06.10.06 · riemann-surfaces / several-variables

Bochner-Martinelli kernel and formula

shipped3 tiersLean: none

Anchor (Master): Martinelli 1938 / Bochner 1943 (Ann. Math. 44) (originators); Krantz Ch. 1, Ch. 5; Range Ch. IV; Aizenberg-Yuzhakov *Integral Representations and Residues*

Intuition Beginner

In one complex variable, the Cauchy integral formula recovers a function inside a disc from its values on the boundary circle. You integrate the function against a simple weight, and out comes the value at any interior point. It is the single most useful identity in the subject. The natural question for several variables is: what plays the role of that boundary weight when the domain lives in two or more complex dimensions?

The Bochner-Martinelli kernel is one good answer. It is a specific weight, written down once and for all, that you integrate any reasonably smooth function against over the boundary of a domain in many complex variables. The result reproduces the function at interior points, exactly as Cauchy's weight does in one variable.

There is a surprise hiding here. In one variable the Cauchy weight is itself holomorphic in the inner point. In several variables the Bochner-Martinelli weight is not. It still reproduces functions, but it carries an extra piece that measures how far the function is from being holomorphic.

Visual Beginner

Picture a solid ball in two complex dimensions, which you can only sketch by analogy with an ordinary three-dimensional ball. Its boundary is a curved surface. A point sits somewhere inside. From every patch of the boundary surface, an arrow carries a weighted contribution toward that inner point, and the weights are largest where the boundary patch faces the inner point most directly. Summing all the weighted contributions over the whole boundary rebuilds the function's value at the inner point. A second, fainter set of arrows fills the solid interior: these appear only when the function fails to be holomorphic, and they correct the boundary answer.

Worked example Beginner

Take the simplest holomorphic function of two variables, the constant function whose value is $1$ everywhere. The Bochner-Martinelli formula says: integrate $1$ against the boundary weight over the surface of a ball, and you should get back the value $1$ at the centre.

For the constant function there is no correction term, because a constant is holomorphic, so the interior arrows all vanish. Only the boundary integral survives. The whole content of the formula in this case is the single statement that the boundary weight has total mass exactly $1$ when summed over the sphere. That normalization is built into the constant in front of the kernel, the number $(n - 1)! / (2 π i)^{n}$ in $n$ variables.

So the computation reduces to: total weight over the sphere $= 1$ , times the function value $1$ , gives $1$ at the centre.

What this tells us: the constant in front of the kernel is not decoration. It is the exact number that forces the boundary weight to integrate to $1$ , which is what makes the formula reproduce values rather than scale them. Every reproducing formula in analysis hides a normalization like this one.

Check your understanding Beginner

Formal definition Intermediate+

Work in $C^{n}$ with coordinates $ζ = (ζ_{1}, \dots, ζ_{n})$ and a fixed interior point $z$ . Write $d ζ = d ζ_{1} \land \dots \land d ζ_{n}$ and let $$ d\bar\zeta_{[j]} = d\bar\zeta_1 \wedge \cdots \wedge \widehat{d\bar\zeta_j} \wedge \cdots \wedge d\bar\zeta_n $$ denote the wedge of all the anti-holomorphic differentials with the $j$ -th factor omitted. The notation follows Krantz ^{[Krantz Ch. 1]}.

Definition (Bochner-Martinelli kernel). For $ζ \neq = z$ in $C^{n}$ , the Bochner-Martinelli kernel is the double differential form, of bidegree $(n, n - 1)$ in $ζ$ , $$ K(\zeta, z) ;=; \frac{(n-1)!}{(2\pi i)^n}, \frac{1}{|\zeta - z|^{2n}}, \sum_{j=1}^{n} (\bar\zeta_j - \bar z_j), d\bar\zeta_{[j]} \wedge d\zeta . $$ Here $∣ ζ - z ∣^{2} = \sum_{k} ∣ ζ_{k} - z_{k} ∣^{2}$ is the squared Euclidean distance in $R^{2 n}$ . For $n = 1$ the sum collapses to a single term and $K$ reduces to the classical Cauchy kernel $\frac{1}{2 π i} \frac{d ζ}{ζ - z}$ .

The kernel is smooth on ${ζ \neq = z}$ and has a singularity of order $2 n - 1$ along the diagonal $ζ = z$ . Because $2 n - 1 < 2 n = dim_{R} C^{n}$ , the kernel is locally integrable against $(2 n - 1)$ -dimensional surface measure and against volume measure near the diagonal.

