The Stability of the Euler-Einstein System with a Positive Cosmological Constant Jared Speck jspeck@math.princeton.edu University of Cambridge April 7, 2010 Einstein equations (for a spacetime (M, gµν ) of signature (-,+,+,+)) (fluid) Gµν + Λgµν = Tµν Λ>0 µν Bianchi identities =⇒ Dµ T(fluid) =0 Our spacetimes: [0, T ] × T3 Red-shift effect Light emitted by distant objects is red-shifted Hubble-Humason “law" (1929): galaxies are receding from Earth, and their velocities are proportional to their distances from it ; “suggests" an expanding universe . “Expansion of the universe” a(t) is the scale factor g = −dt 2 + a2 (t) 3 X (dx k )2 , k =1 ȧ(t) > 0 Leads to red-shift effect Why Λ > 0? 1990’s: experimental data (type IA supernovae, ...) suggest accelerated expansion. Λ > 0 =⇒ ∃ solutions with accelerated expansion. e.g. P g = −dt 2 + e2Ht 3a=1 (dx a )2 , p H = Λ/3. Stability Background solution (Ours will be FLRW type) Initial value problem formulation; wave coordinates (de Donder 1921, Choquet-Bruhat 1952) Goal: show that if we slightly perturb the background data, then the resulting solution exists for all t ≥ 0 and that the spacetime is future causally geodesically complete We show convergence as t → ∞ Proofs based on “energy" estimates for quasilinear wave equations and the Euler equations Previous stability results For Λ > 0 using the conformal method Vacuum Einstein (Friedrich, 1986) Einstein-Maxwell & Einstein-Yang-Mills (Friedrich, 1991) Vacuum Einstein in (1 + n) dimensions, n odd (Anderson, 2005) For Λ = 0 using wave coordinates Einstein-scalar field with a nonlinearity V (Φ) in (1 + n) dimensions (Ringström, 2008) Φ = V 0 (Φ) V (Φ) emulates Λ > 0 : V (0) > 0, V 0 (0) = 0, V 00 (0) > 0 Previous stability results cont. Newtonian cosmological models with Λ > 0 (Brauer, Rendall, & Reula, 1994) Irrotational Euler-Einstein with Λ > 0 (Rodnianski & Speck, 2009) Indices Latin (spatial) indices: a, b, i, j, k ∈ {1, 2, 3} Greek (spacetime) indices: α, β, κ, µ, ν ∈ {0, 1, 2, 3} def Summation convention: ua u a = u1 u 1 + u2 u 2 + u3 u 3 , etc. ∂t = ∂0 ∂ = (∂t , ∂1 , ∂2 , ∂3 ) ∂¯ = (∂1 , ∂2 , ∂3 ) The relativistic Euler equations µν Dµ T(fluid) = 0, Dµ (nu µ ) = 0, µν T(fluid) = (ρ + p)g µν + pg µν , u 0 > 0. uα u α = −1, To close the system: combined first and second law of thermodynamics: ρ+p =n dρ , dn and an equation of state: p = cs2 ρ. Friedmann-Lemaître-Robertson-Walker background solutions Ansatz: e = −dt 2 + a2 (t) g 3 X (dx a )2 , a(0) = 1, a=1 e=p e(t), p def e(0) = p p̄ > 0. Plug into Einstein equations, and get: s p̄ d a = a H + 2 3(1+c 2 ) , s dt 3cs a r Λ H= , 3 2 e(t) = a−3(1+cs ) p̄. p Cosmetics and Psychology Cosmetics: def Ω(t) = ln a(t) ≈ Ht, ȧ(t) def d ≈ H. ω(t) = Ω(t) = dt a(t) P e = −dt 2 + e2Ω(t) 3a=1 (dx a )2 exponentially grows; g e = p̄e−3(1+cs2 )Ω(t) exponentially decays. p Rescale: def hjk = e−2Ω(t) gjk , def 2 P = e3(1+cs )Ω(t) p. Coordinate gauge Wave coordinates Γµ = e Γµ = 3ω(t)δ0µ ≈ 3Hδ0µ Use to “modify” the Einstein equations (without changing them!) Γµ |t=0 = e Γµ |t=0 =⇒ Γµ ≡ e Γµ Proof: (≈ Choquet-Bruhat) Γµ − e Γµ satisfies a system of wave equations with 0 data The