Let $ a $, $ b $ and $ c $ be real constants such that $ \Delta \stackrel{\text{df}}{=} a c - b^{2} > 0 $. The Beltrami Equations are defined as the following system of PDE’s on the domain $ \Bbb{R}^{2} $: $$ u_{x} = \frac{1}{\sqrt{\Delta}} (b v_{x} + c v_{y}), \qquad u_{y} = - \frac{1}{\sqrt{\Delta}} (a v_{x} + b v_{y}). $$ Boundary conditions are not imposed.

Question: Are solutions $ u $ and $ v $ of this system necessarily smooth?

I have been told that the Elliptic Regularity Theorem answers my question, but the theorem applies only to even-order elliptic operators and what we have here are only first-order equations, so we do not know if solutions $ u $ and $ v $ have even second-order partial derivatives. Observe that we have the same scenario for the Cauchy-Riemann Equations (in PDE form).

I understand that my question may be too basic for MathOverflow, so I understand if anyone wishes to close it. Thank you!

  • $\begingroup$ consider the extreme case of $ a=c=1$ and $b=0$. Then $u$ is harmonic. Can maybe do something similar in the more general case and still get an elliptic second order equation?? $\endgroup$
    – Math604
    Sep 29, 2015 at 5:44
  • $\begingroup$ @Math604: Hi Math604. Pardon me for being a little blunt here, but there’s a problem with your comment. If we don’t know that $ u $ has second-order partial derivatives, how can we claim that $ u $ is harmonic? $\endgroup$ Sep 29, 2015 at 6:11
  • 1
    $\begingroup$ It suffices to use weak second order partial derivatives (i.e., distributions). A weak solution to a linear elliptic PDE is necessarily smooth. Or you can work with the first order system directly, since it is itself an elliptic system. $\endgroup$
    – Deane Yang
    Sep 29, 2015 at 6:47
  • 1
    $\begingroup$ Not only the solutions are smooth, but your system rewrites $du=dv\circ J$, where $J^2=-Id$. Hence in coordinates where $J$ is $(s,t)\mapsto (-t,s)$, $u+iv$ is holomorphic in $s+it$. Moreover, Ahlfors and Bers proved that this generalizes to bounded measurable $J$ (but the solutions are no longer smooth if $J$ isn't, of course). See cimat.mx/~mmoreno/teaching/fall10/AhlforsBers.pdf $\endgroup$
    – BS.
    Sep 29, 2015 at 8:31
  • 1
    $\begingroup$ @Deane: Hi Deane. Your comment was really helpful, as it reminded me of the fact that distributions have weak derivatives of all orders. If $ u $ and $ v $ are distributions corresponding to locally-integrable functions, and they satisfy the Beltrami Equations in the weak sense, then by elliptic regularity, these functions are a.e.-equal to smooth functions. Hence, continuous solutions of the Beltrami Equations, viewed in the ordinary sense, are already smooth. $\endgroup$ Oct 2, 2015 at 16:19

1 Answer 1


As Deane Yang suggested you can regard the first order Beltrami system as an elliptic system, see Chapter 7.3 in the following book

Astala, Kari; Iwaniec, Tadeusz; Martin, Gaven Elliptic partial differential equations and quasiconformal mappings in the plane. Princeton Mathematical Series, 48. Princeton University Press, Princeton, NJ, 2009.

Also Chapter 5 and 6 give the answer for your question.

Since your equation is linear, you could have a representation formula for your equation, see for instance the following nice paper

Bojarski, B. V. Generalized solutions of a system of differential equations of the first order and elliptic type with discontinuous coefficients. Translated from the 1957 Russian original. With a foreword by Eero Saksman. Report. University of Jyväskylä Department of Mathematics and Statistics, 118. University of Jyväskylä, Jyväskylä, 2009.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.