Morrey–Kohn–Hörmander Formula (Theorem # 3685)
Theorem
Let $\Omega \subset \mathbb{C}^n$ be a bounded domain with $C^\infty$ boundary, and let $\rho \in C^\infty(\mathbb{C}^n; \mathbb{R})$ be a defining function for $\Omega$, i.e., $\Omega = \{\rho < 0\}$ and $\nabla\rho \neq 0$ on $\partial\Omega$. Define the weighted surface measure on $\partial\Omega$ by
\begin{align*}
dS_\rho := \frac{d\sigma}{|\nabla\rho|},
\end{align*}
where $d\sigma$ is the induced Euclidean surface measure on $\partial\Omega$.
Fix an integer $q$ with $1 \leq q \leq n$. For multi-indices, let $\sum'$ denote summation over strictly increasing multi-indices. For an arbitrary multi-index $(i_1,\dots,i_p)$, extend coefficients $u_{i_1\dots i_p}$ by total antisymmetry, so $u_{i_1\dots i_p} = 0$ whenever any two indices coincide.
Let $u \in C^\infty_{(0,q)}(\overline{\Omega})$ be a smooth $(0,q)$-form on $\overline{\Omega}$, written in coordinates as
\begin{align*}
u = \sum_{|J|=q}{}' u_J \, d\bar z_J, \qquad u_J \in C^\infty(\overline{\Omega}; \mathbb{C}).
\end{align*}
Assume $u$ satisfies the **$\bar\partial$-Neumann boundary condition**
\begin{align*}
\sum_{j=1}^n \frac{\partial \rho}{\partial z_j}(x)\, u_{jK}(x) = 0 \qquad \text{for all } x \in \partial\Omega \text{ and all } K \text{ with } |K| = q-1.
\end{align*}
Let $\bar\partial^* u$ be defined by the interior formula
\begin{align*}
\bar\partial^* u := -\sum_{|K|=q-1}{}' \left(\sum_{j=1}^n \frac{\partial u_{jK}}{\partial z_j}\right) d\bar z_K,
\end{align*}
which coincides with the Hilbert-space adjoint of $\bar\partial$ on smooth forms satisfying the above boundary condition. Then
\begin{align*}
\|\bar\partial u\|_{L^2(\Omega)}^2 + \|\bar\partial^* u\|_{L^2(\Omega)}^2
&= \sum_{|J|=q}{}' \sum_{j=1}^n \left\|\frac{\partial u_J}{\partial \bar z_j}\right\|_{L^2(\Omega)}^2 \\
&\quad + \sum_{|K|=q-1}{}' \int_{\partial\Omega} \sum_{j,k=1}^n \frac{\partial^2 \rho}{\partial z_j \, \partial \bar z_k}(x)\, u_{jK}(x)\, \overline{u_{kK}(x)} \, dS_\rho(x).
\end{align*}
Discussion
This identity decomposes the L2 energy of dbar and its adjoint into interior derivative terms plus a boundary Levi form contribution. It is a core estimate in dbar theory because it connects analytic norms with the geometry of the boundary.
Proof
[proofplan]
We compute $\|\bar\partial u\|_{L^2}^2$ and $\|\bar\partial^* u\|_{L^2}^2$ in coordinates. A combinatorial expansion of the wedge product reduces $\|\bar\partial u\|_{L^2}^2$ to a "diagonal" piece $\sum_{J,j} \|\partial_{\bar z_j} u_J\|^2$ minus a "cross" piece $\sum_{K,j,k}(\partial_{\bar z_j} u_{kK}, \partial_{\bar z_k} u_{jK})$, while $\|\bar\partial^* u\|_{L^2}^2$ contributes a "divergence" piece $\sum_{K,j,k}(\partial_{z_j} u_{jK}, \partial_{z_k} u_{kK})$. Integrating by parts twice converts the difference of cross and divergence pieces into a pure boundary integral; the $\bar\partial$-Neumann boundary condition annihilates one of the resulting boundary terms, and tangential differentiation of the conjugate boundary condition converts the remaining boundary term into the complex Hessian (Levi form) of $\rho$ paired against the boundary values of $u$.
