Gerhard Rein. Mathematisches Institut der Universitat Munchen, Alan D. Rendall. Max-Planck-Institut fur Astrophysik, and

Size: px

Start display at page:

Download "Gerhard Rein. Mathematisches Institut der Universitat Munchen, Alan D. Rendall. Max-Planck-Institut fur Astrophysik, and"

Godfrey Stone
6 years ago
Views:

1 A regularity theorem for solutions of the spherically symmetric Vlasov-Einstein system Gerhard Rein Mathematisches Institut der Universitat Munchen, Theresienstr. 39, Munchen, Germany, Alan D. Rendall Max-Planck-Institut fur Astrophysik, Karl-Schwarzschild-Str. 1, Garching, Germany, and Jack Schaeer Department of Mathematics, Carnegie-Mellon University, Pittsburgh, PA 1513, USA Abstract We show that if a solution of the spherically symmetric Vlasov- Einstein system develops a singularity at all then the rst singularity has to appear at the center of symmetry. The main tool is an estimate which shows that a solution is global if all the matter remains away from the center of symmetry. 1 Introduction This paper is concerned with the long-time behaviour of solutions of the spherically symmetric Vlasov-Einstein system. In [4] a continuation criterion was obtained for solutions of this system, and it was shown that for small initial data the corresponding solution exists globally in time. In the following we investigate what happens for large initial data. In the coordinates used in [4] the equations to be solved are as follows: v t f +e ; xf ; 1+v r _ +e ;p 1+v 0 x vf =0 (1:1) e ; (r 0 ;1)+1=8r (1:) 1

2 e ; (r 0 +1);1=8r p (1:3) (t x)=z p1+v f(t x v)dv (1:4) p(t x)= Z xv r dv f(t x v) p : (1:5) 1+v Here x and v belong to IR 3, r := jxj, xv denotes the usual inner product of vectors in IR 3,andv := v v. The distribution function f is assumed to be invariant under simultaneous rotations of x and v, and hence and p can be regarded as functions of t and r. Spherically symmetric functions of t and x will be identied with functions of t and r whenever it is convenient. In particular and are regarded as functions of t and r, and the dot and prime denote derivatives with respect to t and r respectively. It is assumed that f(t) has compact support for each xed t. We are interested in regular asymptotically at solutions which leads to the boundary conditions that (t 0) = 0 lim (t r)=0 (1:6) r!1 for each xed t. The main aim of this paper is to show that if singularities ever develop in solutions of the system (1.1){(1.5) with the above boundary conditions then the rst singularity must be at the center of symmetry. In order to do this we consider solutions of (1.1){(1.5) on a certain kind of exterior region with dierent boundary conditions and prove that for the modied problem there exists a global in time solution for any initial data. One of the most interesting implications of the results of [4] is that the naked singularities in spherically symmetric solutions of the Einstein equations coupled to dust can be cured by passing to a slightly dierent matter model, namely that described by the Vlasov equation, in the case of small initial data. The results of this paper strengthen this conclusion to say that shell-crossing singularities are completely eliminated. The following remarks put our results in context. For the Vlasov-Poisson system, which is the non-relativistic analogue of the Vlasov-Einstein system, it is known that global existence holds for boundary conditions which are the analogue of the requirement of asymptotic atness in the relativistic case [, 3, 6] and also in a cosmological setting [5]. No symmetry assumptions are necessary. For the relativistic Vlasov-Poisson system with an attractive force spherically symmetric solutions with negative energy develop singularities in nite time [1]. It is easy to show that in these solutions the rst singularity

3 occurs at the center of symmetry. On the other hand it was also shown in [1] that spherically symmetric solutions of the relativistic Vlasov-Poisson system with a repulsive force never develop singularities. The latter are in one-to-one correspondence with spherically symmetric solutions of the Vlasov-Maxwell system. The paper is organized as follows. Sect. contains the main estimates together with a proof that they imply global existence in the case that all the matter remains away from the center. In Sect. 3 a local existence theorem and continuation criterion for the exterior problem are proved. It is then shown that the estimates of Sect. imply a global existence theorem for the exterior problem. Finally, in Sect. 4, these results are combined to give the main theorem. The restricted regularity theorem The goal of this section is to show that a solution may be extended as long as f vanishes in a neighborhood of the center of symmetry. Consider an initial datum f 0 which is spherically symmetric, C 1, compactly supported, and satises Z Z p1+v f(x v)dvdx < r= (:1) jxj<r for all r>0. By Theorem 3.1 of [4] a regular solution (f ) of the system (1.1){(1.5) with boundary conditions (1.6) and initial datum f exists on [0 T[IR 6 for some T>0. Theorem.1 Let f and T be asabove (T nite). Assume there exists >0 such that f(t x v)=0 if 0 t<t and jxj: (:) Then (f ) extends to a regular solution on [0 T 0 [ for some T 0 >T: Dene n P (t):=sup jvj :(x v) suppf(t) o (:3) then Theorem 3. of [4] states that Theorem.1 above follows once P (t) is shown to be bounded on [0 T[. We need a few other facts from [4]: e ; =1;m=r (:4) 3