modified Euler-Einstein system def def 2 def def hjk = e−2Ω gjk , P = e3(1+cs )Ω p W = 3cs2 , H 2 = def ˆg = g αβ ∂α ∂β ; “friction", “error", “dangerous" Λ 3 ˆ g (g00 + 1) = 5H∂t g00 + 6H 2 (g00 + 1) + 400 , ˆ g g0j = 3H∂t g0j + 2H 2 g0j − Hg ab Γajb + 40j , ˆ g hjk = 3H∂t hjk + 4jk , −1 Pua ∂t u a + (1 + cs2 )P∂a u a = 4, u ν ∂ν (P − p̄) + (1 + cs2 ) u0 u ν ∂ν u j + cs2 Πjα ∂α (P − p̄) = (W − 2)ωu j + 4j . (1 + cs2 )P Norms (N ≥ 3, 0 < q 1, Ω(t) ≈ Ht, hjk = e−2Ωgjk ) ¯ 00 kH N + eqΩ kg00 + 1kH N , Sg00 +1;N = eqΩ k∂t g00 kH N + e(q−1)Ω k∂g 3 X def ¯ 0j kH N + e(q−1)Ω kg0j kH N , e(q−1)Ω k∂t g0j kH N + e(q−2)Ω k∂g Sg0∗ ;N = def j=1 def Sh∗∗ ;N = 3 X ¯ jk kH N + k∂h ¯ jk kH N−1 , eqΩ k∂t hjk kH N + e(q−1)Ω k∂h j,k =1 v u 3 uX def UN= t e2(1+q)Ω ku j k2H N , j=1 def S N = eΩ ku j kH N + kP − p̄kH N , def QN = Sg00+1 ;N + Sg0∗ ;N + Sh∗∗ ;N + UN−1 + SN . | {z } Measures the deviation from the FLRW solution The global existence theorem Theorem (J.S. 2010) Assume that N ≥ 3 and 0 < cs2 < 1/3. Then there exist 0 > 0 and C > 1 such that for all ≤ 0 , if QN (0) ≤ C −1 , then there exists a global future causal geodesically complete solution to the modified Euler-Einstein system. Furthermore, QN (t) ≤ holds for all t ≥ 0. def QN (t) = supτ ∈[0,t] Q N (τ ). Asymptotics of the solution Theorem (J.S. 2010) Under the assumptions of the global existence theorem, with the additional assumption N ≥ 5, there exist (∞) q > 0, a smooth Riemann metric gjk with corresponding (∞) jk Christoffel symbols Γijk and inverse g(∞) on T3 such that (∞) ke−2Ω gjk (t, ·) − gjk kH N ≤ Ce−qHt , jk ke2Ω g jk (t, ·) − g(∞) kH N ≤ Ce−qHt , (∞) ke−2Ω ∂t gjk (t, ·) − 2ωgjk kH N ≤ Ce−qHt , (∞) ab kg0j (t, ·) − H −1 g(∞) Γajb kH N−3 ≤ Ce−qHt , k∂t g0j (t, ·)kH N−3 ≤ Ce−qHt , Asymptotics cont. Theorem (Continued) kg00 + 1kH N ≤ Ce−qHt , k∂t g00 kH N ≤ Ce−qHt , (∞) ke−2Ω Kjk (t, ·) − ωgjk kH N ≤ Ce−qHt . In the above inequality, Kjk is the second fundamental form of the hypersurface {t = const}. Similar results hold for the fluid variables. Fundamental idea of proof Proposition The norm QN (t) has the following property: if the solution blows up at a time Tmax , then limt↑Tmax QN (t) = ∞. To prove global existence: we show that QN (t) remains uniformly bounded for all t ≥ 0. The energies Basic idea of analysis: analyze “energies" that are ≈ to the norms. Energies are defined using the vectorfield method: vectorfield z }| { e ν [v , v ]X µ , J ν [v , v ] = T µ ∂ν (J ν [v , v ]) = lower order. e ν [v , v ] = canonical stress corresponding to v ; T µ encodes “almost-"conservation laws e ν [v , v ]) = Qµ [v , v ]; (have to use the equations) ∂ν (T µ X = suitable timelike vectorfield R 0 J [v , v ]d 3 x is the square of the “energy" of v T3 Time derivative of energy is controlled by the divergence theorem Two sources of stress Energy-norm equivalence Lemma If QN is small enough, then the energies and norms are equivalent: Eg00 +1;N Eg0∗ ;N Eh∗∗ ;N EN + UN−1 QN ≈ Sg00 +1;N , ≈ Sg0∗ ;N , ≈ Sh∗∗ ;N , ≈ SN , ≈ QN . Gronwall estimates for energies Lemma The following system of inequalities (for t ≥ t1 ) holds for the modified equations. Furthermore, it features