Notes on Complex Analysis
Export PDFGitHub

§11.1 Consequences of Holomorphy

11.1 Consequences of Holomorphy

Obviously, we will first formally define the concept of holomorphy in higher dimensions.

Definition 11.1.1.

A function 𝑓:Ω⊆ℂ𝑛→ℂ is holomorphic if it is holomorphic in each variable when the others are held constant.

If we consider 𝑓 to be a function of 𝑧1,𝑧1‾,𝑧2,𝑧2‾,…,𝑧𝑛,𝑧𝑛‾, then 𝑓 is holomorphic iff 𝜕𝑓𝜕𝑧𝑘‾≡0 for all 1≤𝑘≤𝑛 and 𝑓 has all continuous partial derivatives.

Theorem 11.1.2 (Cauchy’s Integral Formula on Polydisks).

Fix 𝒂=(𝑎1,…,𝑎𝑛)∈ℂ𝑛 arbitrarily and suppose 𝑟1,𝑟2,…,𝑟𝑛>0 are the radii of the polydisk defined by Ω=∏𝑘=1𝑛𝐷(𝑎𝑘,𝑟𝑘) (where the product here is the Cartesian product). Suppose 𝑓:Ω‾→ℂ is holomorphic. For fixed 𝑘1,𝑘2,…,𝑘𝑛∈ℤ≥0, we have that

(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒛)=∏𝑗=1𝑛𝑘𝑗!(2πi)𝑛∮𝜕𝐷(𝑎1,𝑟1)⋯∮𝜕𝐷(𝑎𝑛,𝑟𝑛)𝑓(𝜁1,…,𝜁𝑛)d𝜁𝑛⋯d𝜁1∏𝑗=1𝑛(𝜁𝑗−𝑧𝑗)𝑘𝑗+1
for any 𝑧=(𝑧1,𝑧2,…,𝑧𝑛)∈Ω.
Proof.

By Cauchy–Goursat (Theorem 3.2.1Theorem 3.2.1Theorem 3.2.1), we have

𝜕𝑘1𝜕𝑧1𝑘1𝑓(𝒛)=𝑘1!2πi∮𝜕𝐷(𝑎1,𝑟1)𝑓(𝜁1,𝑧2,…,𝑧𝑛)(𝜁1−𝑧1)𝑘1+1d𝜁1

for 𝑧∈Ω, which is holomorphic. Thus, by the same application on 𝜕𝑘1𝜕𝑧1𝑘1𝑓(𝒛), we have

𝜕𝑘2𝜕𝑧2𝑘2𝜕𝑘1𝜕𝑧1𝑘1𝑓(𝒛)=𝑘2!𝑘1!(2πi)2∮𝜕𝐷(𝑎2,𝑟2)∮𝜕𝐷(𝑎1,𝑟1)𝑓(𝜁1,𝜁2,𝑧3,…,𝑧𝑛)d𝜁1d𝜁2(𝜁1−𝑧1)𝑘1+1(𝜁2−𝑧2)𝑘2+1.
By reiterating 𝑛 times and reversing the order of differentiation and integration, the conclusion follows. ⁠□

By the boundedness assumption for 𝑓, we have:

Corollary 11.1.2.1 (Cauchy’s Estimate on Polydisks).

Let 𝒂=(𝑎1,𝑎2,…,𝑎𝑛)∈ℂ𝑛 be fixed and suppose 𝑟1,𝑟2,…,𝑟𝑛>0 are the radii of the polydisk defined by Ω=∏𝑘=1𝑛𝐷(𝑎𝑘,𝑟𝑘) (where the product here is the Cartesian product). Suppose 𝑓:Ω‾→ℂ is holomorphic. For fixed 𝑘1,𝑘2,…,𝑘𝑛∈ℤ≥0, we have that

|(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒛)|≤∏𝑗=1𝑛(𝑘𝑗!𝑟𝑗𝑘𝑗)sup𝜻∈∏𝑗=1𝑛𝜕𝐷(𝑎𝑗,𝑟𝑗)|𝑓(𝜻)|
for any 𝒛=(𝑧1,𝑧2,…,𝑧𝑛)∈Ω.
Proof.