4 where m(t r):=4 Z r (equations (.11) and (.1) of [4]). Also where (equations (3.37) and (3.38) of [4]). The following notation will be used: 0 s (t s)ds (:5) _ = ;4re + j (:6) Z xv j(t r):= f(t x v)dv (:7) r u := jvj w:= r ;1 xv F := jx^vj = r u ;(xv) : Dierentiation along a characteristic of the Vlasov equation is denoted by D t so v D t x = e ; p 1+u and It follows that xv D t v = ; +e r _ ;p 1+u 0 x r : D t F =0 (:8) w D t r = e ; p 1+u (:9) and D t w = ;r ; (D t r)xv +r ;1 (D t x)v +r ;1 x(d t v): Substitution for D t r D t x and D t v and simplication yields D t w = F r 3p 1+u e; ;w _ ;e ;p 1+u 0 : (:10) The letter C will denote a generic constant which changes from line to line, and may depend only on f, andt. Now we make a few preliminary estimates. The values of f are conserved along characteristics so 0 f sup f = C: 4

5 Also we claim that Z Z (t x)dx = (0 x)dx = C (:11) To show this multiply (1.1) by p 1+v and integrate in v which yields (after simplication) 0=@ t +e ; div Z t +div e ; Z fvdv +(+p) +je _ ; 0 fvdv +(+p) +je _ ; ( ): Now substituting (.6) and (1.), (1.3) this becomes 0=@ t +div Z e ; fvdv (:1) and (.11) follows. Also by (.5) Z 0 m(t r) (t x)dx C r 0: (:13) It follows from (1.), (1.3) that and from (1.6) and (.4) that so and Note that on the support of f, so lim r!1 (+)=0 ; + 0 e ; e + 1: (:14) 0 F C u = w + F r w + C = w +C: (:15) 5

6 Hence we will focus on w. Dene P i (t):=inf n o w : 9x v with f(t x v) 6= 0 and w = r ;1 xv (:16) and P s (t):=sup n o w : 9x v with f(t x v) 6=0 and w = r ;1 xv (:17) then if P i (t) and P s (t) are bounded, it follows that P (t) is bounded. Also note that n measure v :(x v) suppf(t) o C ; P s (t);p i (t) jxj: (:18) Next we focus on the characteristic equation for w. Note rst that by (1.3) and (.4) 0 = 1 e (8r p+1);1 = 1 r r e 8r p+1;1+ m r = e (r ; m+4rp): Using this and (.6) in (.10) yields D t w = F r 3p 1+u e; ;r ;p 1+u e + m+4re + (wj ; p 1+u p) By (.) and (.14) we may bound the rst term of (.19) by 0 F r 3p 1+u e; ;3 C = C (:19) on the support of f. By (.), (.15), (.14), and (.13) we may bound the second term of (.19) by 0 r ;p 1+u e + m ;p C +w C C p C +w on the support of f. Hence, (.19) becomes ;C p C +w +4re + (wj ; p 1+u p) D t w (.0) C +4re + (wj; p 1+u p): On the support of f 0 4re + C 6

7 so we must consider the quantity wj ; p 1+u p. Let us denote ~w := r ;1 x ~v ~u := j~vj ~ F := jx^ ~vj = r ~u ;(x ~v) : Then wj ; p Z 1+u p = f(t x ~v) w ~w ; p ~w 1+u p d~v: (:1) The next step is to use (.0) and (.1) to derive an a upper bound for w (on the support of f), and hence for P s (t). This may be done without a bound on P i (t): Then (.0), (.1), and the bound for P s (t) will be used to derive alower bound for w, and hence for P i (t): Then by (.15) and (.3) a bound for P (t) follows, and the solution may be extended. To bound P s (t) suppose P s (t) > 0 and consider w (in suppf) with w>0: For ~w 0wehave w ~w; p 1+u ~w p 0: For ~w>0wehave w ~w; p ~w 1+u p = ~w w (1+ ~u ); ~w (1+u ) p w p +~w p 1+u = ~w p w (1+ ~ Fr ; ); ~w (1+Fr ; ) w p +~w p 1+u : Note that in the last step a term of \w ~w " canceled, which is crucial. Hence w ~w ; p ~w 1+u p ~w w (1+C ; ) p w p C w ~w 1+ ~w : Now using the above and (.18) we have Z f w ~w ; p ~w Z 1+u p d~v fcw ~w 0< ~w<p s(t) 1+ ~w d~v Z Ps(t) C ; Cw ~w 0 1+ ~w d ~w = Cwln(1+P (t)) CP s s(t)ln(1+p (t)) s 7