[/proofplan]
[step:Set up complex coordinates and the integration by parts identity]
Write $z_j = x_j + i y_j$ and use the Wirtinger derivatives
\begin{align*}
\frac{\partial}{\partial z_j} = \tfrac{1}{2}\!\left(\frac{\partial}{\partial x_j} - i\frac{\partial}{\partial y_j}\right), \qquad
\frac{\partial}{\partial \bar z_j} = \tfrac{1}{2}\!\left(\frac{\partial}{\partial x_j} + i\frac{\partial}{\partial y_j}\right).
\end{align*}
For $f, g \in C^\infty(\overline\Omega; \mathbb{C})$, applying the [Divergence Theorem](/theorems/3614) componentwise to the real and imaginary parts of $f\bar g$ and combining yields
\begin{align*}
(\partial_{z_j} f, g)_{L^2(\Omega)} &= -(f, \partial_{\bar z_j} g)_{L^2(\Omega)} + \int_{\partial\Omega} f\, \bar g \, \frac{\partial \rho}{\partial z_j} \, dS_\rho, \\
(\partial_{\bar z_j} f, g)_{L^2(\Omega)} &= -(f, \partial_{z_j} g)_{L^2(\Omega)} + \int_{\partial\Omega} f\, \bar g \, \frac{\partial \rho}{\partial \bar z_j} \, dS_\rho,
\end{align*}
where $(\phi, \psi)_{L^2(\Omega)} := \int_\Omega \phi \, \bar\psi \, d\mathcal{L}^{2n}$. We use these formulas repeatedly below.
[guided]
The Wirtinger calculus is essential: it lets us treat $z_j$ and $\bar z_j$ as independent. To verify the integration-by-parts formula, write
\begin{align*}
\partial_{z_j}(f \bar g) = (\partial_{z_j} f)\, \bar g + f\, (\partial_{z_j} \bar g) = (\partial_{z_j} f)\, \bar g + f\, \overline{\partial_{\bar z_j} g},
\end{align*}
where the last equality uses $\partial_{z_j} \bar g = \overline{\partial_{\bar z_j} g}$. Integrating over $\Omega$ and applying the [Divergence Theorem](/theorems/3614) to the vector field whose Cartesian components recombine into $\partial_{z_j}(f\bar g)$ (specifically: $\partial_{z_j} = \frac{1}{2}(\partial_{x_j} - i \partial_{y_j})$ contributes the half-sum of the $x_j$ and $-i y_j$ divergences) gives
\begin{align*}
\int_\Omega \partial_{z_j}(f\bar g) \, d\mathcal{L}^{2n} = \int_{\partial\Omega} f \bar g \cdot \tfrac{1}{2}\!\left(\frac{\partial_{x_j}\rho}{|\nabla\rho|} - i \frac{\partial_{y_j}\rho}{|\nabla\rho|}\right) d\sigma = \int_{\partial\Omega} f\bar g \, \frac{\partial\rho}{\partial z_j} \, dS_\rho,
\end{align*}
since the outer unit normal at $x \in \partial\Omega$ is $\nabla \rho(x)/|\nabla\rho(x)|$ (because $\rho < 0$ inside $\Omega$ and $\rho = 0$ on $\partial\Omega$). Rearranging the product-rule identity yields the first formula; the second follows by complex conjugation. Throughout the proof we will need both formulas: $\partial_{z_j}$ and $\partial_{\bar z_j}$ are formal adjoints (up to a sign and boundary term) of $-\partial_{\bar z_j}$ and $-\partial_{z_j}$ respectively.
