From a potential to a holomorphic function
If is holomorphic, the real and imaginary parts are not arbitrary independent functions. The Cauchy–Riemann equations connect their first derivatives, and consequently both are harmonic. The reverse construction is useful in conformal mapping: starting from a real harmonic potential , can we find so that is holomorphic? Locally the answer is yes. Globally, a loop may prevent the locally constructed values of from agreeing.
Let be a domain and .
Definition 3.5.1
(Harmonic function and harmonic conjugate)
The function is harmonic if on . A real-valued function is a harmonic conjugate of if is holomorphic on .
The relation is not symmetric, and the order of and matters.
Common mistake
(Swapping the roles of and )
If is a conjugate of , it does not follow that is a conjugate of : , which is holomorphic only when is constant on the connected domain. The correct statement is that is a conjugate of , because . Likewise is not a conjugate of , since .
The differential form dictated by Cauchy–Riemann
The equations and force
The form is closed because
Closedness is a local compatibility condition; being the differential of a globally single-valued function is a stronger statement on a domain with holes.
Theorem 3.5.2
(Periods decide global conjugates)
Let be real and harmonic. It has a global harmonic conjugate if and only if
for every piecewise closed curve in . When a conjugate exists, it is unique up to an additive real constant. In particular, a harmonic function on a simply connected domain has a harmonic conjugate.
Proof
(Integrating the forced differential)
If exists, the Cauchy–Riemann equations give , whose closed-curve integrals vanish. Conversely, fix and define
Connected open subsets of the plane are polygonally connected. The closed-curve hypothesis makes this integral independent of the chosen piecewise smooth path. On a small disk around each point, comparing with short horizontal and vertical segments gives and . These derivatives are continuous, so the Cauchy–Riemann equations imply that is holomorphic. If are conjugates, their difference has both first partial derivatives zero and is constant on the connected domain.
On a simply connected domain, is holomorphic by the Cauchy–Riemann equations and . Along any closed curve, . The simply connected-domain form of Cauchy's theorem makes its integral zero, so the imaginary part gives . This is where the topology converts local compatibility into a global conclusion.
The proof has one analytic step and one topological step, and they should be kept apart.
Proof X-ray
Local compatibility versus a global conjugate
The equation says the form forced by Cauchy–Riemann is closed. Integrating it between two points produces a candidate conjugate. Two path choices differ by a closed curve, so vanishing periods are exactly what makes the candidate single-valued. Once the integral is path-independent, short coordinate segments recover the two Cauchy–Riemann equations.
Examples and a period obstruction
Worked example 3.5.3
(A polynomial potential)
Let on . Since and ,
Thus and . Reversing the sign of would give , which is not holomorphic as a function of .
Worked example 3.5.4
(An exponential potential)
For , we have and , so the form is . A conjugate is , and .
A domain with a hole does not by itself obstruct a conjugate; what matters is the value of the period.
Worked example 3.5.5
(A zero period on the punctured plane)
On let . The holomorphic function equals , and . Around the unit circle , so
because has residue at the origin. Every closed curve in gives the period times its winding number about , so all periods vanish and a global conjugate exists: . The same computation for uses , with residue , and gives the period found below.
Counterexample mode
Harmonic does not always mean globally the real part of one holomorphic function
On , the function is harmonic. Its form is
For the positively oriented unit circle , , the integral is . The nonzero period rules out a single-valued global conjugate. On a sufficiently small disk avoiding zero, a conjugate exists; following it once around the origin changes its value by . This is the same obstruction that prevents a global logarithm of on the punctured plane.
Why this helps with maximum principles
Theorem 3.5.6
(Strong maximum principle for harmonic functions)
A nonconstant real harmonic function on a connected domain cannot attain an interior maximum.
Proof
(A local conjugate and the maximum modulus principle)
Suppose attained its maximum at . A small disk about is simply connected, so on that disk for a holomorphic . The function has modulus , which has an interior maximum; by the maximum-modulus principle, is constant on the disk, and hence so is . Let be the maximum and . Continuity makes closed in , and the same disk argument at every point of makes it open. Since is connected and is nonempty, , so is constant, a contradiction.
The local conjugate is enough for this argument; a global conjugate is not required. For a bounded domain and a harmonic function continuous on its closure, compactness and the strong statement imply that the maximum is attained on the boundary. This conclusion needs continuity on the closure, not merely harmonicity in the interior.
Common mistake
(Using the boundary maximum principle on an unbounded domain)
On the half-plane the function is harmonic, continuous on the closure and zero on the boundary line, yet it is not identically zero and has no maximum at all. The boundary version needs a bounded domain, so that the maximum over the compact closure exists.
Practice and solutions
Quick check
(Find a harmonic conjugate of on .)
Integrate in , then use .
Solution · Answer
gives , and forces constant. So , and the resulting function is .
Quick check
(Find a conjugate of that vanishes at .)
Adjust the additive constant.
Solution · Answer
, which is zero at ; every other conjugate differs by a nonzero constant.
Quick check
(Does have a global conjugate on ? Explain without drawing a branch cut.)
Compute one period of the forced form.
Solution · Answer
No. The period of its forced differential form around the unit circle is , whereas an exact differential has zero closed-curve integral.
Quick check
(If is harmonic on a bounded connected domain, continuous on its closure, and zero on the boundary, what is ?)
Apply the maximum principle to and to .
Solution · Answer
The maximum and minimum principles give and throughout the domain; hence .