## IvJ xu2x tdx 2ilm J dxuux0dx

and therefore if Im J dxuu = 0 then

(v) J r(x,t)dx = J T(x,0)dx, (L1 ,H~X/2), with r the torsion of the X-curve in the binormal flow.

In the above list we include whithin brackets the Sobolev Hs and Lebesgue Lp spaces which are invariant under the same scaling of the corresponding quantity. Because of (2) we are mainly interested in L1 — H1/2 scalings which are the ones of the (iv) and (v) quantities.

4. Symmetries

Among the transformations which leave invariant the set of solutions of (1) we have the following ones.

(ii) Dilations: u\ = Xu(Xx, X2t) is a solution if so is u\

In terms of the initial condition we have u\(x,0) — Xuo(Xx).

Therefore the solutions which are invariant with respect to scaling should be related to initial data which are homogeneous of degree — 1:

(iii) Multiplication by a constant of modulus one;

(iv) Galilean transformations: given N e R then uN(x, t) - e~itN2+iNxu(x - 2Nt, t) (7)

is a solution if u is a solution. The corresponding initial conditions are related by the identity uN(x,0) = eiNxu(x,0). (8)

Therefore the solutions which are invariant with respect to the Galilean transformations should verify that elNxu(x, 0) = u(x, 0) for all N. Hence u(x, 0) = aS. In fact if

UN = u for all N we can differentiate in N in both sides to conclude that ir]u(rj,t) — 2tdnu(T],t) = 0.

And from (1) and some simple manipulations [11] we get a ( |x|2\

1 .jaji

But —41 converges as t goes to zero to a constant times the Vi i-function. As a consequence ua has no limit at t = 0 and the IVP:

either has no limit or has more than one solution [11]. 5. Stability

Recall that the equation coming from the binormal flow (5) has an extra term A(t) with respect to equation (9). In fact u(s,t) = a^=eia2/4t, (10)

This suggests that the natural problem related to the ¿-function as initial datum has to be modified as in (10). In fact with S. Gutierrez [6] and G. Perelman [12] we have proved that (9) has no solution within the set of self-similar solutions according to scaling.

At the same time a natural question is if the explicit solution (10) of (11) is stable in any sense. The problem suggests different possibilities that we shall try to explain.

The first issue we want to consider is which is the minimal regularity needed to solve the problem f tut + d2xu ± \u\2u = 0

The question is solved if uo is in the Sobolev space Hs, s > 0 [13] [14] [15]. The way of doing it is to find by a fixed point argument a solution to the corresponding integral equation f4

Hence one has to look for good estimates for the solutions of the linear problem

with v the solution of (14). The smallness condition necessary to obtain a fixed point is obtained by taking the time of existence T small enough. But T = Tdluolli2)» and from the L2-conservation law the local result is extended to all t > 0.

It is easy to see using the Galilean transformations [11], that H3 with s < 0 is not very well adapted to solve (12) (TP, s < 0, is not invariant under translations of Fourier space cf. (8)).

Instead of using Sobolev spaces in a joint work with A. Vargas [16] we proposed to use the space

This space is not easy to handle. A sufficient condition to belong to X is for example to assume that Uq G C1 and dj - tt\

with a > 1/6. Then we prove [16] that if uq G X then (12) is locally well posed in time. Notice that if uq = aS then uq verifies (17) with a = 0.

We also obtain a global result following some previous ideas developed by J. Bourgain [17]. For this result we need to be able to decompose uo as u0 = fN + gN-, fN e L2, gNeX,

for all TV > 1. We find [16] such a decomposition by imposing some extra properties. I think that to find good ways of achieving (18) it is a very interesting question in Fourier Analysis.

Another approach to "decrease" the regularity is to look at tut +d2± M7u = 0, (19)

and to look for selfsimilar solutions with respect to the corresponding scaling. The question is then which is the minimal 7 so that those solutions can be constructed. In a joint work with T. Cazenave and M. C. Vilela we improve previous results obtained by T. Cazenave and F. Weissler [18] and later extended by F. Planchon [19] to the classical Hs, s > 0. We extend [20] Planchon's result to s = 0 in dimensions one and two, and define some alternative spaces which behave as Hs, s < 0, in all dimensions. In n = 1

g the lowest 7 obtained by this method is 7 = -, still far from the cubic o equation 7 = 2.

The spaces introduced in [20] are quite simple:

Interestingly A. Griinrock [21] has seen that this type of spaces are quite well adapted to deal with the cubic (and others) non-linearity. He has announced the result of local wellposedness of (12) with u0£Yp p < 00.

Therefore he is "just" missing the initial data we are interested in. Namely uq = aS + bp.v. —.

Recall that as we said before for this type of initial condition the cubic non-linearity should be modified by an extra term as in (11).