[/guided]
[/step]
[step:Expand $\|\bar\partial u\|_{L^2}^2$ in coordinates]
Since $\bar\partial u = \sum_{J}{}' \sum_j \partial_{\bar z_j} u_J \, d\bar z_j \wedge d\bar z_J$, we claim
\begin{align*}
\|\bar\partial u\|_{L^2(\Omega)}^2 = \sum_{|J|=q}{}' \sum_{j=1}^n \|\partial_{\bar z_j} u_J\|_{L^2(\Omega)}^2 \;-\; \sum_{|K|=q-1}{}' \sum_{j,k=1}^n (\partial_{\bar z_j} u_{kK},\, \partial_{\bar z_k} u_{jK})_{L^2(\Omega)},
\tag{$\ast$}
\end{align*}
where in the second sum the components $u_{jK}, u_{kK}$ are extended by total antisymmetry.
[claim:Coordinate formula for $\|\bar\partial u\|^2$]
Equation $(\ast)$ holds.
[/claim]
[proof]
For $|L| = q+1$ and $j \in L$, let $\epsilon_{j,L} := \mathrm{sgn}(\pi)$ where $\pi$ is the permutation sending $(j, L \setminus j)$ (with $L \setminus j$ in increasing order) to $L$ in increasing order. Then
\begin{align*}
\bar\partial u = \sum_{|L|=q+1}{}' (\bar\partial u)_L \, d\bar z_L, \qquad (\bar\partial u)_L = \sum_{j \in L} \epsilon_{j,L}\, \partial_{\bar z_j} u_{L \setminus j}.
\end{align*}
Therefore
\begin{align*}
\|\bar\partial u\|_{L^2}^2 = \sum_{|L|=q+1}{}' \int_\Omega \left|\sum_{j \in L} \epsilon_{j,L}\, \partial_{\bar z_j} u_{L\setminus j}\right|^2 d\mathcal{L}^{2n} = \mathrm{D} + \mathrm{O},
\end{align*}
where $\mathrm{D}$ and $\mathrm{O}$ are the diagonal ($j = k$) and off-diagonal ($j \neq k$) contributions of the expanded square.
**Diagonal.** Reparametrising $L = \{j\} \cup J$ with $J = L \setminus j$ ordered increasingly and $j \notin J$,
\begin{align*}
\mathrm{D} = \sum_{|L|=q+1}{}' \sum_{j \in L} \|\partial_{\bar z_j} u_{L \setminus j}\|_{L^2}^2 = \sum_{|J|=q}{}' \sum_{j \notin J} \|\partial_{\bar z_j} u_J\|_{L^2}^2.
\end{align*}
**Off-diagonal.** For distinct $j, k \in L$, set $K := L \setminus \{j, k\}$ ordered increasingly, $|K| = q-1$. For $a \notin K$, define $\delta_{a,K} := \operatorname{sgn}((a,K) \to \{a\} \cup K)$, where $\{a\} \cup K$ is written in increasing order. The antisymmetric extension gives $u_{aK} = \delta_{a,K} u_{\{a\}\cup K}$. Since $L \setminus j = \{k\} \cup K$ and $L \setminus k = \{j\} \cup K$,
\begin{align*}
\partial_{\bar z_j}u_{L\setminus j} = \delta_{k,K}\,\partial_{\bar z_j}u_{kK}, \qquad
\partial_{\bar z_k}u_{L\setminus k} = \delta_{j,K}\,\partial_{\bar z_k}u_{jK}.
\end{align*}
Moreover,
\begin{align*}
\epsilon_{j,L}\delta_{k,K} = \operatorname{sgn}((j,k,K) \to L), \qquad
\epsilon_{k,L}\delta_{j,K} = \operatorname{sgn}((k,j,K) \to L).
\end{align*}
The ordered lists $(j,k,K)$ and $(k,j,K)$ differ by exactly one transposition, so
\begin{align*}
(\epsilon_{j,L}\delta_{k,K})(\epsilon_{k,L}\delta_{j,K}) = -1.
