Averaging, now for the real part alone
In the Cauchy-formula rung you met the mean-value property for a holomorphic f: stand at the center z_0 of a circle and the value f(z_0) is exactly the average of f around the rim. The first two guides of this rung taught you that the real part u of a holomorphic f = u + i v is a harmonic function — a solution of Laplace's equation u_xx + u_yy = 0 — and that, given such a u, you can build its harmonic partner v. The natural next question is whether the averaging law survives when we throw away v and keep only the real, physical quantity u. It does, perfectly, and that survival is the engine of everything in this guide.
To see it, take the holomorphic mean-value identity f(z_0) = (1/(2 pi)) times the integral over theta from 0 to 2 pi of f(z_0 + r e^(i theta)), and simply read off real parts. The real part of an average is the average of the real parts, so u(z_0) equals the average of u around the circle. The same statement named for and credited to Gauss is Gauss's mean-value theorem, the harmonic mean-value property: the value of a harmonic function at the center of a disk is the average of its boundary values. Crucially, this does not secretly need a holomorphic partner — any function satisfying Laplace's equation obeys it, whether or not it is anyone's real part.
Gauss's mean-value theorem (u harmonic on and inside the circle of radius r about z_0):
u(z_0) = (1 / (2 pi)) * integral_0^{2 pi} u(z_0 + r e^(i theta)) d theta
The center value is the plain average of the ring values -- for EVERY radius r that fits.
There is a solid-disk version too: u(z_0) is also the area-average over the whole disk.Why a steady plate has no interior hot spot
Here is the physical picture that makes the law feel inevitable. Think of u as the steady temperature of a thin metal plate — heat has finished flowing, nothing is warming or cooling anymore. Could the single hottest point sit somewhere in the interior? Intuitively no: a tiny interior peak would be hotter than all its neighbors, so heat would stream away from it down the temperature gradient, and the peak would melt down into its surroundings. A genuinely steady state cannot tolerate an isolated interior summit. The mean-value law is exactly this intuition made into arithmetic — a center that equals the average of its ring cannot stick up above that ring.
Now turn the intuition into a clean argument. Suppose u is harmonic on a region and reaches its largest value M at some interior point z_0, so u(z_0) = M. On any small circle around z_0, every boundary value is at most M. But the mean-value property says u(z_0) is the average of those boundary values, and an average of numbers all at most M can equal M only if every one of them equals M. So u equals M on that whole circle, and — sweeping over all small radii — on a whole little disk around z_0. The peak is not a peak at all; it is a plateau.
That plateau then spreads. The set of points where u = M is closed (u is continuous) and, by the argument just given, open (around any such point u is M on a whole disk). On a connected region the only sets that are both open and closed are the empty set and everything, so if the maximum is attained anywhere inside, u is the constant M on the entire region. Contrapositive: a non-constant harmonic function attains its maximum nowhere in the interior. This is the maximum principle for harmonic functions.
Minimum too — and that is the real difference from |f|
Run the same argument for the smallest value and you get the minimum principle for free: a non-constant harmonic function attains its minimum only on the boundary as well. The slick way to see it is that if u is harmonic then so is -u, and the maximum of -u is the minimum of u. So a harmonic function on a bounded region, continuous up to the boundary, takes both its largest and its smallest values on the boundary rim — the interior is squeezed between the extremes set on the edge. The coldest and the hottest spots of the steady plate both live on its rim.
The two principles are not rivals, though; they are cousins. If u is harmonic and never zero you could exponentiate it into the modulus of a holomorphic function and recover the harmonic bound from the modulus bound — and conversely, the real part of any holomorphic f is harmonic, so the harmonic maximum principle is the more flexible statement of the two. Whenever you next see |f| forbidden a peak, remember that under the hood it is log|f| (a harmonic function away from the zeros of f) doing the averaging.
The payoff: a Dirichlet problem has at most one answer
The maximum principle is not just a pretty picture; it is the tool that makes potential theory well-posed. Recall the Dirichlet problem that frames this whole rung: paint continuous boundary data g on the rim of a region D and ask for the harmonic function u inside that matches g on the boundary. Whether such a u EXISTS is a separate, harder question — the Poisson integral of the next guide answers it for a disk. But the maximum principle settles UNIQUENESS in three lines, before any solution is even built.
- Suppose two solutions, u_1 and u_2, are both harmonic in D, continuous up to the boundary, and both equal the same data g on the boundary. Form their difference w = u_1 - u_2; since Laplace's equation is linear, w is harmonic too.
- On the boundary w = g - g = 0. By the maximum principle, w's largest value over the closed region is attained on the boundary, where it is 0; so w <= 0 everywhere inside.
- By the minimum principle, w's smallest value is also attained on the boundary, where it is 0; so w >= 0 everywhere inside.
- Squeezed between w <= 0 and w >= 0, the difference w is identically 0, so u_1 = u_2. Two solutions with the same boundary data must coincide — the answer, if it exists, is unique.
Honest limits and what comes next
Two honesty notes before we move on. First, the boundary version of the principle — extremes on the rim — needs the region to be bounded and u to be continuous up to its boundary; on an unbounded domain a harmonic function can grow off toward infinity and escape the rule, which is why sharper substitutes like the Phragmen-Lindelof principle exist for strips and half-planes. Second, 'no interior maximum' does not mean 'no interior critical point': a saddle of u, where the surface dips one way and rises the other, is perfectly allowed — what is forbidden is a true local max or min, a point higher (or lower) than all its neighbors at once.
It is also worth marking how far this reaches beyond the plane. The mean-value property and the maximum principle hold for harmonic functions in any dimension and underpin the theory of elliptic partial differential equations across mathematics and physics — the two-dimensional case you are learning here is the visual, complex-analytic gateway to all of it, with the bonus that here harmonic functions are exactly the real parts of holomorphic ones.
With uniqueness secured, the rung now turns to existence and refinement. The next guide builds the missing solution explicitly with the Poisson integral, a weighted average of boundary values that produces the harmonic function inside a disk. From there, two sharper instruments grow out of the same averaging soil: Harnack's inequality, which says a positive harmonic function cannot be huge in one spot and tiny next door, and the Schwarz reflection principle, which extends a harmonic or holomorphic function across a straight edge by mirroring. All of them trace back to the one law you proved today: the center is the average of the ring.