12(t+R∗).
(i) By Lemma B.6 we can then estimatekφk2L2(bΣt) by the energy flux through Στ=1
2(t−R∗), and apply Prop. 1.37 to the higher order energies of Lemma 1.54.
(ii) Here we extend the integral only toτ+R∗ ≤v∗ ≤τ+R∗+τ3 and apply Lemma B.8 to obtain
Z R∗
r0∗
dr∗ Z
Sn−1
dµ◦
γn−1(∂tφ)2rn−1≤C(n, m)R2 Z
Στ∩{r∗≤R∗+τ3}
JT(∂tφ), n
+C(n, m)R2 r
Z
Sn−1
rn−1(∂tφ)2|(u∗=τ,v∗=τ+R∗+τ3). (1.6.11) As in the proof of Lemma 1.42 we obtain by integrating from infinity and Cauchy’s inequality that
Z
Sn−1
dµγ◦
n−1rn−2(∂tφ)2(τ, τ +R∗)≤ C(n, m) 1−R2mn−2
Z
Στ
JT(∂tφ), n
(1.6.12) which decays by Prop. 1.37 with a rate τ−2. Moreover, as in the proof of Lemma 1.42,
Z
Sn−1
dµγ◦
n−1rn−1(∂tφ)2|(u∗=τ,v∗=τ+R∗+τ3) =
= Z
Sn−1
dµ◦
γn−1rn−1(∂tφ)2|(u∗=τ,v∗=τ+R∗)+
Z τ+R∗+τ3
τ+R∗
dv∗ Z
Sn−1
dµ◦
γn−12∂tψ ∂∂tψ
∂v∗ |u∗=τ
(1.6.13) and
Z τ+R∗+τ3
τ+R∗
dv∗ Z
Sn−1
dµ◦
γn−1∂tψ ∂∂tψ
∂v∗ |u∗=τ ≤
≤
sZ ∞
τ+R∗
Z
Sn−1
dµγ◦
n−1
1
r2(∂tφ)2rn−1×
sZ ∞
τ+R∗
Z
Sn−1
dµγ◦
n−1r2∂rn−12 ∂tφ
∂v∗ 2
, (1.6.14) the first factor decaying with a rateτ−1 by Lemma B.6 and Prop. 1.37, and the second factor bounded by the weighted energy inequality for rn−12 ∂tφ in place of ψ with p = 2.
Therefore Z
Sn−1
rn−1(∂tφ)2|(u∗=τ,v∗=τ+R∗+τ3)≤ C(n, m) 1− R2mn−2
D
τ . (1.6.15)
By virtue of Prop. 1.39, compare in particular Remark 1.44 on page 98, the first term on the right hand side of (1.6.11) decays with a rate of τ4−4δ, and this is matched by the second term in view of the prefactor r−1 = (R∗ +τ3)−1, which is the result of our choice of powers of τ in the extension of the integral (1.44). Lemma 1.54 applied to the solution∂tφ of (1.1.1) then yields the pointwise decay result (1.6.4) after having applied Prop. 1.39 to the higher order energies on the right hand side of (1.6.10).
Interpolation. We shall now interpolate between the results Prop. 1.53 (i) and (ii) to improve the pointwise estimate for |φ|. Our argument can in some sense be compared to the proof of improved decay in [36]. The basic observation underlying this argument is that for r0 < r < R and t1 > t0
rn−2φ2(r, t1) =rn−2φ2(r, t0) + Z t1
t0
2φ(t, r)∂φ
∂t(t, r)rn−2dt
≤rn−2φ2(r, t0) + 1 t1−2δ0
Z t1
t0
φ2(t, r)rn−2dt+t1−2δ0 Z t1
t0
∂φ
∂t 2
(t, r)rn−2dt . (1.6.16) Moreover, as a consequence of Lemma 1.55,
rn−2φ2(t, r)≤Rn−2φ2(t, R) + 1− 2m r0n−2
−1Z R∗ r∗
∂φ
∂r∗ 2
rn−1dr∗, (1.6.17) we obtain an estimate for the timelike integrals in terms of the corresponding integrals at r=R and spacetime integrals, using the Sobolev inequality on the sphere:
Z t1
t0
rn−2φ2(t, r) dt≤ Z t1
t0
dt Z
Sn−1
dµγ◦
n−1
X
|α|≤[n2]+1
Rn−2 Ωαφ2
(t, R)
+ 1− 2m r0n−2
−1Z t1
t0
dt Z R∗
r∗
dr∗ Z
Sn−1
dµγ◦
n−1rn−1 X
|α|≤[n2]+1
∂Ωαφ
∂r∗ 2
(t, r) (1.6.18)
Lemma 1.55. Let a < b∈R and φ ∈C1([a, b]) then an−2φ2(a)≤bn−2φ2(b) +
Z b
a
dφ dx
2
xn−1dx (1.6.19) for all n ≥3.