\end{align*}
Hence
\begin{align*}
\mathrm{O} = -\sum_{|K|=q-1}{}' \sum_{j \neq k} (\partial_{\bar z_j} u_{kK},\, \partial_{\bar z_k} u_{jK})_{L^2}.
\end{align*}
The antisymmetric extension makes $u_{jK} = 0$ whenever $j \in K$, so adding the $j = k$ diagonal of the second sum costs $\sum_{K}{}' \sum_j \|\partial_{\bar z_j} u_{jK}\|^2 = \sum_{J}{}'\sum_{j \in J}\|\partial_{\bar z_j} u_J\|^2$ (parametrising $J = \{j\} \cup K$). Combining,
\begin{align*}
\mathrm{D} + \mathrm{O} = \sum_{J}{}'\!\sum_{j \notin J}\|\partial_{\bar z_j} u_J\|^2 + \sum_{J}{}'\!\sum_{j \in J}\|\partial_{\bar z_j} u_J\|^2 - \sum_{K}{}'\sum_{j,k}(\partial_{\bar z_j} u_{kK}, \partial_{\bar z_k} u_{jK})_{L^2},
\end{align*}
which is precisely $(\ast)$.
[/proof]
[/step]
[step:Compute $\|\bar\partial^* u\|_{L^2}^2$ from the interior formula]
By the interior formula defining $\bar\partial^* u$,
\begin{align*}
\|\bar\partial^* u\|_{L^2(\Omega)}^2 = \sum_{|K|=q-1}{}' \int_\Omega \left|\sum_{j=1}^n \partial_{z_j} u_{jK}\right|^2 d\mathcal{L}^{2n} = \sum_{|K|=q-1}{}' \sum_{j,k=1}^n (\partial_{z_j} u_{jK},\, \partial_{z_k} u_{kK})_{L^2(\Omega)}.
\tag{$\ast\ast$}
\end{align*}
Combining $(\ast)$ and $(\ast\ast)$,
\begin{align*}
\|\bar\partial u\|_{L^2}^2 + \|\bar\partial^* u\|_{L^2}^2 = \sum_{|J|=q}{}'\sum_{j=1}^n \|\partial_{\bar z_j} u_J\|_{L^2}^2 + \Theta,
\end{align*}
where
\begin{align*}
\Theta := \sum_{|K|=q-1}{}' \sum_{j,k=1}^n \Big[(\partial_{z_j} u_{jK},\, \partial_{z_k} u_{kK})_{L^2} - (\partial_{\bar z_j} u_{kK},\, \partial_{\bar z_k} u_{jK})_{L^2}\Big].
\end{align*}
It remains to show that $\Theta$ equals the boundary integral of the complex Hessian.
[/step]
[step:Convert $\Theta$ to a boundary integral via double integration by parts]
Fix $K$ with $|K| = q-1$ and $j, k \in \{1, \dots, n\}$. Apply the first integration-by-parts identity of Step 1 to move $\partial_{z_j}$ off $u_{jK}$:
\begin{align*}
(\partial_{z_j} u_{jK}, \partial_{z_k} u_{kK})_{L^2} = -(u_{jK}, \partial_{\bar z_j} \partial_{z_k} u_{kK})_{L^2} + \int_{\partial\Omega} u_{jK}\, \overline{\partial_{z_k} u_{kK}}\, \frac{\partial \rho}{\partial z_j}\, dS_\rho.
