One derivative wearing three hats
We built one derivative in the last chapter. But mathematics is full of things that look like other derivatives — the growth rate a banker uses, the generator that drives an iteration — and they seem to belong to different subjects. They don't. They are the one derivative, read through a coordinate. That coordinate is the object worth naming.
To compare a function to a shifted copy of itself, you first have to decide how you are comparing — and that decision is a coordinate on the output line. Call it a chart α. The derivative in the chart α is just the ordinary derivative taken after re-coordinatizing the output with α:
That is the whole instrument. One rule with one slot in it — the chart — and everything below is a matter of what you drop into the slot.
Choose the plainest chart, α = id, the one that changes nothing, and the instrument gives back the ordinary slope:
This is the structural differential from the last chapter, the coefficient you read off grade 1. Now change the chart to α = log. Because (log ∘ f)′ = f′/f, the same instrument now returns the growth rate — the quantity a population, a bank balance, or a decaying isotope actually obeys:
Nothing about the function changed. We changed the ruler on the output, and "rate of change" became "rate of proportional change." This is the multiplicative twin from the last chapter, and it is why log turns products into sums: measured through log, multiplying is adding.
The third hat is the one textbooks file under a different subject entirely. Ask not "how does f change" but "what happens when I iterate it" — compare f(f(x)) to f(x). That looks nonlinear and awkward, until you find the chart in which f becomes a plain shift. Take f(x) = x²: in the log chart it reads as
and so iterating f is just doubling that coordinate over and over — f applied n times is x^(2ⁿ), read straight off 2ⁿ · log x. The wild-looking iteration was an ordinary "×2 each step," wearing a coordinate. Iteration is not a third kind of change; it is the same instrument in a chart chosen so that f stands still.
This should feel familiar. Back at the addition dial, one family of charts — raise to the power p — swept out every addition, ordinary and harmonic and the rest. It is the same family here: the chart α = xᵖ that indexed the additions also indexes the derivatives. That is not a rhyme. "Compare outputs in a given group" and "differentiate in the chart that flattens that group" are the same instruction spoken twice — once about addition, once about the derivative. The dial and the three hats are one object seen from two sides.
The three charts above — id, log, xᵖ — are ones you can look up in a table. Elementary functions with elementary inverses; you write the chart, you conjugate, you are done, exactly and reversibly. Most of classical manipulation never leaves this comfortable room.
But nothing guaranteed the chart would be writable. Sometimes the coordinate that flattens f is fixed uniquely by a condition and yet has no closed form — you know it exists, you can compute with it, you simply cannot write it down. The condition is an equation for the chart itself: "in the coordinate α, f is a shift," which pins α as tightly as any formula would, without handing you one. And past that, there are charts that cannot be made to cover the whole line at all — where the coordinate is honest only locally, and refuses to close up globally. Those failures are not accidents; they come in a short, exact list.
The derivative has stopped being a menu. There is one difference operator, and the slope, the growth rate, and the iteration generator are it — read through a chart you either look up or solve for. The question worth asking is no longer "which derivative do I want," but "which coordinate am I in, and can I write it?" Answering that second half — the ladder from charts you tabulate to charts you can only approach — is the map the rest of this part follows.