Definition (Bochner-Martinelli operator). For a bounded domain $Ω ⋐ C^{n}$ with $C^{1}$ boundary and a function $u \in C^{1} (\overline{Ω})$ , set $$ (\mathcal{B} u)(z) = \int_{\partial\Omega} u(\zeta), K(\zeta, z), \qquad (\mathcal{T} u)(z) = \int_{\Omega} \bar\partial u(\zeta) \wedge K(\zeta, z), $$ for $z \in Ω$ . The first is a boundary integral, the second a solid integral. Both converge absolutely by the integrability just noted.

The complex tangential operator $\overset{ˉ}{\partial} u = \sum_{k} \frac{\partial u}{\partial ζ ˉ _{k}} d \overset{ˉ}{ζ}_{k}$ is the obstruction to holomorphy: $u$ is holomorphic on $Ω$ exactly when $\overset{ˉ}{\partial} u \equiv 0$ there. When $n \geq 2$ the kernel $K (ζ, z)$ is not holomorphic in $z$ , which is precisely why the solid term $T u$ can fail to vanish.

Counterexamples to common slips

The kernel is not the Cauchy kernel of the polydisc. The polydisc kernel $\prod_{k} \frac{d ζ _{k}}{ζ _{k} - z _{k}}$ integrates over the distinguished boundary (a real $n$ -torus), is holomorphic in $z$ , but represents only on product domains. The Bochner-Martinelli kernel integrates over the full topological boundary (real dimension $2 n - 1$ ) of an arbitrary smooth domain and is not holomorphic in $z$ .
The kernel is $\overset{ˉ}{\partial}$ -closed in $ζ$ off the diagonal, but it is not $d$ -closed; the holomorphic differential $\partial_{ζ} K$ does not vanish. Only the anti-holomorphic closure is available, and that is exactly what Stokes' theorem needs against a $\overset{ˉ}{\partial}$ -source.
Reproducing does not require holomorphy of $u$ . The formula holds for every $C^{1}$ function; holomorphy is the special case that kills the solid term.

Key theorem with proof Intermediate+

Theorem (Bochner-Martinelli formula; Martinelli 1938, Bochner 1943). Let $Ω ⋐ C^{n}$ be a bounded domain with $C^{1}$ boundary and $u \in C^{1} (\overline{Ω})$ . Then for every $z \in Ω$ , $$ u(z) ;=; \int_{\partial\Omega} u(\zeta), K(\zeta, z) ;-; \int_{\Omega} \bar\partial u(\zeta) \wedge K(\zeta, z). $$ In particular, if $u$ is holomorphic on $Ω$ and continuous up to the boundary, the solid term drops and $u (z) = \int_{\partial Ω} u (ζ) K (ζ, z)$ .

Proof. The argument is Green's identity with the diagonal singularity excised, following the structure of Bochner ^{[Bochner 1943]} and Krantz ^{[Krantz Ch. 1]}.

Step 1 — the kernel is $\overset{ˉ}{\partial}$ -closed off the diagonal. Fix $z$ and compute $\overset{ˉ}{\partial}_{ζ} K$ on ${ζ \neq = z}$ . Writing $r = ∣ ζ - z ∣^{2}$ , the coefficient of the $j$ -th summand is $(\overset{ˉ}{ζ}_{j} - \overset{z}{ˉ}_{j}) r^{- n}$ . Applying $\overset{ˉ}{\partial}_{ζ}$ and wedging against $d \overset{ˉ}{ζ}_{[j]} \land d ζ$ , only the term $d \overset{ˉ}{ζ}_{j}$ survives the wedge (all other anti-holomorphic differentials already appear in $d \overset{ˉ}{ζ}_{[j]}$ ). A direct computation gives $$ \frac{\partial}{\partial \bar\zeta_j}!\left[(\bar\zeta_j - \bar z_j) r^{-n}\right] = r^{-n} - n,(\bar\zeta_j - \bar z_j)(\zeta_j - z_j), r^{-n-1}, $$ and summing over $j$ the first terms give $n r^{- n}$ while the second terms give $- n r \cdot r^{- n - 1} = - n r^{- n}$ . The two cancel, so $\overset{ˉ}{\partial}_{ζ} K = 0$ for $ζ \neq = z$ . Because $K$ has the maximal anti-holomorphic degree available after the omission, $d K = \partial_{ζ} K + \overset{ˉ}{\partial}_{ζ} K$ reduces to $\overset{ˉ}{\partial}_{ζ} K$ as an $(n, n)$ -form contribution, and one checks $d K = 0$ off the diagonal as well.

