The promise from the last guide, now paid in full
Earlier in this rung you glimpsed a bonus falling out of the Cauchy-Riemann equations: the two parts u and v of a holomorphic function each satisfy Laplace's equation. That was a teaser. Now we earn it slowly and honestly, because this fact is the single most useful bridge between complex analysis and the physical world. Write the function as f = u + i v, where u(x, y) and v(x, y) are its real and imaginary parts, and recall the equations the last rung pinned down: u_x = v_y and u_y = -v_x.
First, one honest precaution that the teaser glossed over. To differentiate the Cauchy-Riemann equations a second time, we need u and v to have continuous second partials so the mixed partials are equal. On the real line that would be a genuine extra assumption. In the complex world it comes for free: being holomorphic forces infinite differentiability — a single complex derivative on an open set hands you all the derivatives you could ever want. So we are allowed to differentiate as many times as the proof asks, and the smoothness we lean on is not assumed but inherited.
Two lines of algebra, and Laplace appears
Here is the whole derivation, slow enough to follow with a pencil. Take the first equation u_x = v_y and differentiate both sides with respect to x: that gives u_xx = v_yx. Take the second equation u_y = -v_x and differentiate both sides with respect to y: that gives u_yy = -v_xy. Now add the two results. On the right you meet v_yx + (-v_xy); but for a smooth function the order of mixed partials does not matter, so v_yx = v_xy and they cancel exactly. What survives on the left is u_xx + u_yy = 0.
Cauchy-Riemann: u_x = v_y u_y = -v_x
d/dx of the first: u_xx = v_yx
d/dy of the second: u_yy = -v_xy
add, and use v_yx = v_xy :
u_xx + u_yy = 0 (Laplace's equation)
same trick on v gives:
v_xx + v_yy = 0The expression u_xx + u_yy is so central it has its own name, the Laplacian of u, often written as the triangle symbol applied to u. The equation u_xx + u_yy = 0 is Laplace's equation, and any function obeying it on an open set is called harmonic. Run the identical trick starting from the v-versions — differentiate u_x = v_y by y and u_y = -v_x by x, then subtract — and you find v_xx + v_yy = 0 as well. So both parts are harmonic, and that is the headline: the real and imaginary parts of a holomorphic function are each harmonic.
Not just harmonic — locked into a partnership
It would be a mistake to file this away as "u is harmonic, and separately v is harmonic." The Cauchy-Riemann equations bind them far more tightly than that. They say u and v change in step: the slope of u in the x-direction equals the slope of v in the y-direction, and the slope of u in the y-direction is the negative of v's slope in x. Two harmonic functions tied together this way are called harmonic conjugates: v is a harmonic conjugate of u when the pair solves the Cauchy-Riemann equations together. The relationship is not symmetric in a careless way — if v is a harmonic conjugate of u, then u is a harmonic conjugate of -v, the minus sign being the fingerprint of the equations.
A vivid picture seals the partnership. The level curves of u — the contours where u is constant — and the level curves of v form two families that cross each other at right angles everywhere. This is exactly the amplitwist from the previous guide seen through the eyes of u and v: because a holomorphic derivative is a pure rotation-and-scaling that preserves angles, the grid lines of the input get carried to a grid that is still perpendicular, and that perpendicular grid is precisely the contours of u against the contours of v. In a heat problem the u-contours are isotherms and the v-contours are the lines heat flows along; they meet at ninety degrees, just as physics demands.
Working backwards: rebuild f from just one part
Here is the payoff that makes harmonic conjugates a tool rather than a curiosity. Suppose someone hands you a single harmonic function u and asks you to find a holomorphic f whose real part is exactly that u. Astonishingly, you almost can — you can recover the whole function from its real part alone, up to one additive constant. The Cauchy-Riemann equations are the recipe: they tell you what the partial derivatives of the missing partner v must be, and then you integrate.
- Take u = x^2 - y^2 (you can check u_xx + u_yy = 2 + (-2) = 0, so it is harmonic). Compute its partials: u_x = 2x and u_y = -2y.
- Use the first Cauchy-Riemann equation u_x = v_y, so v_y = 2x. Integrate in y, treating x as constant: v = 2x y + g(x), where g is an unknown function of x alone, the constant of integration.
- Pin down g using the second equation u_y = -v_x. Differentiate your v in x: v_x = 2y + g'(x). The equation says this must equal -u_y = 2y. So g'(x) = 0, meaning g is a genuine constant C.
- Assemble: v = 2x y + C, so f = (x^2 - y^2) + i (2x y + C). You may recognise x^2 - y^2 + i (2 x y) as exactly z^2, so f = z^2 + i C — the holomorphic function whose real part is u, unique up to the constant.
Notice two things from the worked example. First, the answer was forced: the only freedom was the harmless additive constant C, which just shifts f vertically and changes nothing about its behaviour. Second, the construction quietly assumed we could integrate v_y across the region without contradiction. That works smoothly on a simply connected domain — one with no holes — but on a region with a hole the partner v can fail to close up into a single-valued function. The classic warning is the harmonic function log|z| on the punctured plane: it has no single-valued harmonic conjugate, because its would-be partner is arg z, which jumps by 2 pi each time you loop the origin. That is the same multivaluedness you met with the complex logarithm, wearing a harmonic-conjugate disguise.
Why anyone outside mathematics cares
Laplace's equation is not some obscure curiosity; it is arguably the most important partial differential equation in physics. Steady-state temperature with no internal heat source is harmonic. The electrostatic potential in a charge-free region is harmonic. The velocity potential of an ideal, incompressible, irrotational fluid is harmonic. So the moment you write down any holomorphic f, you have for free TWO solutions to a real physical problem — its real part and its imaginary part — already paired as potential and flow. Engineers exploit this directly: a clever holomorphic function is a complex potential whose real part gives the potential and whose imaginary part traces the streamlines.
Once u is harmonic it inherits a whole estate of rigidity from the holomorphic world. By the mean value property, the value of u at the centre of any disk equals its average over the boundary circle — a harmonic function can never bulge above its surroundings. From this follows the maximum principle for harmonic functions: a non-constant harmonic function attains neither a maximum nor a minimum in the interior of its region; the extremes always sit on the boundary. That is physically obvious in hindsight — the hottest point of a uniformly conducting plate with no heat source must be on its edge — but here it drops out as a theorem, courtesy of the link to holomorphy.
The chain you have forged
Step back and look at the whole rung as one argument. A single complex derivative, demanded across an open set, was so severe that it could only be satisfied by a holomorphic function. That demand forced the Cauchy-Riemann equations. Those equations forced the derivative to act as a pure rotation-and-scaling, and — differentiated once more — forced each of u and v to satisfy Laplace's equation, bound together as harmonic conjugates. So the parts of a holomorphic function are not free; they are two halves of one rigid object, each determining the other up to a constant.
One last honest word so you do not over-promise. Everything here lived inside a region, an open set, with the boundary deliberately set aside; behaviour right at an edge or at a singular point is a separate, more delicate story. And the harmonic-conjugate construction needed the region to be simply connected, or the partner may refuse to be single-valued, exactly as log|z| showed. With those caveats honoured, you now hold the central truth of this rung: complex differentiability is not a small upgrade on real differentiability — it is a contract so strong that one derivative buys you a power series, infinite smoothness, angle preservation, and a matched pair of solutions to the deepest equation in two-dimensional physics. The next rung turns this rigidity loose on integrals, where it becomes downright magical.