small-data global bounds. U 2N−1 (t)≤ U 2N−1 (t1 ) <0 t Z }| { z 2(W − 1 + q) e(1+q)Hτ U 2N−1 + CQN U N−1 dτ , + E 2N (t)≤ E 2N (t1 ) + C Z τ =t1 t e−qHτ Q2N dτ , t1 E2g00 +1;N (t) E2g0∗ ;N (t) E2h∗∗ ;N (t) ≤ ≤ ≤ Z E2g00 +1;N (t1 ) + E2g0∗ ;N (t1 ) Z E2h∗∗ ;N (t1 ) t t1 −4qHE2g00 +1;N + Ce−qHτ QN Eg00 +1;N dτ, t −4qHE2g0∗ ;N + CEh∗∗ ;N Eg0∗ ;N + Ce−qHτ QN Eg0∗ ;N dτ, + t1 Z t + t1 He−qHτ E2h∗∗ ;N + Ce−qHτ QN Eh∗∗ ;N dτ. The fluid energy currents def Given any variations (Ṗ, u̇ 1 , u˙2 , u˙3 ), we set u̇ 0 = − u10 ua u̇ a and define the fluid energy current: def J̇ µ = uµ (1 + cs2 )Pu µ 2 µ Ṗ + 2 u̇ Ṗ + gαβ u̇ α u̇ β (1 + cs2 )P cs2 J̇ 0 = “density of energy" J̇ 0 ≈ Ṗ 2 + e2Ω δab u̇ a u̇ b The divergence of J̇ h i uµ (1 + cs2 )Pu µ i 2 Ṗ + ∂µ gαβ u̇ α u̇ β (1 + cs2 )P cs2 (1 + cs2 )Pu a (1 + cs2 )Pu 0 α β + (∂ g ) u̇ u̇ + (∂t g00 )(u̇ 0 )2 a αβ cs2 cs2 2(1 + cs2 )Pu 0 2(1 + cs2 )P(u 0 − 1) 0 α + (∂ g ) u̇ u̇ + (∂t gab )u̇ a u̇ b t 0α cs2 cs2 h i ua (1 + cs2 )P − 2 ∂t u̇ a Ṗ + (∂t gab − 2ωgab )u̇ a u̇ b u0 cs2 2(1 + cs2 )P 2F 2(1 + cs2 )P + (W − 1)ωgab u̇ a u̇ b + Ṗ + gαβ Gα u̇ β 2 2 cs (1 + cs )P cs2 | {z } h ∂µ J̇ = ∂µ µ <0 Control of k∂µJ̇ µkL1 Using the formula for ∂µ (J̇ µ ) plus bootstrap assumption QN (t) ≤ , we can show: k∂µ J̇ µ kL1 ≤ Ce−qHt QN 2 Simple consequence: d dt Z 0 3 Z J̇ d x = T3 T3 ∂µ J̇ µ d 3 x ≤ Ce−qHt QN 2 . Comparison with flat spacetime Christodoulou’s monograph “The Formation of Shocks in 3-Dimensional Fluids" shows that on the Minkowski spacetime background, shock singularities can form in solutions to the Euler equations arising from data that are arbitrarily close to that of a uniform quiet state. Conclusion: Exponentially expanding spacetimes can stabilize fluids. Topologies beyond T3 The study of wave equations arising from metrics featuring accelerated expansion is a “very local" problem. A patching argument can “very likely" be used to allow for many of the topologies considered by Ringström: Unimodular Lie Groups different from SU2; H3 ; H2 × R; · · · Future directions p = 0 : preliminary analysis suggests that it is stable The case p = cs2 ρ with cs2 ≥ 1/3 is unclear; numerical evidence could be helpful Other equations of state Sub-exponential expansion rates Thank you gµν building block energies If α > 0, β ≥ 0, and φ is a solution to ˆ g φ = αH∂t φ + βH 2 φ + F , then we control solutions to this equation using an energy Z def 1 2 E(γ,δ) [φ, ∂φ] = {−g 00 (∂t φ)2 + g ab (∂a φ)(∂b φ) − 2γHg 00 φ∂t φ + δH 2 φ2 } d 3 x 2 T3 γ, δ can be adjusted so that ∃η > 0 such that Z n o d 2 2 (E(γ,δ) [φ, ∂φ]) ≤ −ηHE(γ,δ) [φ, ∂φ] + − (∂t φ + γHφ)F + 4E;(γ,δ) [φ, ∂φ] d 3 x, dt T3 4E;(γ,δ) [φ, ∂φ] = −γH(∂a g ab )φ∂b φ − 2γH(∂a g 0a )φ∂t φ − 2γHg 0a (∂a φ)(∂t φ) 1 − (∂a g 0a )(∂t φ)2 − (∂a g ab )(∂b φ)(∂t φ) − (∂t g 00 )(∂t φ)2 2 1 ab ab + ∂t g + Hg (∂a φ)(∂b φ) − γH(∂t g 00 )φ∂t φ − γH(g 00 + 1)(∂t φ)2 . 2 Energies for gµν def E2g00 +1;N = X 2 e2qΩ E(γ [∂α~ (g00 + 1), ∂(∂α~ (g00 + 1))] 00 ,δ00 ) |~ α|≤N E2g0∗ ;N def = 3 X X 2 e2(q−1)Ω E(γ [∂α~ g0j , ∂(∂α~ g0j )] 0∗ ,δ0∗ ) |~ α|≤N j=1 E2h∗∗ ;N def = 3 X X |~ α|≤N j,k =1 2 [0, ∂(∂α~ hjk )] e2qΩ E(0,0) 1 + 2 Z 2 3 cα~ H (∂α~ hjk ) d x T3