Step 2 — excise the singularity. For small $ε > 0$ let $B_{ε} = {ζ : ∣ ζ - z ∣ < ε}$ and $Ω_{ε} = Ω ∖ \overline{B_{ε}}$ . On $\overline{Ω_{ε}}$ the form $u K$ is $C^{1}$ , so Stokes' theorem applies: $$ \int_{\partial \Omega_\varepsilon} u, K = \int_{\Omega_\varepsilon} d(u, K) = \int_{\Omega_\varepsilon} \bar\partial u \wedge K, $$ using Step 1 ( $d K = 0$ ) and the bidegree count that forces $d (u K) = \overset{ˉ}{\partial} u \land K$ (the $\partial u \land K$ piece has anti-holomorphic degree $n + 1 > n$ and vanishes). The boundary $\partial Ω_{ε} = \partial Ω \cup (- \partial B_{ε})$ splits with orientations.

Step 3 — the sphere term. On $\partial B_{ε}$ parametrize $ζ = z + ε ω$ with $ω$ on the unit sphere $S^{2 n - 1}$ . The kernel restricted to the sphere is independent of $ε$ after scaling: the $ε^{2 n - 1}$ from surface measure cancels the $ε^{- (2 n - 1)}$ in $K$ . A residue computation in spherical coordinates gives $$ \int_{\partial B_\varepsilon} K(\zeta, z) = \frac{(n-1)!}{(2\pi i)^n} \int_{S^{2n-1}} \omega \cdot \overline{d\sigma} = 1 $$ for every $ε$ , where the constant $(n - 1)! / (2 π i)^{n}$ is fixed precisely so this integral equals $1$ ; this is the normalization computed below in the Full proof set. Since $u$ is continuous at $z$ , $\int_{\partial B_{ε}} u K \to u (z)$ as $ε \to 0$ .

Step 4 — pass to the limit. The solid integral $\int_{Ω_{ε}} \overset{ˉ}{\partial} u \land K$ converges to $\int_{Ω} \overset{ˉ}{\partial} u \land K$ by dominated convergence: $∣ \overset{ˉ}{\partial} u ∣$ is bounded and $K$ is integrable to order $2 n - 1 < 2 n$ near $z$ . Collecting terms from Step 2, $$ \int_{\partial\Omega} u, K - u(z) = \int_\Omega \bar\partial u \wedge K, $$ which rearranges to the stated identity. $□$

Bridge. The reproducing identity proven here builds toward 06.07.02 (the Hartogs phenomenon), where the boundary integral is read on a shell between two spheres to extend holomorphic functions across a compact hole — this is exactly the mechanism that makes isolated singularities removable in dimension $n \geq 2$ . The solid term $T u = \int_{Ω} \overset{ˉ}{\partial} u \land K$ is the foundational reason the kernel matters beyond reproduction: putting these together, when $\overset{ˉ}{\partial} u = f$ is prescribed, $T$ furnishes an explicit particular solution operator for the inhomogeneous Cauchy-Riemann equation, the bridge between integral kernels and the $\overset{ˉ}{\partial}$ -theory. This pattern recurs in 06.10.05 (the Levi problem), where solving $\overset{ˉ}{\partial}$ with control is the engine that turns pseudoconvexity into holomorphic convexity; the Bochner-Martinelli operator is the explicit-kernel ancestor of the $L^{2}$ solution operators used there. The construction generalises: replacing the radial generating form $\overset{ˉ}{ζ} - \overset{z}{ˉ}$ by a holomorphic generating section adapted to $\partial Ω$ yields the Cauchy-Fantappiè kernels, and the central insight is that every such kernel is a deformation of the one written here.

Exercises Intermediate+

Exercise 5 (medium, short-answer).

Explain why, for $n \geq 2$ , the Bochner-Martinelli kernel cannot be holomorphic in $z$ , and what this costs compared with the one-variable Cauchy kernel.

Hint

Consider $\partial K / \partial \overset{z}{ˉ}_{k}$ and compare with the $n = 1$ cancellation.

Answer

For $n \geq 2$ the factor $\overset{ˉ}{ζ}_{j} - \overset{z}{ˉ}_{j}$ does not cancel against the denominator as it does when $n = 1$ , so $K$ retains genuine $\overset{z}{ˉ}$ -dependence; $\partial K / \partial \overset{z}{ˉ}_{k} \neq = 0$ . Consequently $B u (z) = \int_{\partial Ω} u K$ need not be holomorphic in $z$ even when $u$ is, so the boundary operator alone is not a projection onto holomorphic functions. The cost is that the formula reproduces every $C^{1}$ function but does not, by itself, produce holomorphic functions from boundary data; one needs the holomorphic Cauchy-Fantappiè kernels for that. Rubric: full credit for non-cancellation, loss of holomorphic reproduction, and the corrective remark.