8 and by (.1) and (.0) D t w C +CP s (t)ln(1+p s (t)) (:) for w>0 in the support of f. Denote the values of w along a characteristic by w() and let t 0 := inf n 0:w(s) 0fors ] t[ o : Then either t 0 =0 or w(t 0 ) = 0 and in either case Hence by (.) Dening we may write and hence w(t) w(t 0 )+ C +C Z t t 0 Z t w(t) C +C P s (t) C +C w(t 0 ) C: C +CP s ()ln(1+p s ()) d t 0 P s ()ln(1+p s ())d: P s () := maxf0 P s ()g Z t 0 Z t 0 P s ()ln(1+p s())d P s ()ln(1+p s())d: We assumed that P s (t) > 0 but note that the last inequality isvalid in all cases. It now follows that on t [0 T[: To bound P i (t) from below suppose and consider w (in support f) with P s (t) P s (t) exp(e ct ) C (:3) P i (t) < 0 P i (t) <w<0: 8

9 For ~w 0wehave w ~w ; p ~w 1+u p = ~w w p (1+ Fr ~ ; ); ~w (1+Fr ; ) w p +~w p 1+u = j ~wj p w (1+ ~ Fr ; ); ~w (1+Fr ; ) jwj p +j ~wj p 1+u p j ~wj (; ~w )(1+Fr ; ) j ~wj p 1+u ; ~w (1+C ; ) p p 1+ ~w 1+w ;Cp j ~wj : 1+w For w<0 < ~w P s (t) wehave (using (.3)) w ~w; p 1+u ~w p P s(t)w ; p 1+w +Fr ; ~w p 1+ ~w Hence (using (.18) as before) Z Cw;j~wj p 1+C ; +w Cw;P s (t) p C +w ;C p C +w : f w ~w; p ~w 1+u p d~v Z f ;C j ~wj Z p d~v + f ;C p C +w d~v ~w0 1+w ~w>0 ; C p 1+w C; Z 0 ;P i (t) 1 = ;CP i (t) p ;CP s(t) p C +w : 1+w Now by (.1), (.0), and (.3) j ~wjd ~w;c p Z P s(t) C +w C ; d ~w D t w ;C p 1 C +w ;CP i (t) p ;CP s(t) p C +w 1+w ;C p C +w ;CP i (t) 1 p 1+w : 0 9

10 Since we have assumed 0 >w>p i (t) it is convenient to write this as D t (w )=wd t w C(;w) p C +w +CP i (t) p (;w) 1+w q CjP i (t)j C +P i (t)+cp(t) i C +CP i (t): As before dene then so t 1 := inf n 0:w(s) 0fors ] t[ o 0 w(t 1 ) ;C It follows that Z t w (t) C + (C +CP i ())d t 1 C +C Z t 0 P i (t) C +C if P i (t) < 0: But if P i (t) 0 then P i ()d: Z t 0 P i (t) P s (t) C 0 P i ()d (:4) so (.4) holds in this case, too. Now by Gronwall's inequality it follows that P i (t) ect C on [0 T[: Finally a bound for P (t) follows from (.15) and (.3), and the proof is complete. 3 The exterior problem If r 1 and T are positive real numbers, dene the exterior region W (T r 1 ):=f(t r):0 t<t r r 1 +tg: 10

