Consider the Monge-Ampere problem:
\begin{equation}
\left{
det D 2 u = 1 \det D^2u = 1 detD2u=1
D u ( Ω ) = Ω ∗ Du(\Omega)=\Omega^* Du(Ω)=Ω∗
\right.
\end{equation}
If u : Ω → Ω ∗ u:\Omega\to\Omega^* u:Ω→Ω∗ is a convex function, and x ∈ Ω x\in\Omega x∈Ω is a Lebesgue point of u u u, then u u u is continuous at x x x.
If u : Ω → Ω ∗ u:\Omega\to\Omega^* u:Ω→Ω∗ is a convex function, and D u ( Ω ) ⊂ Ω ∗ Du(\Omega)\subset\Omega^* Du(Ω)⊂Ω∗, then we have ∀ E ⊂ B ( Ω ) , ∂ u ( E ) ⊂ Γ ( Ω ∗ ) \forall E\subset \mathscr{B}(\Omega), \partial u(E)\subset \Gamma(\Omega^*) ∀E⊂B(Ω),∂u(E)⊂Γ(Ω∗)where Γ ( Ω ∗ ) \Gamma(\Omega^*) Γ(Ω∗) is the convex hull of Ω ∗ \Omega^* Ω∗.
If u , Ω , Ω ∗ u,\Omega,\Omega^* u,Ω,Ω∗ as above, and under the following 3 conditions:
(i) u ( x ) = sup α L α ( x ) , u(x)=\sup \limits_{\alpha} L_\alpha (x), u(x)=αsupLα(x), for each x x x, where { L α } \{L_\alpha\} { Lα} is a family of affine functions, satisfying D L α ( x ) ∈ Ω ∗ DL_\alpha (x)\in\Omega^* DLα(x)∈Ω∗ for each α \alpha α.
(ii) λ ( Γ ( Ω ∗ ) − Ω ∗ ) = 0 \lambda(\Gamma(\Omega^*)-\Omega^*)=0 λ(Γ(Ω∗)−Ω∗)=0
(iii) ∀ h ∈ C ( R n ) \forall h \in C(\R^n) ∀h∈C(Rn) , one has the Brenier’s Monge-Ampere inequality: C 1 ∫ Ω h ( D u ( x ) ) d x ≤ ∫ Ω ∗ h ( y ) d y ≤ C 2 ∫ Ω h ( D u ( x ) ) d x C_1\int_{\Omega}h(Du(x))dx \leq\int_{\Omega^*}h(y)dy\leq C_2\int_\Omega h(Du(x))dx C1∫Ωh(Du(x))dx≤∫Ω∗h(y)dy≤C2∫Ωh(Du(x))dx
———————————————————————————————————
Then, one has the Aleksandrov’s Monge-Ampere inequality: C 1 χ Ω ≤ det D 2 u ≤ C 2 χ Ω C_1 \chi_\Omega\leq \det D^2 u\leq C_2 \chi _\Omega C1χΩ≤detD2u≤C2χΩin R n \R^n Rn.