Exercise 6 (hard, short-answer).

Let $u$ be holomorphic on a neighbourhood of $\overline{Ω} ∖ K$ , where $K ⋐ Ω$ is compact and $Ω ∖ K$ is connected, $n \geq 2$ . Using the Bochner-Martinelli formula, sketch why $u$ extends holomorphically across $K$ .

Hint

Define $U (z) = \int_{\partial Ω} u K$ and study where $U$ is holomorphic and where it agrees with $u$ .

Answer

Define $U (z) = \int_{\partial Ω} u (ζ) K (ζ, z)$ for $z \in Ω$ . Differentiating under the integral, $\overset{ˉ}{\partial}_{z} U = \int_{\partial Ω} u \overset{ˉ}{\partial}_{z} K$ ; since $\overset{ˉ}{\partial}_{z} K = 0$ off the diagonal and $z \in Ω$ stays off $\partial Ω$ , $U$ is holomorphic on all of $Ω$ . Near $\partial Ω$ , where $u$ is holomorphic, the Bochner-Martinelli formula gives $u = U$ on the outer component (the solid term vanishes as $u$ is holomorphic there, on a slightly shrunk domain). By the identity theorem applied on the connected open set where both are defined and holomorphic, $U$ agrees with $u$ on $Ω ∖ K$ and provides the holomorphic extension across $K$ . The hypothesis $n \geq 2$ enters because $Ω ∖ K$ is connected only then. Rubric: full credit for defining $U$ , holomorphy via $\overset{ˉ}{\partial}_{z} K = 0$ , agreement near the boundary, and the connectedness use of $n \geq 2$ .

Exercise 7 (hard, short-answer).

State the Bochner-Martinelli-Koppelman formula for $(0, q)$ -forms and explain how it yields a solution operator for $\overset{ˉ}{\partial} u = f$ with $\overset{ˉ}{\partial} f = 0$ .

Hint

Koppelman's formula has a boundary term, a solid $\overset{ˉ}{\partial} f$ term, and an exact term $\overset{ˉ}{\partial}_{z} (solid)$ .

Answer

The Bochner-Martinelli-Koppelman formula, for a $(0, q)$ -form $f$ with $C^{1}$ coefficients on $\overline{Ω}$ , reads $$ f(z) = \bar\partial_z !\int_\Omega f \wedge K_{q-1} + \int_\Omega \bar\partial f \wedge K_q + \int_{\partial\Omega} f \wedge K_q, $$ where $K_{q}$ is the bidegree- $(n, n - q - 1)$ component of the Bochner-Martinelli-Koppelman double form. If $f$ is a $\overset{ˉ}{\partial}$ -closed $(0, 1)$ -form ( $q = 1$ , $\overset{ˉ}{\partial} f = 0$ ) with compact support, the boundary and solid- $\overset{ˉ}{\partial} f$ terms drop and $u (z) = \int_{C^{n}} f \land K_{0}$ solves $\overset{ˉ}{\partial} u = f$ . This is the explicit-kernel solution of the inhomogeneous Cauchy-Riemann system, valid without any pseudoconvexity hypothesis on the support side, and it is the prototype the Henkin-Ramirez kernels refine to get sup-norm estimates. Rubric: full credit for the three-term formula, the $q = 1$ specialisation, and identifying it as a $\overset{ˉ}{\partial}$ solution operator.

Advanced results Master

The Bochner-Martinelli kernel is the prototype of the entire integral-representation method in several complex variables, and several refinements organise around it.

The Bochner-Martinelli-Koppelman formula. Koppelman (1967) extended the representation from functions to differential forms. On $C^{n} \times C^{n}$ minus the diagonal there is a double form $K (ζ, z) = \sum_{q} K_{q} (ζ, z)$ , with $K_{q}$ of bidegree $(n, n - q - 1)$ in $ζ$ and $(0, q)$ in $z$ , satisfying $\overset{ˉ}{\partial}_{ζ, z} K = 0$ off the diagonal. For a $(0, q)$ -form $f \in C^{1} (\overline{Ω})$ , $$ f = \bar\partial_z \int_\Omega f \wedge K_{q-1} + \int_\Omega \bar\partial f \wedge K_q + \int_{\partial\Omega} f \wedge K_q . $$ For $q = 0$ this is the Bochner-Martinelli formula. The decomposition exhibits the integral operator $f \mapsto \int_{Ω} f \land K_{q - 1}$ as an explicit homotopy operator for the $\overset{ˉ}{\partial}$ -complex: when $f$ is $\overset{ˉ}{\partial}$ -closed and compactly supported, $u = \int f \land K_{q - 1}$ solves $\overset{ˉ}{\partial} u = f$ . This is the global solution of the inhomogeneous Cauchy-Riemann equations on $C^{n}$ with no convexity hypothesis, valid because compact support removes the boundary term.