\end{align*}
Since $u_{kK} \in C^\infty(\overline\Omega; \mathbb{C})$, the mixed partials commute: $\partial_{\bar z_j} \partial_{z_k} = \partial_{z_k} \partial_{\bar z_j}$. Now apply the formula $(f, \partial_{z_k} g)_{L^2} = -(\partial_{\bar z_k} f, g)_{L^2} + \int_{\partial\Omega} f \bar g \, \partial_{\bar z_k} \rho \, dS_\rho$ (obtained by complex-conjugating the second identity of Step 1) with $f = u_{jK}$ and $g = \partial_{\bar z_j} u_{kK}$:
\begin{align*}
(u_{jK}, \partial_{z_k} \partial_{\bar z_j} u_{kK})_{L^2} = -(\partial_{\bar z_k} u_{jK}, \partial_{\bar z_j} u_{kK})_{L^2} + \int_{\partial\Omega} u_{jK}\, \overline{\partial_{\bar z_j} u_{kK}}\, \frac{\partial \rho}{\partial \bar z_k}\, dS_\rho.
\end{align*}
Substituting,
\begin{align*}
(\partial_{z_j} u_{jK}, \partial_{z_k} u_{kK})_{L^2} = (\partial_{\bar z_k} u_{jK}, \partial_{\bar z_j} u_{kK})_{L^2} + B_{j,k,K}^{(1)} - B_{j,k,K}^{(2)},
\end{align*}
where
\begin{align*}
B_{j,k,K}^{(1)} &:= \int_{\partial\Omega} u_{jK}\, \overline{\partial_{z_k} u_{kK}}\, \frac{\partial \rho}{\partial z_j}\, dS_\rho, \\
B_{j,k,K}^{(2)} &:= \int_{\partial\Omega} u_{jK}\, \overline{\partial_{\bar z_j} u_{kK}}\, \frac{\partial \rho}{\partial \bar z_k}\, dS_\rho.
\end{align*}
Under the swap $j \leftrightarrow k$, the inner product $(\partial_{\bar z_k} u_{jK}, \partial_{\bar z_j} u_{kK})_{L^2}$ becomes $(\partial_{\bar z_j} u_{kK}, \partial_{\bar z_k} u_{jK})_{L^2}$. Hence after summing over $j$ and $k$,
\begin{align*}
\sum_{j,k}(\partial_{\bar z_k} u_{jK}, \partial_{\bar z_j} u_{kK})_{L^2} = \sum_{j,k}(\partial_{\bar z_j} u_{kK}, \partial_{\bar z_k} u_{jK})_{L^2},
\end{align*}
so the volume terms in $\Theta$ cancel and
\begin{align*}
\Theta = \sum_{|K|=q-1}{}' \sum_{j,k=1}^n \big[B_{j,k,K}^{(1)} - B_{j,k,K}^{(2)}\big].
\end{align*}
[guided]
Fix $K$ with $|K| = q-1$ and fix $j,k \in \{1,\dots,n\}$. We want to compare the divergence inner product
\begin{align*}
(\partial_{z_j} u_{jK}, \partial_{z_k} u_{kK})_{L^2(\Omega)}
\end{align*}
with the cross inner product
\begin{align*}
(\partial_{\bar z_j} u_{kK}, \partial_{\bar z_k} u_{jK})_{L^2(\Omega)}.
\end{align*}
The mechanism is two integrations by parts, with the commutation of mixed Wirtinger derivatives used between them.
Apply the first integration-by-parts identity from Step 1 with $f = u_{jK}$ and $g = \partial_{z_k}u_{kK}$. The functions are smooth on $\overline\Omega$, so the identity applies. We obtain
\begin{align*}
(\partial_{z_j} u_{jK}, \partial_{z_k} u_{kK})_{L^2(\Omega)}
&= -(u_{jK}, \partial_{\bar z_j}\partial_{z_k}u_{kK})_{L^2(\Omega)} \\
&\quad + \int_{\partial\Omega} u_{jK}\,\overline{\partial_{z_k}u_{kK}}\,\frac{\partial\rho}{\partial z_j}\,dS_\rho.
\end{align*}
Since $u_{kK} \in C^\infty(\overline\Omega;\mathbb{C})$, the real partial derivatives commute, and therefore the Wirtinger derivatives commute:
\begin{align*}
\partial_{\bar z_j}\partial_{z_k}u_{kK} = \partial_{z_k}\partial_{\bar z_j}u_{kK}.
