### November 27, 2006

<ph f="cmbx">Local Analytic Hypoellipticity for a sum of squares of complex vector fields with large loss of derivatives</ph>

### David S. Tartakoff

5 rue de la Juviniere, 78350 Les Loges en Josas, FRANCE E-mail address : Makhlouf.Derridj@univ-rouen.fr Department of Mathematics, University of Illinois at Chicago, m/c 249, 851 S. Morgan St., Chicago IL 60607, USA E-mail address : dst@uic.edu
• Abstract. A very recent paper of Kohn studies the hypoellipticity for a sum of squres of complex vector fields which exhibit a large loss of derivatives. We give an elementary proof of the analytic hypoellipticity of this operator. The proof requires the analysis of [2.
M. Derridj and D.S.Tartakoff

1 Introduction

In [1, J.J. Kohn proved the hypoellipticity of the operator $P={L}^{*}L+\left({\overline{z}}^{k}L\right)\left({\overline{z}}^{k}L{\right)}^{*},L=\frac{\partial }{\partial z}-i\overline{z}\frac{\partial }{\partial t},$  for which there is a large loss of derivatives indeed in the a priori estimate one bounds only the Sobolev norm of order $-\left(k+1\right)/2.$  We show in this note that solutions of $Pu=f$  with $f$  real analytic are themselves real analytic in any open set where $f$  is. For simplicity we shall write out in detail the case for $k=1$  only, as larger values of $k$  pose no difficulties.
The a priori estimate with which we will work is $\parallel Lv{\parallel }_{0}^{2}+\parallel z\overline{L}v{\parallel }_{0}^{2}\lesssim |\left(Pv,v{\right)}_{{L}^{2}}|,v\in {C}_{0}^{\infty }$  since the loss of derivatives and extra low order term on the left is of no use to us and this estimate is immediate from the definition of $P.$  Our first observation is that we know the analyticity of the solution for $z\ne 0$  from the earlier work of the second author [2, [3and Treves [4.
Our second observation is that it suffices to bound derivatives measured in terms of high powers of the vector fields $L$  and $\overline{L}$  in ${L}^{2}$  norm, by standard arguments, and indeed estimating high powers of $\overline{L}$  can be reduced to bounding high powers of $L$  and powers of $T=2i\partial /\partial t=\left[L,\overline{L}\right]$  of half the order, by repeated integration by parts. Thus our overall scheme will be to start with high powers (order $2p$  ) of $L$  or $\overline{L},$  use integration by parts and the a priori estimate repeatedly to reduce to treating ${T}^{p}u$  in a slightly larger set.
And to do this, as in many other papers of ours, we will create a (new) special localization of ${T}^{p}.$  The new localization of ${T}^{p}$  may be written in the simple form:
Local Analyticity for complex vector fields $\left({T}^{{p}_{1},{p}_{2}}{\right)}_{\phi }={\sum }_{\genfrac{}{}{0}{}{a\le {p}_{1}}{b\le {p}_{2}}}\frac{{L}^{a}{z}^{a}{T}^{{p}_{1}-a}{\phi }^{\left(a+b\right)}{T}^{{p}_{2}-b}{\overline{z}}^{b}{\overline{L}}^{b}}{a!b!},$  Here by ${\phi }^{\left(r\right)}$  we mean $\left(i\partial /\partial t{\right)}^{r}\phi \left(t\right)$  since near $z=0$  we may take the localizing function independent of $z.$  We have the commutation relations:
$\left[L,\left({T}^{{p}_{1},{p}_{2}}{\right)}_{\phi }\right]\equiv L\circ \left({T}^{{p}_{1}-1,{p}_{2}}{\right)}_{{\phi }^{\prime }},\left[\overline{L},\left({T}^{{p}_{1},{p}_{2}}{\right)}_{\phi }\right]\equiv \left({T}^{{p}_{1},{p}_{2}-1}{\right)}_{{\phi }^{\prime }}\circ \overline{L},$  $\left[\left({T}^{{p}_{1},{p}_{2}}{\right)}_{\phi },z\right]=\left({T}^{{p}_{1}-1,{p}_{2}}{\right)}_{{\phi }^{\prime }}\circ z,and\left[\left({T}^{{p}_{1},{p}_{2}}{\right)}_{\phi },\overline{z}\right]=\overline{z}\circ \left({T}^{{p}_{1},{p}_{2}-1}{\right)}_{{\phi }^{\prime }},$  where the $\equiv$  denotes modulo ${C}^{{p}_{1}-{p}_{1}^{\prime }+{p}_{2}-{p}_{2}^{\prime }}$  terms of the form $\frac{{L}^{{p}_{1}-{p}_{1}^{\prime }}\circ {z}^{{p}_{1}-{p}_{1}^{\prime }}\circ {T}^{{p}_{1}^{\prime }}\circ {\phi }^{\left({p}_{1}^{\prime }+{p}_{2}^{\prime }+1\right)}\circ {T}^{{p}_{2}^{\prime }}\circ {\overline{z}}^{{p}_{2}-{p}_{2}^{\prime }}\circ {\overline{L}}^{{p}_{2}-{p}_{2}^{\prime }}}{\left({p}_{1}-{p}_{1}^{\prime }+{p}_{2}-{p}_{2}^{\prime }\right)!}$  with either ${p}_{1}^{\prime }=0$  or ${p}_{2}^{\prime }=0.$  Note that if we start with ${p}_{1}={p}_{2}=p/2,$  and iteratively apply these commutation relations, the number of $T$  derivatives not necessarily applied to $\phi$  is eventually at most $p/2;$  in our previous works the number dropped to zero (i.e., only $L$  and $\overline{L}$  derivatives remained and at most $p$  of them, half the number of $L$  and $\overline{L}$  derivatives we started with, but any fraction of the original number would do as well.