Proof. Since, by integration by parts, Z b
a
2φ(x)dφ
dx(x)xn−2dx= 2φ2(x)xn−2|ba
− Z b
a
2φ(x)dφ
dx(x)xn−2dx− Z b
a
2φ2(x)(n−2)xn−3dx , it clearly follows, with Cauchy’s inequality,
an−2φ2(a)≤bn−2φ2(b) + Z b
a
dφ dx
2
xn−1dx +
1−(n−2)Z b a
1
x2φ2(x)xn−1dx .
Proposition 1.56 (Improved interior pointwise decay). Let φ be a solution of the wave equation (1.1.1), with initial data on Στ0 (τ0 >1) satisfying
D .
= Z ∞
τ0+R∗
dv∗ Z
Sn−1
dµ◦
γn−1 × X2
k=0
X
|α|≤[n2]+1
r4−δ∂(Tk·Ωαχ)
∂v∗
2
+ X5
k=0
X
|α|≤[n2]+1
r2∂TkΩαψ
∂v∗ 2
+ X4
k=0
X
|α|≤[n2]+2
r2∂TkΩαψ
∂v∗
2
u∗=τ0
+ Z
Στ0
X6
k=0
X
|α|≤[n2]+1
JN(TkΩαφ) + X5
k=0
X
|α|≤[n2]+2
JN(TkΩαφ), n
<∞. (1.6.20)
for some0< δ < 14, whereR > n−2q
8nm
δ , n≥3. Then there exists a constantC(n, m, δ, R) such that for n−2√
2m < r0 < r < R,
rn−22 |φ|(t, r)≤ C D
t32−δ . (1.6.21)
Proof. Let ¯t0 = 2(τ0+τ0)+R∗and ¯t1 = ¯t0+2τ0then by (1.6.18), Prop. 1.14 and Prop. 1.11 Z ¯t1
¯t0
φ2(t, r)rn−2dt≤C(n, m, R) Z
Σ2τ0
X1
k=0
X
|α|≤[n2]+1
JT[TkΩαφ], n
; (1.6.22)
hence by Prop. 1.37 there exists t′0 ∈(¯t0,¯t1) such that rn−2φ2(t′0, r)≤ C(n, m, R)D
¯t30 . (1.6.23)
Now set τ0′ = 12(t′0−R∗) and τj′ = 2τj−1′ (j ∈ N), and t′j = 2τj′ +R∗ (j ∈ N); note that t′j+1−t′j = 12(t′j −R∗). Now consider (1.6.16) with t1 = t′j+1, t0 = t′j; since by (1.6.18), together with Prop. 1.11 and Prop. 1.14,
Z t′j+1
t′j
rn−2φ2(t, r) dt≤C(n, m, R) Z
Στ′ j
X1
k=0
X
|α|≤[n2]+1
JT[TkΩαφ], n
, (1.6.24)
and by Prop. 1.32 and Prop. 1.33, Z t′j+1
t′j
rn−2(∂tφ)2(t, r) dt≤
≤C(n, m, R) Z
Στ′
j∩{r∗≤R∗+(τj′)3}
X2
k=1
X
|α|≤[n2]+1
JT[TkΩαφ], n +
Z
Sn−1
dµγ◦
n−1
X
|α|≤[n2]+1
rn−2(Ωα∂tφ)2|(u∗=τj′,v∗=R∗+τj′+(τj′)3)
, (1.6.25)
which decays with the rateτ4−4δ as is shown in the proof of Prop. 1.53 (ii), we obtain rn−2φ2(r, t′j+1)≤
≤rn−2φ2(r, t′j) + C(n, m, R) (t′j)1−2δ
D
(τj′)2 +C(n, m, δ, R)(t′j)1−2δ D (τj′)4−4δ ≤
≤rn−2φ2(r, t′j) + C(n, m, δ, R)D
(t′j)3−2δ . (1.6.26) In fact, by induction onj ∈N using (1.6.23) for j = 0, we have shown
rn−2φ2(r, t′j)≤ C(n, m, δ, R)D
(t′j)3−2δ (j ∈N∪ {0}). (1.6.27) Finally for any t ≥ t′0 we may choose j ∈ N∪ {0} such that t ∈ (t′j, t′j+1) and conclude the proof by applying (1.6.27) and (1.6.26) which holds with t in place oft′j+1.