\end{align*}
Thus the remaining volume term is
\begin{align*}
(u_{jK}, \partial_{z_k}\partial_{\bar z_j}u_{kK})_{L^2(\Omega)}.
\end{align*}
Now use the rearranged second integration-by-parts identity from Step 1,
\begin{align*}
(f,\partial_{z_k}g)_{L^2(\Omega)} = -(\partial_{\bar z_k}f,g)_{L^2(\Omega)} + \int_{\partial\Omega} f\,\bar g\,\frac{\partial\rho}{\partial\bar z_k}\,dS_\rho,
\end{align*}
with $f = u_{jK}$ and $g = \partial_{\bar z_j}u_{kK}$. This gives
\begin{align*}
(u_{jK}, \partial_{z_k}\partial_{\bar z_j}u_{kK})_{L^2(\Omega)}
&= -(\partial_{\bar z_k}u_{jK}, \partial_{\bar z_j}u_{kK})_{L^2(\Omega)} \\
&\quad + \int_{\partial\Omega} u_{jK}\,\overline{\partial_{\bar z_j}u_{kK}}\,\frac{\partial\rho}{\partial\bar z_k}\,dS_\rho.
\end{align*}
Define the two boundary terms
\begin{align*}
B_{j,k,K}^{(1)} &:= \int_{\partial\Omega} u_{jK}\,\overline{\partial_{z_k}u_{kK}}\,\frac{\partial\rho}{\partial z_j}\,dS_\rho, \\
B_{j,k,K}^{(2)} &:= \int_{\partial\Omega} u_{jK}\,\overline{\partial_{\bar z_j}u_{kK}}\,\frac{\partial\rho}{\partial\bar z_k}\,dS_\rho.
\end{align*}
Substituting the second integration-by-parts formula into the first one yields
\begin{align*}
(\partial_{z_j} u_{jK}, \partial_{z_k} u_{kK})_{L^2(\Omega)}
= (\partial_{\bar z_k}u_{jK}, \partial_{\bar z_j}u_{kK})_{L^2(\Omega)} + B_{j,k,K}^{(1)} - B_{j,k,K}^{(2)}.
\end{align*}
This is the exact point where the volume terms prepare to cancel: after summing over $j$ and $k$, the finite relabeling $(j,k) \mapsto (k,j)$ gives
\begin{align*}
\sum_{j,k=1}^n(\partial_{\bar z_k}u_{jK}, \partial_{\bar z_j}u_{kK})_{L^2(\Omega)}
= \sum_{j,k=1}^n(\partial_{\bar z_j}u_{kK}, \partial_{\bar z_k}u_{jK})_{L^2(\Omega)}.
\end{align*}
Therefore, in the quantity
\begin{align*}
\Theta := \sum_{|K|=q-1}{}'\sum_{j,k=1}^n\Big[(\partial_{z_j}u_{jK},\partial_{z_k}u_{kK})_{L^2(\Omega)}-(\partial_{\bar z_j}u_{kK},\partial_{\bar z_k}u_{jK})_{L^2(\Omega)}\Big],
\end{align*}
the summed volume contribution cancels with the cross term. What remains is precisely the boundary contribution
\begin{align*}
\Theta = \sum_{|K|=q-1}{}'\sum_{j,k=1}^n\big[B_{j,k,K}^{(1)}-B_{j,k,K}^{(2)}\big].
\end{align*}
This completes the conversion of $\Theta$ from an interior expression into boundary terms.
[/guided]
[/step]
[step:Annihilate $B^{(1)}$ using the $\bar\partial$-Neumann boundary condition]
For each $K$ with $|K| = q-1$, factor the $j$-sum out of $B^{(1)}_{j,k,K}$:
\begin{align*}
\sum_{j,k=1}^n B_{j,k,K}^{(1)} = \sum_{k=1}^n \int_{\partial\Omega} \left(\sum_{j=1}^n \frac{\partial \rho}{\partial z_j} u_{jK}\right) \overline{\partial_{z_k} u_{kK}} \, dS_\rho.