For each 𝑗, let 𝜀𝑗 satisfy 𝐷(𝑧𝑗,𝜀𝑗)⊆𝐷(𝑎𝑗,𝑟𝑗). By Cauchy’s Integral Formula (Theorem 11.1.2Theorem 11.1.2Theorem 11.1.2), we have

|(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒛)|≤∏𝑗=1𝑛𝑘𝑗!(2π)𝑛∮𝜕𝐷(𝑎1,𝑟1)⋯∮𝜕𝐷(𝑎𝑛,𝑟𝑛)|𝑓(𝜁1,…,𝜁𝑛)∏𝑗=1𝑛(𝜁𝑗−𝑧𝑗)𝑘𝑗+1||d𝜁𝑛|⋯|d𝜁1|.

For each 𝑗, let 𝜁𝑗=𝑎𝑗+𝑟𝑗ei𝑡𝑗, and it follows that d𝜁𝑗=i𝑟𝑗ei𝑡𝑗d𝑡𝑗. Because |𝜁𝑗−𝑧𝑗|>𝜀𝑗, we have, after substitution,

|(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒛)|≤∏𝑗=1𝑛(𝑟𝑗𝑘𝑗!)(2π)𝑛∫02π⋯∫02π|𝑓(𝜁1,…,𝜁𝑛)∏𝑗=1𝑛𝜀𝑗𝑘𝑗+1|d𝑡𝑛⋯d𝑡1≤∏𝑗=1𝑛(𝑟𝑗𝑘𝑗!2π𝜀𝑗𝑘𝑗+1)sup𝜻∈∏𝑗=1𝑛𝜕𝐷(𝑎𝑗,𝑟𝑗)|𝑓(𝜻)|∫⋯∫[0,2π]𝑛d𝑡𝑛⋯d𝑡1≤∏𝑗=1𝑛(𝑘𝑗!𝑟𝑗𝑘𝑗)sup𝜻∈∏𝑗=1𝑛𝜕𝐷(𝑎𝑗,𝑟𝑗)|𝑓(𝜻)|,
since 𝜀𝑗≤𝑟𝑗 for all 𝑗. ⁠□

Similar to the univariate case, there are Taylor expansions of holomorphic functions in several complex variables.

Theorem 11.1.3.

Let 𝑓:ℂ𝑛→ℂ be holomorphic on (a neighborhood of) the closure Ω‾ of a polydisk Ω=∏𝑘=1𝑛𝐷(𝑎𝑘,𝑟𝑘) centered at 𝒂=(𝑎1,𝑎2,…,𝑎𝑛)∈ℂ𝑛. Then, for any 𝒛=(𝑧1,𝑧2,…,𝑧𝑛)∈Ω, we have the expansion

𝑓(𝒛)=∑𝑘1=0∞⋯∑𝑘𝑛=0∞𝑎𝑘1,…,𝑘𝑛(𝑧1−𝑎1)𝑘1⋯(𝑧𝑛−𝑎𝑛)𝑘𝑛,(11.1.1)

where ∀𝑘1,…,𝑘𝑛∈ℤ≥0,

𝑎𝑘1,…,𝑘𝑛=1∏𝑗=1𝑛𝑘𝑗!(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒂).
The series converges absolutely and uniformly on Ω.
Proof.

By Theorem 11.1.2Theorem 11.1.2Theorem 11.1.2 we have

𝑓(𝒛)=1(2πi)𝑛∮𝜕𝐷(𝑎1,𝑟1)⋯∮𝜕𝐷(𝑎𝑛,𝑟𝑛)𝑓(𝜻)(𝜁1−𝑧1)⋯(𝜁𝑛−𝑧𝑛)d𝜁𝑛⋯d𝜁1.

For each 𝑗, since |𝑧𝑗−𝑎𝑗|<𝑟𝑗=|𝜁𝑗−𝑎𝑗| on 𝜕𝐷(𝑎𝑗,𝑟𝑗), the geometric series expansion holds:

1𝜁𝑗−𝑧𝑗=1𝜁𝑗−𝑎𝑗⋅11−𝑧𝑗−𝑎𝑗𝜁𝑗−𝑎𝑗=∑𝑘𝑗=0∞(𝑧𝑗−𝑎𝑗)𝑘𝑗(𝜁𝑗−𝑎𝑗)𝑘𝑗+1,

which converges uniformly in 𝜁𝑗 on 𝜕𝐷(𝑎𝑗,𝑟𝑗). Hence, we have

𝑓(𝒛)=1(2πi)𝑛∑𝑘1=0∞∮𝜕𝐷(𝑎1,𝑟1)⋯∮𝜕𝐷(𝑎𝑛,𝑟𝑛)𝑓(𝜻)(𝑧1−𝑎1)𝑘1d𝜁𝑛⋯d𝜁1(𝜁1−𝑎1)𝑘1+1(𝜁2−𝑧2)⋯(𝜁𝑛−𝑧𝑛)

where uniform convergence has allowed the interchange of summation and integration. Reiteration of this process gives

