🧮 Brain Teaser

The Entire Function That Squares to Another

Let f:C→Cf: \mathbb{C} \to \mathbb{C} be an entire function satisfying

f(z)2=z2−1f(z)^2 = z^2 - 1

for all z∈Cz \in \mathbb{C}. Does such ff exist?

entire functionszeros and orderswinding numberbranch cutssquare root obstruction

Answer: The Entire Function That Squares to Another

Key Idea / Intuition

The function z2−1z^2 - 1 has two zeros, at z=1z = 1 and z=−1z = -1. If f2=z2−1f^2 = z^2 - 1, then ff must vanish at these points. But near each zero of z2−1z^2 - 1 (which is a simple zero), ff would have to be a "half-zero" — a zero of order 1/21/2 — which cannot be holomorphic. More precisely, the issue is topological: the function z2−1z^2 - 1 changes sign as you loop around one of its zeros, so there is no consistent holomorphic square root globally on C\mathbb{C}.


Formal Proof / Solution

Suppose for contradiction that such an entire ff exists with f(z)2=z2−1f(z)^2 = z^2 - 1.

Step 1: Zeros of ff.

At z=1z = 1: we have f(1)2=0f(1)^2 = 0, so f(1)=0f(1) = 0. Similarly f(−1)=0f(-1) = 0.

Write z2−1=(z−1)(z+1)z^2 - 1 = (z-1)(z+1). Both zeros are simple (order 1). Since f2=z2−1f^2 = z^2-1, the order of vanishing satisfies 2 ordz=1(f)=12\,\mathrm{ord}_{z=1}(f) = 1, which is impossible for a holomorphic function (orders must be non-negative integers).

Conclusion from Step 1: No entire ff can satisfy f2=z2−1f^2 = z^2-1.


Alternative (topological) argument: Suppose such ff is entire (hence continuous). Consider the loop γ(t)=eit\gamma(t) = e^{it} for t∈[0,2π]t \in [0, 2\pi], which winds once around both z=1z = 1 and z=−1z = -1 together. Actually, let us use a smaller loop.

Consider γ(t)=1+ϵeit\gamma(t) = 1 + \epsilon e^{it} winding once around z=1z = 1 (with ϵ\epsilon small enough to exclude z=−1z = -1). Along γ\gamma,

f(γ(t))2=γ(t)2−1=(γ(t)−1)(γ(t)+1).f(\gamma(t))^2 = \gamma(t)^2 - 1 = (\gamma(t)-1)(\gamma(t)+1).

Near z=1z = 1, we have (γ(t)−1)=ϵeit(\gamma(t) - 1) = \epsilon e^{it} (winds once around 0) while (γ(t)+1)≈2(\gamma(t)+1) \approx 2 (nonzero, winds zero times). So f(γ(t))2f(\gamma(t))^2 winds once around 0.

But f(γ(t))2f(\gamma(t))^2 winds an even number of times around 0 for any closed curve (since winding number of w2w^2 is twice the winding number of ww). Contradiction.

More precisely: The winding number of f∘γf \circ \gamma around 0 equals 12⋅(winding number of f2∘γ around 0)=12⋅1=12\frac{1}{2}\cdot(\text{winding number of } f^2 \circ \gamma \text{ around } 0) = \frac{1}{2}\cdot 1 = \frac{1}{2}, which is not an integer — impossible for a continuous closed curve.


Summary: No entire function ff satisfies f(z)2=z2−1f(z)^2 = z^2 - 1.

The obstruction is that z2−1z^2 - 1 has simple zeros, so its square root would require a branch cut; there is no single-valued holomorphic square root of z2−1z^2-1 on any domain that includes both ±1\pm 1. (By contrast, z2+1z^2+1 restricted to the right half-plane does admit a holomorphic square root, since you can avoid the branch cut.)

Remark: If instead the equation were f(z)2=(z2−1)2f(z)^2 = (z^2-1)^2, then f(z)=±(z2−1)f(z) = \pm(z^2-1) works (an entire function). The key difference is that even-order zeros allow holomorphic square roots, but odd-order zeros do not.

Type: Complex AnalysisEdit on GitHub ↗