Extension to the horizon. Note that for n−2√
2m ≤ r < r0 the same interpolation (1.6.16) by integration along lines of constant radiusr < r0 can be carried out. However, on the right hand sides of (1.6.17) and (1.6.18) a new term results from the integration on v∗ = 12(t0 +r0∗) from the radius r < r0 to r = r0; but we infer from the explicit construction (1.3.18) that the resulting integrand
2 1− r2mn−2
∂φ
∂u∗ 2
≤T[φ](Y, Y)≤
JN[φ], N
(1.6.28) is controlled by Cor. 1.13 and the proof of Prop. 1.56 above extends to that of Thm. 2 by replacing JT by JN on the right hand sides of (1.6.22), (1.6.24) and (1.6.25).
Chapter 2
Linear waves on expanding
Schwarzschild de Sitter spacetimes
2.1 Overview
It is the purpose of the work presented in this Chapter to initiate the global study of linear waves on cosmological spacetimes.
A common feature of expanding spacetimes is suggested by the global causal geometry of the simplest explicitly known solutions to the vacuum Einstein equations with positive cosmological constant. The Schwarzschild de Sitter family exhibits a region of spacetime which is bounded in the past by two cosmological horizons and in the future by aspacelike hypersurface of unbounded area.
We provide a suitably robust approach to the analysis of linear waves in these regions.
More precisely, we establish uniform energy estimates for general solutions to the linear wave equation in the expanding regions of de Sitter and Schwarzschild de Sitter spacetimes which extend by a stable redshift mechanism to a global estimate.
Statement of the Main Result. Let Σ+ be the timelike future boundary of a chosen expanding region in a subextremal Schwarzschild de Sitter spacetime (M, g). Σ+ is a spacelike hypersurface with topologyR×S2 endowed with the standard metric g◦ of the cylinder. Let Σ ⊂ J−(Σ+) be a spacelike hypersurface in the past of Σ+ such that Σ+ is in the domain of dependence of Σ and such that Σ crosses the cosmological and event horizons to the future of the bifurcation spheres (see figure 2.1). We consider the Cauchy problem
gψ = 0 (2.1.1)
121
Σ+
Σ C¯+ C+
Figure 2.1: Cauchy problem (2.1.1) with initial data on Σ. The expanding region is bounded in the past by the cosmological horizons ¯C+ ∪ C+ and in the future by the spacelike hypersurface Σ+.
with initial data prescribed on Σ. Let T be the energy momentum tensor of (2.1.1) and n the normal to Σ. We show if the energy of a solution to (2.1.1) is initially finite
D[ψ] .
= Z
Σ
T(n, n)<∞, (2.1.2)
then it is globally bounded in the expanding region and has a limit on Σ+. Furthermore, the limit of ψ on Σ+ as a function on R×S2 satisfies
Z
Σ+
∇◦ ψ 2◦
g dµ◦g ≤C(M,Σ)D[ψ] (2.1.3) whereC is a constant that only depends on the given manifoldsMand Σ, and∇◦ denotes the gradient on the standard cylinder R×S2.
Remark 2.1. The expanding region can be foliated by spacelike hypersurfaces (Σr, gr) which are conformal to the standard cylinder; in fact gr =r2 g◦ +O(1r). Since the future boundary can be viewed as the set
Σ+ =\
r
J+(Σr), (2.1.4)
the statement (2.1.3) is a decay result for the induced derivatives on Σr as they approach Σ+.
Remark 2.2. The wave equation on “asymptotically de Sitter-like spaces” was previously studied by [47] however with results that are local in nature.
The precise statement of the main result and an overview of our proof is given in Sec- tion 2.3. The reader is advised however to first familiarize herself with our treatment of linear waves on de Sitter spacetimes in Section 2.2 where many of the ideas for our approach originate and their application is seen more clearly. Moreover, Section 2.2 offers a self-contained global analysis of linear waves on de Sitter spacetimes including results for the Klein-Gordon equation.