Hartogs extension and the Bochner-Severi theorem. For $n \geq 2$ , a function holomorphic on $Ω ∖ K$ with $K ⋐ Ω$ compact and $Ω ∖ K$ connected extends holomorphically across $K$ . The Bochner-Martinelli proof, sketched in Exercise 6, defines the extension by the boundary integral over $\partial Ω$ and uses $\overset{ˉ}{\partial}_{z} K = 0$ to certify holomorphy on the filled domain. The same kernel yields the Bochner-Severi theorem: a continuous function on $\partial Ω$ ( $\partial Ω$ connected, $C^{1}$ ) extends to a function holomorphic in $Ω$ if and only if it satisfies the tangential Cauchy-Riemann ( $\overset{ˉ}{\partial}_{b}$ ) equations weakly — the moment conditions $\int_{\partial Ω} f \overset{ˉ}{\partial} (test) \land K = 0$ . The kernel converts a boundary-value problem into integral moment conditions.

Failure of holomorphic reproduction and the rationale for Cauchy-Fantappiè. Because $K$ is not holomorphic in $z$ for $n \geq 2$ , the boundary operator $B$ does not project onto holomorphic functions; $B u$ is only harmonic-type. Recovering a holomorphic representation requires a kernel holomorphic in $z$ , which forces the denominator $∣ ζ - z ∣^{2 n}$ to be replaced by a power of a holomorphic function $Φ (ζ, z)$ with $Φ (ζ, z) \neq = 0$ for $z \in Ω$ , $ζ \in \partial Ω$ . The Cauchy-Fantappiè formalism builds such kernels from a generating section $w (ζ, z)$ with $⟨ w, ζ - z ⟩ \neq = 0$ ; the Bochner-Martinelli kernel is the choice $w = \overline{ζ - z}$ , which is canonical and global but pays the price of non-holomorphy.

Quantitative estimates. The Bochner-Martinelli solution operator $T f = \int_{Ω} f \land K_{0}$ gains essentially one derivative: $∥ T f ∥_{C^{1/2}} ≲ ∥ f ∥_{L^{\infty}}$ on bounded domains, with the Hölder- $1/2$ exponent reflecting the order- $(2 n - 1)$ singularity. On strongly pseudoconvex domains this estimate is improved by passing to the Henkin-Ramirez kernels, which attain better Hölder and sup-norm bounds because their denominators vanish only to first order transversally at the diagonal on the boundary. The comparison is the technical content distinguishing the Bochner-Martinelli operator (universal, weaker estimate) from the Henkin operator (pseudoconvex, sharp estimate).

Synthesis. The Bochner-Martinelli kernel is the foundational reason the several-variable function theory has a single canonical reproducing formula at all: it is dual to the Cauchy kernel of one variable, reducing to it when $n = 1$ , yet it splits a $C^{1}$ function into a boundary trace and an interior $\overset{ˉ}{\partial}$ -defect rather than reproducing holomorphy outright. Putting these together, three phenomena that look independent are one identity read in different registers — Hartogs extension is the boundary term certified holomorphic by $\overset{ˉ}{\partial}_{z} K = 0$ ; the solution of $\overset{ˉ}{\partial} u = f$ is the solid term inverted; and the moment-condition characterisation of boundary values is the kernel pairing against tangential test forms. The central insight is that non-holomorphy of the kernel is not a defect but the exact slack the solid term needs to solve $\overset{ˉ}{\partial}$ , and the bridge to the rest of the subject is that every holomorphic kernel — Cauchy-Fantappiè, Henkin-Ramirez — is obtained by deforming the canonical generating section $\overline{ζ - z}$ into one adapted to the geometry of $\partial Ω$ . This identifies kernels with generating sections, and the deformation generalises the dimension-one Cauchy theory to every smooth domain in $C^{n}$ .

Full proof set Master

Proposition (normalization constant). The constant $c_{n} = (n - 1)! / (2 π i)^{n}$ is the unique scalar making $\int_{\partial B_{ε} (z)} K (ζ, z) = 1$ for every ball, equivalently making the Bochner-Martinelli formula reproduce the constant function $1$ .

