Books Sample

Inverses

A function may or may not have an inverse. The obstruction is already in Definition 1.1: each input has exactly one output, but two inputs may share an output. If they do, I cannot run the rule backwards.

Definition 1.2injective
𝑓:𝐴→𝐵𝑓:𝐴\to 𝐵 is injective if 𝑓(𝑎1)=𝑓(𝑎2)𝑓({𝑎}_{1})=𝑓({𝑎}_{2}) implies 𝑎1=𝑎2{𝑎}_{1}={𝑎}_{2}.
Definition 1.3inverse
If 𝑓:𝐴→𝐵𝑓:𝐴\to 𝐵 is injective, the inverse 𝑓−1:𝑓(𝐴)→𝐴{𝑓}^{-1}:𝑓(𝐴)\to 𝐴 is the function that sends 𝑓(𝑎)𝑓(𝑎) back to 𝑎𝑎. Equivalently, 𝑓−1∘𝑓=𝐼𝐴{𝑓}^{-1}\circ 𝑓={𝐼}_{𝐴} and 𝑓∘𝑓−1=𝐼𝑓(𝐴)𝑓\circ {𝑓}^{-1}={𝐼}_{𝑓(𝐴)}.

The second line uses Proposition 1.1. I have to write the codomain of 𝑓−1{𝑓}^{-1} as the image 𝑓(𝐴)𝑓(𝐴), not as 𝐵𝐵, unless 𝑓𝑓 hits everything. Surjectivity onto 𝐵𝐵 is a separate request; injectivity is what makes a left inverse exist.

Theorem 1.2
An injective function has at most one inverse.
Proof
Suppose 𝑔𝑔 and hh both invert 𝑓:𝐴→𝐵𝑓:𝐴\to 𝐵 on 𝑓(𝐴)𝑓(𝐴). For 𝑏=𝑓(𝑎)𝑏=𝑓(𝑎) in the image, 𝑔(𝑏)=𝑔(𝑓(𝑎))=𝑎=h(𝑓(𝑎))=h(𝑏)𝑔(𝑏)=𝑔(𝑓(𝑎))=𝑎=h(𝑓(𝑎))=h(𝑏). So 𝑔=h𝑔=h as functions 𝑓(𝐴)→𝐴𝑓(𝐴)\to 𝐴.

Cutting the domain until it works

Absolute value on R\mathbb{R} is not injective: ∣−2∣=∣2∣|-2|=|2|. There is no inverse on the whole line. Restrict to [0,∞)[0,\infty ) and 𝑥↦∣𝑥∣𝑥\mapsto |𝑥| becomes the identity, which is its own inverse. Restrict to (−∞,0](-\infty ,0] and the inverse is 𝑥↦−𝑥𝑥\mapsto -𝑥. Same formula, two domains, two different invertible functions. That is Definition 1.1 again: the domain is part of the data.

Example 1.3
𝑠(𝑥)=𝑥2𝑠(𝑥)={𝑥}^{2} on R\mathbb{R} is not injective. On [0,∞)[0,\infty ) it is, and the inverse is 𝑥↦𝑥𝑥\mapsto \sqrt{𝑥}. On (−∞,0](-\infty ,0] the inverse is 𝑥↦−𝑥𝑥\mapsto -\sqrt{𝑥}. Compare Example 1.2 and Composition: after the restriction, 𝑟∘𝑠𝑟\circ 𝑠 and 𝑠∘𝑟𝑠\circ 𝑟 become identities on the cut domains.

Inverse of a composite

If I can invert each factor, I can invert the composite, but the order flips.

Proposition 1.2
If 𝑓:𝐴→𝐵𝑓:𝐴\to 𝐵 and 𝑔:𝐵→𝐶𝑔:𝐵\to 𝐶 are injective, then (𝑔∘𝑓)−1=𝑓−1∘𝑔−1{(𝑔\circ 𝑓)}^{-1}={𝑓}^{-1}\circ {𝑔}^{-1} on the image of 𝑔∘𝑓𝑔\circ 𝑓.
Proof
(𝑓−1∘𝑔−1)∘(𝑔∘𝑓)=𝑓−1∘(𝑔−1∘𝑔)∘𝑓=𝑓−1∘𝑓=𝐼𝐴({𝑓}^{-1}\circ {𝑔}^{-1})\circ (𝑔\circ 𝑓)={𝑓}^{-1}\circ ({𝑔}^{-1}\circ 𝑔)\circ 𝑓={𝑓}^{-1}\circ 𝑓={𝐼}_{𝐴} by (1.4) and Definition 1.3. The other composite is the identity on the image by the same parentheses.

I will use this later without thinking: the chain rule differentiates a composite in the same reversed order that inversion undoes it.

Exercise 1.3
Show that 𝑓(𝑥)=𝑥3−𝑥𝑓(𝑥)={𝑥}^{3}-𝑥 on R\mathbb{R} is not injective. Find three points with the same image, or argue from the derivative after you have it.
Solution to Exercise 1.3
𝑓(−1)=0=𝑓(0)=𝑓(1)𝑓(-1)=0=𝑓(0)=𝑓(1). Three distinct inputs, one output, so Definition 1.2 fails. After (2.1) one can also note that 𝑓′{𝑓}^{\prime } changes sign, but three roots of 𝑓(𝑥)=0𝑓(𝑥)=0 are enough.