𝑓(𝒛)=∑𝑘1=0∞⋯∑𝑘𝑛=0∞∏𝑗=1𝑛(𝑧𝑗−𝑎𝑗)𝑘𝑗(2πi)𝑛∮𝜕𝐷(𝑎1,𝑟1)⋯∮𝜕𝐷(𝑎𝑛,𝑟𝑛)𝑓(𝜻)d𝜁𝑛⋯d𝜁1∏𝑗=1𝑛(𝜁𝑗−𝑎𝑗)𝑘𝑗+1.

By the Cauchy Integral Formula (Theorem 11.1.2Theorem 11.1.2Theorem 11.1.2), we have

(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒂)=∏𝑗=1𝑛𝑘𝑗!(2πi)𝑛∮𝜕𝐷(𝑎1,𝑟1)⋯∮𝜕𝐷(𝑎𝑛,𝑟𝑛)𝑓(𝜻)d𝜁𝑛⋯d𝜁1∏𝑗=1𝑛(𝜁𝑗−𝑎𝑗)𝑘𝑗+1,

and hence if we let

𝑎𝑘1,…,𝑘𝑛=1∏𝑗=1𝑛𝑘𝑗!(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓(𝒂),

then (11.1.1)(11.1.1)(11.1.1) follows. Cauchy’s Estimate (Corollary 11.1.2.1Corollary 11.1.2.1Corollary 11.1.2.1) gives that

|𝑎𝑘1,…,𝑘𝑛|≤𝑀∏𝑗=1𝑛(1𝜌𝑗𝑘𝑗),

where 𝑀=sup𝜻∈∏𝑗=1𝑛𝜕𝐷(𝑎𝑗,𝑟𝑗)|𝑓(𝜻)| for some 𝜌𝑗>𝑟𝑗 for all 𝑗. Hence,

|∑𝑘1=0∞⋯∑𝑘𝑛=0∞𝑎𝑘1,…,𝑘𝑛∏𝑗=1𝑛(𝑧𝑗−𝑎𝑗)𝑘𝑗|≤∑𝑘1=0∞⋯∑𝑘𝑛=0∞|𝑎𝑘1,…,𝑘𝑛|∏𝑗=1𝑛|𝑧𝑗−𝑎𝑗|𝑘𝑗≤𝑀∑𝑘1=0∞⋯∑𝑘𝑛=0∞∏𝑗=1𝑛|𝑟𝑗𝜌𝑗|𝑘𝑗=𝑀∏𝑗=1𝑛∑𝑘𝑗=0∞|𝑟𝑗𝜌𝑗|𝑘𝑗<∞.

By the Weierstrass 𝑀–Test (Theorem 2.3.3Theorem 2.3.3Theorem 2.3.3), the series converges absolutely and uniformly on Ω. ⁠□

Theorem 11.1.4 (Identity).

Let 𝑓 be a holomorphic function on Ω⊆ℂ𝑛. If the set {𝑧∈Ω:𝑓(𝑧)=0} has an accumulation point in Ω, then 𝑓≡0 on Ω.

Theorem 11.1.5 (Maximum Modulus Principle).

Let Ω⊂ℂ𝑛 be a open bounded region, and suppose that 𝑓:Ω→ℂ is holomorphic. If

𝑀=sup𝜁∈𝜕Ωlim𝑧→𝜁𝑧∈Ω|𝑓(𝑧)|,
then |𝑓(𝑧)|<𝑀 for all 𝑧∈Ω, unless 𝑓 is constant.

Theorem 11.1.6 (Weierstrass).

Suppose that Ω⊆ℂ𝑛 is a region and that {𝑓𝑘}𝑘∈ℕ is a sequence of holomorphic functions Ω→ℂ. If {𝑓𝑘}𝑘∈ℕ converges locally uniformly to 𝑓 on Ω, then 𝑓 is holomorphic on Ω. Moreover, ∀𝑘1,…,𝑘𝑛∈ℤ≥0,

(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓𝑘⇉(∏𝑗=1𝑛𝜕𝑘𝑗𝜕𝑧𝑗𝑘𝑗)𝑓
on compact subsets of Ω.

Theorem 11.1.7 (Montel).

A family ℱ︀ of holomorphic functions on some region Ω⊆ℂ𝑛 is normal iff it is locally uniformly bounded on Ω.