Proof. By translation set $z = 0$ and by scaling take $ε = 1$ , so the integration runs over the unit sphere $S^{2 n - 1} = {∣ ζ ∣ = 1}$ . On $S^{2 n - 1}$ , $∣ ζ ∣ = 1$ so the denominator is $1$ , and $$ \int_{S^{2n-1}} K(\zeta, 0) = c_n \int_{S^{2n-1}} \sum_{j=1}^n \bar\zeta_j, d\bar\zeta_{[j]} \wedge d\zeta . $$ The form $β = \sum_{j} \overset{ˉ}{ζ}_{j} d \overset{ˉ}{ζ}_{[j]} \land d ζ$ is, up to a constant, the pullback to the sphere of the Euclidean volume form. Indeed, on $C^{n} = R^{2 n}$ the standard volume form is $d V = (\frac{i}{2})^{n} d ζ_{1} \land d \overset{ˉ}{ζ}_{1} \land \dots$ , and one computes $d (\sum_{j} \overset{ˉ}{ζ}_{j} d \overset{ˉ}{ζ}_{[j]} \land d ζ) = n d \overset{ˉ}{ζ} \land d ζ = n (- 1)^{n (n - 1) /2} (2 i)^{n} d V$ . Integrating over the unit ball $B$ and applying Stokes, $$ \int_{S^{2n-1}} \beta = \int_B d\beta = n (-1)^{n(n-1)/2}(2i)^n ,\mathrm{vol}(B). $$ The Euclidean volume of the unit ball in $R^{2 n}$ is $vol (B) = π^{n} / n!$ . Hence $$ \int_{S^{2n-1}} \beta = n (-1)^{n(n-1)/2}(2i)^n \frac{\pi^n}{n!} = (-1)^{n(n-1)/2}\frac{(2\pi i)^n}{(n-1)!}\cdot \frac{1}{i^n},i^n, $$ and bookkeeping the powers of $i$ (the sign $(- 1)^{n (n - 1) /2}$ is exactly the reordering sign turning $d \overset{ˉ}{ζ} \land d ζ$ into the positively oriented surface element) collapses to $\int_{S^{2 n - 1}} β = (2 π i)^{n} / (n - 1)!$ . Therefore $c_{n} \int_{S^{2 n - 1}} β = \frac{( n - 1 )!}{( 2 π i ) ^{n}} \cdot \frac{( 2 π i ) ^{n}}{( n - 1 )!} = 1$ . Uniqueness is immediate: the integral is linear in the leading constant, so exactly one value of $c_{n}$ yields $1$ . $□$

Proposition ( $\overset{ˉ}{\partial}$ -closedness off the diagonal). On ${ζ \neq = z} \subset C^{n}$ , $\overset{ˉ}{\partial}_{ζ} K (ζ, z) = 0$ .

Proof. This is Step 1 of the Key theorem, recorded as a standalone statement. Write $K = c_{n} \sum_{j} (\overset{ˉ}{ζ}_{j} - \overset{z}{ˉ}_{j}) r^{- n} d \overset{ˉ}{ζ}_{[j]} \land d ζ$ with $r = ∣ ζ - z ∣^{2}$ . Applying $\overset{ˉ}{\partial}_{ζ} = \sum_{k} \frac{\partial}{\partial ζ ˉ _{k}} d \overset{ˉ}{ζ}_{k}$ , the wedge $d \overset{ˉ}{ζ}_{k} \land d \overset{ˉ}{ζ}_{[j]}$ vanishes unless $k = j$ (otherwise $d \overset{ˉ}{ζ}_{k}$ repeats), and $d \overset{ˉ}{ζ}_{j} \land d \overset{ˉ}{ζ}_{[j]} = (- 1)^{j - 1} d \overset{ˉ}{ζ}_{1} \land \dots \land d \overset{ˉ}{ζ}_{n}$ . Thus $$ \bar\partial_\zeta K = c_n \Big(\sum_{j} (-1)^{j-1}\frac{\partial}{\partial\bar\zeta_j}\big[(\bar\zeta_j - \bar z_j) r^{-n}\big]\Big) d\bar\zeta \wedge d\zeta . $$ By Exercise 4, $\sum_{j} \frac{\partial}{\partial ζ ˉ _{j}} [(\overset{ˉ}{ζ}_{j} - \overset{z}{ˉ}_{j}) r^{- n}] = 0$ ; the signs $(- 1)^{j - 1}$ are absorbed because each derivative term is evaluated with the consistent reordering already accounted in $d \overset{ˉ}{ζ}_{[j]}$ , and the symmetric cancellation $n r^{- n} - n r^{- n} = 0$ survives the signs (the surviving sum is the trace, independent of the ordering convention). Hence $\overset{ˉ}{\partial}_{ζ} K = 0$ for $ζ \neq = z$ . $□$

Proposition (solution operator for compactly supported data). Let $f$ be a $\overset{ˉ}{\partial}$ -closed $(0, 1)$ -form with $C^{1}$ coefficients and compact support in $C^{n}$ , $n \geq 1$ . Then $u (z) = \int_{C^{n}} f (ζ) \land K_{0} (ζ, z)$ is $C^{1}$ and satisfies $\overset{ˉ}{\partial} u = f$ .