\end{align*}
By hypothesis, $\sum_{j} (\partial \rho / \partial z_j)\, u_{jK} = 0$ pointwise on $\partial\Omega$, so $\sum_{j,k} B^{(1)}_{j,k,K} = 0$ for every $K$. Therefore
\begin{align*}
\Theta = -\sum_{|K|=q-1}{}' \sum_{j,k=1}^n B_{j,k,K}^{(2)} = -\sum_{|K|=q-1}{}' \sum_{j,k=1}^n \int_{\partial\Omega} u_{jK}\, (\partial_{z_j} \overline{u_{kK}})\, \frac{\partial \rho}{\partial \bar z_k}\, dS_\rho,
\end{align*}
where we used $\overline{\partial_{\bar z_j} u_{kK}} = \partial_{z_j} \overline{u_{kK}}$.
[/step]
[step:Convert the residual boundary term to the Levi form by tangential differentiation]
Take complex conjugates of the $\bar\partial$-Neumann condition: since $\rho$ is real, $\overline{\partial \rho / \partial z_j} = \partial \rho / \partial \bar z_j$, and so
\begin{align*}
\sum_{k=1}^n \frac{\partial \rho}{\partial \bar z_k}\, \overline{u_{kK}} = 0 \qquad \text{on } \partial\Omega \text{ for every } K \text{ with } |K| = q-1.
\end{align*}
A smooth $\mathbb{C}$-valued function on $\overline\Omega$ vanishing on $\partial\Omega$ is divisible by $\rho$ in $C^\infty(\overline\Omega)$ (Hadamard's lemma applied componentwise), so there exists $h_K \in C^\infty(\overline\Omega; \mathbb{C})$ with
\begin{align*}
\sum_{k=1}^n \frac{\partial \rho}{\partial \bar z_k}\, \overline{u_{kK}} = h_K\, \rho \qquad \text{on } \overline\Omega.
\end{align*}
Apply $\partial_{z_j}$ to both sides:
\begin{align*}
\sum_{k=1}^n \frac{\partial^2 \rho}{\partial z_j\, \partial \bar z_k}\, \overline{u_{kK}} + \sum_{k=1}^n \frac{\partial \rho}{\partial \bar z_k}\, \partial_{z_j} \overline{u_{kK}} = (\partial_{z_j} h_K)\, \rho + h_K\, \frac{\partial \rho}{\partial z_j}.
\end{align*}
Restricting to $\partial\Omega$, where $\rho = 0$, gives
\begin{align*}
\sum_{k=1}^n \frac{\partial \rho}{\partial \bar z_k}\, \partial_{z_j} \overline{u_{kK}} = -\sum_{k=1}^n \frac{\partial^2 \rho}{\partial z_j\, \partial \bar z_k}\, \overline{u_{kK}} + h_K\, \frac{\partial \rho}{\partial z_j} \qquad \text{on } \partial\Omega.
\end{align*}
Multiply by $u_{jK}$ and sum over $j$. The term $h_K \sum_j u_{jK} (\partial \rho/\partial z_j)$ vanishes on $\partial\Omega$ by the boundary condition, yielding the pointwise identity on $\partial\Omega$
\begin{align*}
\sum_{j,k=1}^n u_{jK}\, \frac{\partial \rho}{\partial \bar z_k}\, \partial_{z_j} \overline{u_{kK}} = -\sum_{j,k=1}^n \frac{\partial^2 \rho}{\partial z_j\, \partial \bar z_k}\, u_{jK}\, \overline{u_{kK}}.