Finally let us say a few words of some work in progress done with V. Banica about the stability of the solution

The question is to solve (20) backwards up to time £ = 0. In order to do that we use the conformal transformation u=Vteix2^v(j,11), (21)

and look for solutions of the problem

1 u(s,l) = ui(s), with Vi (s) = a + ei(s). This amounts to solve a2 2a 11 9

From that we obtain

And therefore

On the other hand given eo = eo(£o) £ L2 we define the operator S(F) = e^-^-eo -i f e^'^l [2a2Re F + 2aFRe F + \F\2 F] —

Jto 2 r

with sup ||F(£)||L2 + ||F||L6 < +00, and I = [t0,t0 + h]. Then given T > 0

tei ixR

for any 1 < £0 < T we can take h < c(a)---j

and use (15) in the classical way to obtain a fixed point of (24), and therefore a solution of (23) for 1 < £ < T and all T.

Going back to (22) we easily obtain the conservation law:

J \vs(t)\2ds - I J(M2 - a2)2ds -\J'J(M2 - a2)2ds J = j[\vs(l)\2-(\v(l)\2-a2)2}ds.

It is also easy to prove that if us(l) e L2 then va(t) € L2. But from

Gagliardo-Niremberg's inequality

This can be understood as an orbital stability result of the the solution v = a, a < 1/2, because just ||<9s^||l2 remains small.

Therefore the main remaining question is to identify the limit lim Vteix^4te('-, ]).

References

1. H. Brezis, A. Friedman, Nonlinear parabolic equations involving measures as initial conditions, J. Math. Pures Appl. 62 (1983) 73-97.

2. J. Aguirre, Self-similarity and the singular Cauchy problem for the heat equation with cubic absorption, Appl. Math. Letters 14 (2001) 7-12.

3. G.B. Whitham, Linear and nonlinear waves, Pure and Applied Mathematics, A Wiley Interscience Publication 62 (1974).

4. H. Hasimoto, A soliton in a vortex filament, J. Fluid Mech. 51 (1972) 477485.

5. S. Gutierrrez, J. Rivas, L. Vega, Formation of singularities and self-similar vortex motion under the localized induction approximation, Comm. Part. Diff. Eqns. 28 (2003) 927-968.

6. S. Gutierrrez, L. Vega, Self-similar solutions of the localized induction approximation: singularity formation, Nonlinearity 17 (2004) 2091-2136.

7. L.S. Da Rios, On the motion of an unbounded fluid with a vortex filament of any shape, Rend. Circ. Mat. Palermo 22 (1906) 117.

8. A. Majda, A Bertozzi, Vorticity and incompressible flow. Cambridge texts in Appl. Math., Cambridge University Press (2002)

9. R. Klein, A. Majda, K. Damodaran, Simplified equations for the interaction of nearly parallel vortex filaments, J. Fluid Mech. 288 (1995) 201-248.

10. C. Kenig, G. Ponce, L. Vega, On the interaction of nearly parallel vortex filaments, Comm. Math. Phys. 243 3 (2003) 471-483.

11. C. Kenig, G. Ponce, L. Vega, On the ill-posedness of some canonical nonlinear dispersive equations, Duke Math. J. 106 3 (2001) 716-633.

12. G. Perelman, L. Vega, Selfsimilar solutions for modified KdV equation. Some applications. Preprint.

13. J. Ginibre, G. Velo, Scattering theory in the energy space for a claass of nonlinear Schrddinger equations, J. Math. Pures Appl. 9 64 (1985) 807836.

14. Y. Tsutsumi, L2 -solutions for nonlinear Schrodinger equations and nonlinear groups, Funkcial. Ekvac 30 (1987) 115-125.

15. T. Cazenave, F. B. Wiessler, The Gauchy problem for the critical nonlinear Schrddinger equation in Hs, Nonlinear Anal. 14 (1990) 807-836.

16. A. Vargas, L. Vega, Global well-posedness for Id non-linear Schrodinger equation for data with an infinite L2 norm, J. Math. Pures Appl. 80 10 (2001) 1029-1044.

17. J. Bourgain, Refinements of Strichartzz's inequality and applications to 2D-NLS with critical nonlinearity Internat. Math. Res. Notices 5 (1998) 253283.

18. T. Cazenave, F. B. Wiessler, Asymptotically self-similar global solutions of the nonlinear Schrddinger and heat equations, Math. Z. 228 (1998) 83-120.

19. F. Planchón, On the Cauchy problem for Besov spaces for a nonlinear Schrodinger equation, Comm. Cotemp. 106 3 (2001) 716-633.

20. T. Cazenave, L. Vega, M. C. Vilela, A note on the nonlinear Schrddinger equation in weak Lp spaces Comm. Contemp. Math. 3 1 (2001) 153-162.

21. A. Grünrock, Abstract in Mathematisches Forschungsinstitut Oberwolfach report 50 (2004).

0 0