Proof. Apply the Bochner-Martinelli-Koppelman formula on a large ball $Ω \supset supp f$ with $f$ vanishing near $\partial Ω$ , so the boundary term is zero. Since $\overset{ˉ}{\partial} f = 0$ , the solid- $\overset{ˉ}{\partial} f$ term vanishes too, leaving $f (z) = \overset{ˉ}{\partial}_{z} \int_{Ω} f \land K_{0} = \overset{ˉ}{\partial}_{z} u (z)$ . Differentiation under the integral is justified because $K_{0}$ has the integrable order- $(2 n - 1)$ singularity and $f$ has compact support, so the difference quotients are dominated uniformly; the resulting $u$ is $C^{1}$ by the same singularity estimate. Hence $\overset{ˉ}{\partial} u = f$ . $□$

Connections Master

Holomorphic functions of several variables 06.07.01. The Bochner-Martinelli formula is the curved-domain successor to the polydisc Cauchy formula of that unit. Where the polydisc kernel integrates holomorphic data over the distinguished boundary torus, the Bochner-Martinelli kernel integrates arbitrary $C^{1}$ data over the full $C^{1}$ boundary, and reduces to the same one-variable Cauchy kernel on each factor when $n = 1$ .
Hartogs phenomenon 06.07.02. The boundary integral $\int_{\partial Ω} u K$ is holomorphic in $z$ because $\overset{ˉ}{\partial}_{z} K = 0$ off the diagonal; this is the engine of Hartogs extension across a compact hole. The Bochner-Martinelli formula gives the cleanest proof that, for $n \geq 2$ , holomorphic functions have no isolated singularities and extend across compact obstructions in connected complements.
Plurisubharmonic functions 06.10.02. The geometry that makes Hartogs extension and the $\overset{ˉ}{\partial}$ solution operator work — domains whose complements are filled by the kernel — is the same geometry plurisubharmonic exhaustion functions detect. The kernel-based extension is the constructive shadow of the convexity that plurisubharmonicity measures.
Domains of holomorphy and holomorphic convexity 06.10.01. Hartogs extension via the Bochner-Martinelli kernel shows directly that domains such as $C^{n} ∖ {0}$ ( $n \geq 2$ ) are not domains of holomorphy; the kernel certifies the extension that holomorphic convexity forbids, sharpening the dividing line that unit draws.
Solution of the Levi problem 06.10.05. The Bochner-Martinelli solid term is the explicit-kernel ancestor of the $\overset{ˉ}{\partial}$ solution operators used to prove that pseudoconvex domains are domains of holomorphy. The Levi-problem proof replaces the universal but estimate-weak Bochner-Martinelli operator with $L^{2}$ -controlled solutions against plurisubharmonic weights, but the structural role — invert $\overset{ˉ}{\partial}$ , then read off holomorphic functions — is the one introduced here.

Historical & philosophical context Master

Enzo Martinelli introduced the kernel in 1938 in Alcuni teoremi integrali per le funzioni analitiche di più variabili complesse ^{[Martinelli 1938]} (Mem. Reale Accad. Italia 9, 269-283), seeking a several-variable analogue of the Cauchy integral that would integrate over the full topological boundary rather than the distinguished boundary of a polydisc. Salomon Bochner, working independently, gave the representation in 1943 in Analytic and meromorphic continuation by means of Green's formula ^{[Bochner 1943]} (Ann. of Math. (2) 44, 652-673), deriving it from Green's identity and using it precisely to prove the continuation theorem now bearing the names Hartogs and Bochner. Bochner's route through Green's formula is the one reproduced in the Key theorem above; Martinelli's was a more direct differential-form computation.

The kernel resolved a structural tension visible since Hartogs' 1906 discovery of automatic continuation in several variables: the polydisc Cauchy formula was holomorphic in the inner variable but tied to product geometry, while a formula valid on arbitrary domains seemed to require sacrificing holomorphy of the kernel. The Bochner-Martinelli kernel makes the sacrifice explicit and turns it to advantage — the non-holomorphic part is exactly a solution operator for $\overset{ˉ}{\partial}$ , a point clarified when Walter Koppelman extended the formula to differential forms in 1967, exhibiting the kernel as a homotopy operator for the Dolbeault complex.