So we insert first $v=\left({T}^{p/2,p/2}{\right)}_{\phi }u$  in the a priori inequality, then bring $\left({T}^{p/2,p/2}{\right)}_{\phi }$  to the left of $P=-\overline{L}L-L\overline{z}z\overline{L}$  since $Pu$  is known and analytic. By the above bracket relations, $\left(\left[\overline{L}L+L\overline{z}z\overline{L},\left({T}^{p/2,p/2}{\right)}_{\phi }\right]u,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right)\equiv$  M. Derridj and D.S.Tartakoff $\equiv \left(\left({T}^{p/2,p/2-1}{\right)}_{{\phi }^{\prime }}\overline{L}Lu,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right)+\left(\overline{L}L\left({T}^{p/2-1,p/2}{\right)}_{{\phi }^{\prime }}u,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right)$  $+\left(L\left({T}^{p/2-1,p/2}{\right)}_{{\phi }^{\prime }}\overline{z}z\overline{L}u,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right)+\left(L\overline{z}\left({T}^{p/2,p/2-1}{\right)}_{{\phi }^{\prime }}z\overline{L}u,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right)$  $+\left(L\overline{z}\left({T}^{p/2-1,p/2}{\right)}_{{\phi }^{\prime }}z\overline{L}u,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right)+\left(L\overline{z}z\left({T}^{p/2,p/2-1}{\right)}_{{\phi }^{\prime }}\overline{L}u,\left({T}^{p/2,p/2}{\right)}_{\phi }u\right).$  In every case, the brackets reduce the order of the sum of the two indices ${p}_{1}$  and ${p}_{2}$  by one (here we started with ${p}_{1}={p}_{2}=p/2$  ), pick up one derivative on $\phi ,$  and leave the vector fields over which we have maximal control in the estimate intact and in the correct order. Thus we may bring either $L\overline{z}$  to the right as $z\overline{L},$  (possibly after additional iterations of the brackets to obtain an $\overline{z}$  to the left of $\left({T}^{{q}_{1},{q}_{2}}{\right)}_{\psi }$  ), and use a weighted Schwarz inequality on the result to take maximal advantage of the a priori inequality. Iterations of all of this continue until there remain at most $p/2$  free $T$  derivatives (i.e., the $T$  derivatives on at least one side of $\phi$  are all `corrected' by good vector fields) and perhaps as many as $p/2L$  or $z\overline{L}$  derivatives, and we may continue further until, at worst, the remaining $L$  and $\overline{L}$  derivatives bracket two at a time to produce more $T$  's, one at a time. After all of this, there will be at most ${T}^{3p/4},$  and at this point we introduce a new localizing function of Ehrenpreis type with slightly larger support, geared to $3p/4$  instead of to $p.$  In our previous work, the order dropped by half and we used ${log}_{2}N$  nested open sets, but here we employ ${log}_{4/3}N$  nested open sets of prescribed size growth may be constructed to prove that for derivatives of order no greater than $N,$  $|{D}^{|\alpha |}u|\le C{C}^{N}{N}^{N}\sim {C}^{\prime }{{C}^{\prime }}^{N}N!$  locally with $C$  independent of $N,$  which proves the analyticity of the solution. Local Analyticity for complex vector fields The situation is not different with larger values of $k,$  and we do not write this out.
References

1. J.J. Kohn, Hypoellipticity and loss of derivatives, Annals of Mathematics, to appear.
2. D.S. Tartakoff, Local Analytic Hypoellipticity for ${\square }_{b}$  on Non-Degenerate Cauchy Riemann Manifolds, Proc. Nat. Acad. Sci. U.S.A. 75 (1978), pp. 3027-3028.
3. D.S. Tartakoff, On the Local Real Analyticity of Solutions to ${\square }_{b}$  and the $\overline{\partial }$  -Neumann Problem, Acta Math. 145 (1980), pp. 117-204.
4. F. Treves, Analytic Hypo-ellipticity of a Class of Pseudo-Differential Operators with Double Characteristics and Application to the $\overline{\partial }$  -Neumann Problem, Comm. in P.D.E. 3 (6-7) (1978), pp. 475-642.

5 rue de la Juviniere, 78350 Les Loges en Josas, FRANCE E-mail address : Makhlouf.Derridj@univ-rouen.fr Department of Mathematics, University of Illinois at Chicago, m/c 249, 851 S. Morgan St., Chicago IL 60607, USA E-mail address : dst@uic.edu