MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

We know Frobenius theorem handle pde systems like $\{Xf=0, Yf=0\}$ requiring Lie bracket $[X,Y]\equiv 0 \mod X, Y$ for completely integrability of the system. However, how to handle systems like $\{Xf+Yg=0\}$. For example, how to solve $\frac{\partial u_i}{\partial x_j}=\frac{\partial u_j}{\partial x_i}$ for $i,j=1..n$?

share|cite|improve this question
Your example is about the closedness of the differential 1-form $\omega=u_1dx_1+\ldots+u_ndx_n$, which has the solution $u_i=\frac{\partial F}{\partial x_i}$, with arbitrary $F$. Concerning the general question, I would think in terms of differential consequences and prolongations (see, e.g., Kuranishi's "On E. Cartan's prolongation theorem of exterior differential systems."): in this context, "integrability" means that the prolonged system projects back to the whole original one (you may check that this reduces to the Frobenius theorem in the case of a PDE given by vector fields). – Giovanni Moreno Aug 28 '13 at 9:19
Thank you for the complement. How do you get the solution? By Poincare lemma or other stuffs? And I'm only a beginner on differential geometry and feel it hard to reduce? Could it be successfully reduced or just your suggestion? Would you mind offering more explanation? – Shuchang Aug 29 '13 at 4:06

[too long for a comment] Yes, the solution follows precisely from the Poincare' lemma. To see the relationship between Frobenius theorem and the theory of prolongations, let $\pi:\mathbb{R}^2\times\mathbb{R}\to\mathbb{R}^2$ be the trivial bundle, and $x,y$ coordinates on its basis. Let $$ X=a\partial_x+b\partial_y $$ $$ Y=a'\partial_x+b'\partial_y $$ and regard (any) vector field $X$ as a smooth function $\psi_X$ on the first jet extension $J^1(\pi)$, i.e., the space with coordinates $x,y,u,u_x,u_y$: $$ \psi_X=au_x+bu_y. $$ Recall that the function $\psi_X$ is intrinsically defined by the property $$ \psi_X\circ j_1(f)=X(f) $$ for any function $f=f(x,y)$. In this "jet formalism", your system of equations corresponds to the sub manifold $E=\{\psi_X=0,\psi_Y=0\}$ of $J^1(\pi)$. Now we must introduce the first prolongation $E^{(1)}$ of $E$, which is a submanifold of $J^2(\pi)$: to define it, take the total derivatives of the functions $\psi_X$ and $\psi_Y$: $$ D_x(\psi_X)=a_xu_x+au_{xx}+b_xu_y+bu_{xy} $$ $$ D_y(\psi_X)=a_yu_x+au_{xy}+b_yu_y+bu_{yy} $$ $$ D_x(\psi_Y)=a'_xu_x+a'u_{xx}+b'_xu_y+b'u_{xy} $$ $$ D_y(\psi_Y)=a'_yu_x+a'u_{xy}+b'_yu_y+b'u_{yy}. $$ In terms of ideals, $E$ corresponds to the ideal $I=\mathrm{Span}\{\psi_X,\psi_Y\}$ in $C^\infty(J^1(\pi))$, whereas $E^{(1)}$ corresponds to the ideal $I'=\mathrm{Span}\{\psi_X,\psi_Y,D_x(\psi_X),D_y(\psi_X),D_x(\psi_Y),D_y(\psi_Y)\}$ in $C^\infty(J^2(\pi))$. You should also remember that $C^\infty(J^1(\pi))$ is embedded into $C^\infty(J^2(\pi))$: then, the "algebraic formulation" of formal integrability reads $$ I'\cap C^\infty(J^1(\pi)) = I. $$ (Observe that the "$\supseteq$" inclusion always holds).

Now, you can easily check that $I'\cap C^\infty(J^1(\pi))=\mathrm{Span}\{\psi_X,\psi_Y,\psi_{[X,Y]}\}$: basically, you show that the only way (besides taking $\psi_X$ or $\psi_Y$ themselves) to get an element of $C^\infty(J^1(\pi))$ out of the generators of $I'$ is by means of the following $C^\infty(J^1(\pi))$-linear combination: $$ aD_x(\psi_Y)+bD_y(\psi_Y)-a'D_x(\psi_X)-b'D_y(\psi_X), $$ which turns out to be exactly $\psi_{[X,Y]}$, i.e., the function associated to the commutator $[X,Y]$. So,"formal integrability" holds if and only if $$ \mathrm{Span}\{\psi_X,\psi_Y,\psi_{[X,Y]}\}=\mathrm{Span}\{\psi_X,\psi_Y\}, $$ which is just a paraphrase of Frobenius' condition.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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