\end{align*}
Integrating against $dS_\rho$ and inserting into the formula for $\Theta$ from Step 5,
\begin{align*}
\Theta = -\sum_{|K|=q-1}{}' \int_{\partial\Omega} \sum_{j,k=1}^n u_{jK}\, \frac{\partial \rho}{\partial \bar z_k}\, \partial_{z_j} \overline{u_{kK}} \, dS_\rho = \sum_{|K|=q-1}{}' \int_{\partial\Omega} \sum_{j,k=1}^n \frac{\partial^2 \rho}{\partial z_j\, \partial \bar z_k}\, u_{jK}\, \overline{u_{kK}} \, dS_\rho.
\end{align*}
[guided]
The key obstacle is that $\partial_{z_j}$ is *not* a tangential operator on $\partial\Omega$ in general — applying it directly to the boundary identity $\sum_k \rho_{\bar z_k} \overline{u_{kK}} = 0$ is not valid. Hadamard's lemma resolves this: any smooth function vanishing on $\partial\Omega$ equals $\rho$ times another smooth function, so the boundary identity *extends* to a global equation $\sum_k \rho_{\bar z_k} \overline{u_{kK}} = h_K \rho$. Differentiating the global equation by $\partial_{z_j}$ is now legitimate; the term $(\partial_{z_j} h_K)\rho$ then vanishes when we restrict back to $\partial\Omega$, leaving an honest boundary relation.
The remaining nuisance is the unwanted $h_K \, \partial_{z_j}\rho$ term, which encodes the *normal* component of $\partial_{z_j}$. The boundary condition $\sum_j u_{jK}\, \partial_{z_j} \rho = 0$ is exactly the geometric statement that the vector $(u_{1K}, \dots, u_{nK})$ lies in the complex tangent space (the holomorphic tangent space) of $\partial\Omega$ at each point. Pairing the identity with this tangent vector kills the normal-direction nuisance and leaves the complex Hessian of $\rho$ — that is, the Levi form — paired against the boundary trace of $u$. This is precisely the "contraction condition" mentioned in the strategy: the projection onto the holomorphic tangent space is automatic once the boundary condition is consumed.
[/guided]
[/step]
[step:Assemble the identity]
Substituting the expression for $\Theta$ from Step 6 into the formula
\begin{align*}
\|\bar\partial u\|_{L^2}^2 + \|\bar\partial^* u\|_{L^2}^2 = \sum_{|J|=q}{}' \sum_{j=1}^n \|\partial_{\bar z_j} u_J\|_{L^2}^2 + \Theta
\end{align*}
of Step 3 yields
\begin{align*}
\|\bar\partial u\|_{L^2}^2 + \|\bar\partial^* u\|_{L^2}^2 = \sum_{|J|=q}{}' \sum_{j=1}^n \left\|\frac{\partial u_J}{\partial \bar z_j}\right\|_{L^2}^2 + \sum_{|K|=q-1}{}' \int_{\partial\Omega} \sum_{j,k=1}^n \frac{\partial^2 \rho}{\partial z_j\, \partial \bar z_k}\, u_{jK}\, \overline{u_{kK}}\, dS_\rho,
\end{align*}
which is the Morrey–Kohn–Hörmander formula.
[/step]
Prerequisites (0/3 completed)
Prerequisites Graph
Interactive dependency map showing how this theorem builds on foundational concepts
Loading dependency graph...
Theorem
Definition
Current
Requires
Theorems
Definitions & Concepts
Explore Further
Boundary
Definition
Antisymmetry
Theorem #1468
Divergence Theorem
Theorem #2754
Stokes' Theorem on Manifolds with Boundary
analysis
Nadel Vanishing Theorem
analysis
Coordinate Invariance of Integration of Compactly Supported Top Forms
analysis
Canonical Smooth Structure on the Tangent Bundle
analysis
Serre Duality for Holomorphic Vector Bundles
analysis
Poincaré-Hopf Theorem
analysis
Boundary of a Boundary is Zero
analysis
Hörmander $L^2$ Criterion for Runge Pairs
analysis