11 In this section the initial value problem for (1.1){(1.5) will be studied on a region of this kind. Consider an initial datum f(x v) dened on the region jxjr 1 which is non-negative, compactly supported, C 1, and spherically symmetric. The rst of the boundary conditions (1.6) cannot be used in the case of an exterior region, and so it will be replaced as follows. Let m 1 be any number greater than 4 R 1 r 1 r (0 r)dr. For any solution of (1.1){(1.5) on W (T r 1 ) dene m(t r):=m 1 ;4 Z 1 r s (t s)ds: (3:1) Provided m 1 satises the above inequality, the quantity m(0 r)iseverywhere positive. The replacement for (1.6) is: e ;(t r) =1;m(t r)=r lim (t r)=0: (3:) r!1 The rst of these conditions is a combination of (1.) with a choice of boundary condition. Note that if a solution of the original problem with boundary conditions (1.6) is restricted to W (T r 1 ) then it will satisfy (3.) provided m 1 is chosen to be equal to the ADM mass of the solution on the full space. Just as in the local existence theorem in [4], a further restriction must be imposed on the initial datum. In this case it reads Z Z p1+v m 1 ; f(x v)dvdx < r= r r 1 : (3:3) jxjr This is of course necessary if (3.) is to hold on the initial hypersurface. The nature of the solutions to be constructed is encoded in the following denition. Denition A solution (f ) of (1.1){(1.5) on a region of the form W (T r 1 ) is called regular if (i) f is non-negative, spherically symmetric, and C 1,andf(t) has compact support for each t [0 T[ (ii) 0, and,, 0,and 0 are C 1. A local existence theorem can now be stated. Theorem 3.1 Let m 1 > 0 be a xed real number. Let f 0 be a spherically symmetric function on the region jxjr 1 which is C 1 and has compact support. Suppose that (3.3) holds for all r r 1 and that Z Z p1+v f(x v)dvdx < m 1 : (3:4) jxjr 1 11

12 Then there exists a unique regular spherically symmetric solution of (1.1){ (1.5) on a region W (T r 1 ) with f(0) = f and satisfying (3.). Proof : This is similar in outline to the proof of Theorem 3.1 of [4] and thus will only be sketched, with the dierences compared to that proof being treated in more detail. Dene 0 (t r) and 0 (t r) to be zero. If n and n are dened on the region W (T n r 1 ) then f n is dened to be the solution of the Vlasov equation with and replaced by n and n respectively and initial datum f. In order that this solution be uniquely dened it is necessary to know thatnocharacteristic can enter a region of the form W (T r 1 ) except through the initial hypersurface which is guaranteed if n 0 and n 0 on the region of interest. This will be proved by induction. If f n is given then n+1 and n+1 are dened to be the solutions of the eld equations with n and p n constructed from f n rather than f. The quantities n+1 and n+1 are dened on the maximal region W (T n+1 r 1 ) where 0 <m n (t r) <r= so that n+1 can be dened by the rst equation in (3.) and is positive onw (T n+1 r 1 ) note that 0 <m n (t r) <r=forr large and 0 <m n (0 r) <r= r r 1 by assumption on f so that T n+1 > 0by continuity. It can be shown straightforwardly that the iterates ( n n f n ) are well-dened and regular for all n. The most signicant diculty in proving the corresponding statement in[4]was checking the dierentiability of various quantities at the center of symmetry, and in the exterior problem the center of symmetry is excluded. The next step is to show that there exists some T>0 so that T n T for all n and that the quantities n, n _, and 0 n are uniformly bounded in n on the region W (T r 1 ). Let L n (t r):= r ;1 (1;e ;n(t r) ) ;1.For t [0 T n [ dene P n (t) :=supfjvj :(x v) suppf n (t)g Q n (t) :=ke n(t) k 1 +kl n (t)k 1 : Now it is possible to carry out the same sequence of estimates as in [4] to get a dierential inequality for ~ P n (t):=max 0kn P k (t) and ~ Q n (t):= max 0kn Q k (t) which is independent ofn, and this gives the desired result. The one point which is signicantly dierent fromwhatwas done in [4] is the estimate for _ n. There a partial integration in r must be carried out, and in the exterior problem the limits of integration are changed nevertheless the basic idea goes through. The remainder of the proof is almost identical to the proof of Theorem 3.1 of [4]. It is possible to bound 0 n and _ 0 n on the region W (T r 1 ) and then show that the sequence ( n n f n )converges 1