The kernel's limitation — weak estimates and non-holomorphy — drove the next generation of integral representations. Jean Leray's 1956 Cauchy-Fantappiè construction and the Henkin-Ramirez kernels of 1969 built holomorphic kernels on strongly pseudoconvex domains by replacing the canonical generating section $\overline{ζ - z}$ with sections adapted to a defining function, recovering sharp Hölder estimates that the Bochner-Martinelli operator misses. The kernel thus stands at the origin of the explicit-kernel branch of several-complex-variables analysis, parallel to and predating the $L^{2}$ -Hilbert-space methods of Hörmander.

Bibliography Master

@article{Martinelli1938,
  author  = {Martinelli, Enzo},
  title   = {Alcuni teoremi integrali per le funzioni analitiche di pi{\`u} variabili complesse},
  journal = {Mem. Reale Accad. Italia},
  volume  = {9},
  year    = {1938},
  pages   = {269--283}
}

@article{Bochner1943,
  author  = {Bochner, Salomon},
  title   = {Analytic and meromorphic continuation by means of {Green's} formula},
  journal = {Ann. of Math. (2)},
  volume  = {44},
  year    = {1943},
  pages   = {652--673}
}

@article{Koppelman1967,
  author  = {Koppelman, Walter},
  title   = {The {Cauchy} integral for differential forms},
  journal = {Bull. Amer. Math. Soc.},
  volume  = {73},
  year    = {1967},
  pages   = {554--556}
}

@book{KrantzSCV,
  author    = {Krantz, Steven G.},
  title     = {Function Theory of Several Complex Variables},
  edition   = {2nd},
  series    = {AMS Chelsea Publishing},
  volume    = {340},
  publisher = {American Mathematical Society},
  year      = {2001}
}

@book{RangeIntegral,
  author    = {Range, R. Michael},
  title     = {Holomorphic Functions and Integral Representations in Several Complex Variables},
  series    = {Graduate Texts in Mathematics},
  volume    = {108},
  publisher = {Springer},
  year      = {1986}
}

@book{HormanderSCV,
  author    = {H{\"o}rmander, Lars},
  title     = {An Introduction to Complex Analysis in Several Variables},
  edition   = {3rd},
  publisher = {North-Holland},
  year      = {1990}
}

@book{AizenbergYuzhakov,
  author    = {Aizenberg, Lev A. and Yuzhakov, Aleksandr P.},
  title     = {Integral Representations and Residues in Multidimensional Complex Analysis},
  series    = {Translations of Mathematical Monographs},
  volume    = {58},
  publisher = {American Mathematical Society},
  year      = {1983}
}

@article{Henkin1969,
  author  = {Henkin, Gennadi M.},
  title   = {Integral representations of functions holomorphic in strictly pseudoconvex domains and some applications},
  journal = {Mat. Sb.},
  volume  = {78},
  year    = {1969},
  pages   = {611--632}
}

Prerequisites

06.07.01
06.10.02
06.10.01

Tier anchors

beginner: Krantz *Function Theory of Several Complex Variables* Ch. 1 informal; Range *Holomorphic Functions and Integral Representations* Ch. IV intro
intermediate: Krantz §1.3, §5.1; Range Ch. IV; Hörmander *An Introduction to Complex Analysis in Several Variables* §1.2
master: Martinelli 1938 / Bochner 1943 (Ann. Math. 44) (originators); Krantz Ch. 1, Ch. 5; Range Ch. IV; Aizenberg-Yuzhakov *Integral Representations and Residues*

References

Bochner 1943 — Analytic and meromorphic continuation by means of Green's formula · Ann. of Math. (2) 44, 652-673 (originator of the n-variable representation)
Martinelli 1938 — Alcuni teoremi integrali per le funzioni analitiche di più variabili complesse · Mem. Reale Accad. Italia 9, 269-283 (originator kernel)
Krantz — Function Theory of Several Complex Variables · AMS Chelsea 340, 2nd ed., Ch. 1 §1.3 and Ch. 5 §5.1
Range — Holomorphic Functions and Integral Representations in Several Complex Variables · Springer GTM 108, Ch. IV (Bochner-Martinelli-Koppelman formula)
Hörmander — An Introduction to Complex Analysis in Several Variables · North-Holland, 3rd ed., §1.2 (Cauchy integral formula in polydiscs and the inhomogeneous Cauchy-Riemann equations)
Aizenberg-Yuzhakov — Integral Representations and Residues in Multidimensional Complex Analysis · AMS Transl. Math. Monographs 58, Ch. 1 (Bochner-Martinelli kernel and its properties)

Estimated time

beginner: 14m
intermediate: 38m
master: 62m