13 uniformly to a regular solution of (1.1){(1.5) on that region. Moreover this solution is unique. Next a continuation criterion will be derived. By the maximal interval of existence for the exterior problem is meant the largest region W (T r 1 )on which a solution exists with given initial data and the parameters r 1 and m 1 xed. Theorem 3. Let (f ) be aregular solution of the reduced system (1.1){ (1.5) on W (T r 1 ) with compactly supported initial datum f. If T<1 and W (T r 1 ) is the maximal interval of existence thenp is unbounded. Proof : Note rst that, just as in the case of the problem in the whole space, the reduced equations (1.1){(1.5) imply that the additional eld equation (.6) is satised. The conservation law (.1) can be rewritten in the t (r )+@ r (e ; r j)=0: (3:5) Integrating (3.5) in space shows that the minimum value of m at time t (which occurs at the inner boundary of W (T r 1 )) is not less than at t =0. Hence L is bounded on the whole region. Suppose now that P is bounded. From (.6) it follows that _ is bounded, and if T is nite this gives an upper bound for. Combining this with the lower bound for already obtained shows that Q is bounded. This means that the quantities which inuence the size of the interval of existence are all bounded on W (T r 1 ), and it follows that the solution is extendible. Combining Theorem 3. with the estimates of Sect. gives a proof that the solution of the exterior problem corresponding to an initial datum of the kind assumed in Theorem 3.1 exists globally in time i. e. that T can be chosen to be innity in the conclusion of Theorem 3.1. For we know that it suces to bound P, and the estimates bound the momentum v along any characteristic on which r is bounded below. Moreover the inequality r r 1 holds on W (1 r 1 ). 4 The regularity theorem This section is concerned with the initial value problem for (1.1){(1.5) on the whole space with boundary conditions (1.6). 13

14 Theorem 4.1 Let (f ) be aregular solution of the reduced system (1.1){ (1.5) on a time interval [0 T[. Suppose that there exists an open neighborhood U of the point (T 0) such that supfjvj :(t x v) suppf \(U IR 3 )g < 1: (4:1) Then (f ) extends to a regular solution on [0 T 0 [ for some T 0 >T. Proof : Suppose that the condition (4.1) is satised. Since the equations are invariant under time translations it can be assumed without loss of generality that T is as small as desired. Choosing T suciently small ensures that U contains all points with (T ;t) +r < 4T and 0 t<t. Now let r 1 <T. Then [0 T[IR 3 U [W (T r 1 ). Let f be the restriction of f to the hypersurface t = 0. Restricting f to the region jxjr 1 gives an initial datum for the exterior problem on W (T r 1 ). Let m 1 be the ADM mass of f. There are now two cases to be considered, according to whether f vanishes in a neighborhood of the point (T 0) or not. If it does then by doing a time translation if necessary it can be arranged that the matter stays away from the center on the whole interval [0 T[ so that the results of Sect. are applicable. It can be concluded that the solution extends to a larger time interval in this case. If f does not vanish in a neighborhood of (T 0) then by doing a time translation it can be arranged that (3.4) holds. In this case the results of the previous section show that there exists a global solution on W (1 r 1 ) satisfying (3.). The extended solution must agree on W (T r 1 ) with the solution we started with. Thus a nite upper bound is obtained for jvj on the part of the support of f over W (T r 1 ). But by assumption (4.1) we already have a bound of this type on the remainder of the support of f. Hence supfjvj :(t x v) suppfg < 1: Applying Theorem 3. of [4] now shows that the solution is extendible to a larger time interval in this case too. There is another way of looking at this result. Suppose that (f ) is a regular solution of (1.1){(1.5) on the whole space and that [0 T[ is its maximal interval of existence. Then by a similar argument to the above the solution extends in a C 1 manner to the set ([0 T]IR 3 )nf(t 0)g. Thus if a solution of (1.1){(1.5) develops a singularity at all the rst singularity must be at the center. 14

15 References [1] Glassey, R. T., Schaeer, J.: On symmetric solutions of the relativistic Vlasov-Poisson system. Commun. Math. Phys. 101, 459{473 (1985) [] Lions, P.-L., Perthame, B.: Propagation of moments and regularity for the three dimensional Vlasov-Poisson system. Invent. Math. 105, 415{430 (1991) [3] Pfaelmoser, K.: Global classical solutions of the Vlasov-Poisson system in three dimensions for general initial data. J. Di. Eq. 95, 81{303 (199) [4] Rein, G., Rendall, A. D.: Global existence of solutions of the spherically symmetric Vlasov-Einstein system with small initial data. Commun. Math. Phys. 150, 561{583 (199) [5] Rein, G., Rendall, A. D.: Global existence of classical solutions to the Vlasov-Poisson system in a three dimensional, cosmological setting. Arch. Rational Mech. Anal. 16, 183{01 (1994) [6] Schaeer, J.: Global existence of smooth solutions of the Vlasov-Poisson system in three dimensions. Commun. Partial Di. Eq. 16, 1313{1336 (1991) 15

LARGE TIME BEHAVIOR OF THE RELATIVISTIC VLASOV MAXWELL SYSTEM IN LOW SPACE DIMENSION

Differential Integral Equations Volume 3, Numbers 1- (1), 61 77 LARGE TIME BEHAVIOR OF THE RELATIVISTIC VLASOV MAXWELL SYSTEM IN LOW SPACE DIMENSION Robert Glassey Department of Mathematics, Indiana University