Diagram showing log x defined as the area under the curve y = 1/t from 1 to x, illustrating Spivak's integral definition of the logarithm

What Is log x, Really? Defining the Logarithm Without Exponents

Ask most calculus students what log x means, and you'll get some version of the same answer: "it's the power you raise 10 to, to get x." Exponents first, logarithms second, as the inverse. It feels obvious — until you try to make it rigorous. What, precisely, is 10^x when x is irrational, like x = √2? You can define 10^n for natural numbers n easily enough, extend it to negative and rational exponents by insisting that 10^(a+b) = 10^a · 10^b keeps holding, and get as far as 10^(m/n) = (ⁿ√10)^m. And then, as Spivak puts it, "the program comes to a dead halt." There is no simple algebraic recipe for 10^x when x is irrational. So in Chapter 18 of Calculus, he does something bolder: he throws exponents out entirely, and builds the logarithm first — as an area.

Defining log as an integral, not an inverse

Here is the move. Instead of chasing a definition of 10^x for irrational x and then trying to invert it, Spivak asks a different question first: is there some extremely simple function whose derivative is already the mystery function 1/x? Recall that among all the power functions, ∫ x^n dx is a solved problem for every integer n except n = −1. The integral of 1/t is the one gap in an otherwise complete table. So he simply names it:

log x = ∫₁ˣ (1/t) dt, for x > 0.

That's it. No exponentials, no bases, no inverse functions — just the signed area under the curve y = 1/t, swept out from 1 to x. If x > 1, that area is positive; if 0 < x < 1, you're integrating "backwards," so the area (and hence log x) comes out negative. For x ≤ 0 the definition simply doesn't apply, since 1/t isn't even bounded near 0.

By the Fundamental Theorem of Calculus, this definition hands you the derivative for free: log′(x) = 1/x. And notice what's already true from the picture alone — the region collapses to nothing when x = 1, so log 1 = 0, with no algebra required.

The familiar rules, derived rather than assumed

Now comes the payoff. The identity every algebra student memorizes — log(xy) = log x + log y — usually gets waved through as "a law of logarithms." Spivak proves it as Theorem 1, and the proof is a small gem: fix y > 0 and let f(x) = log(xy). By the Chain Rule, f′(x) = log′(xy) · y = 1/(xy) · y = 1/x. But that's exactly log′(x) too! Two functions with the same derivative everywhere differ only by a constant, so log(xy) = log x + c for some constant c. Plug in x = 1: log(y) = log 1 + c = c. Substitute back and you're done — no algebraic manipulation of exponents anywhere, just the Fundamental Theorem and the fact that equal derivatives mean functions differing by a constant. From there, log(xⁿ) = n log x and log(x/y) = log x − log y fall out as easy corollaries.

This same theorem tells you something about the graph, too: since log(2ⁿ) = n log 2 and log 2 > 0, the function log is unbounded above; a mirror argument shows it's unbounded below on (0, 1). Being continuous and taking every value in between, log is onto all of ℝ — which means it actually has an inverse function.

Getting the exponential function for free

This is where the story flips the usual order on its head. Since log: (0, ∞) → ℝ is increasing and onto, it has an inverse, and Spivak simply defines the exponential function as that inverse:

exp = log⁻¹

Everything you want from an exponential function now falls out as a theorem rather than an assumption. Differentiating an inverse function gives exp′(x) = 1/log′(exp(x)) = 1/(1/exp(x)) = exp(x) — the exponential is its own derivative, proved, not postulated. And applying log(xy) = log x + log y to x′ = exp(x), y′ = exp(y) yields exp(x + y) = exp(x) · exp(y), the other defining property of exponentials, likewise proved rather than assumed.

Only now, at the very end, does the famous constant show up: Spivak defines e = exp(1), which is equivalent to saying log e = ∫₁ᵉ (1/t) dt = 1. Using crude upper and lower sums for 1/t on [1,2] and [1,4], he pins it down immediately to 2 < e < 4. And because exp(x) = [exp(1)]ˣ = eˣ for rational x, defining eˣ := exp(x) for every real x — rational or not — finally solves the problem the chapter opened with. From there, an arbitrary base is handled by one clean definition, aˣ = e^(x log a) for a > 0, and Theorem 4 confirms it obeys every exponent law you'd hope for, including (ab)^c = a^(bc).

Why bother building it this way?

You could reasonably ask why any of this matters, if the end result — eˣ, log rules, the derivative formulas — is the same as what you learned in high school. The answer is that Spivak's route replaces a hand-wave with a proof. "Raise 10 to an irrational power" is not something algebra can define; "the area under 1/t from 1 to x" is something the integral defines rigorously for every real x > 0, no exceptions, no hidden assumptions about continuity smuggled in through the back door. Every property that follows — the addition law, the derivative, the very existence of e — is then a genuine theorem with a genuine proof, built from tools (the Fundamental Theorem, the Mean Value Theorem's corollary about equal derivatives, the Intermediate Value Theorem) that were already established in earlier chapters. That's the deeper lesson of Chapter 18: sometimes the most "obvious" definition in mathematics is not the most useful place to start, and the version that looks backwards on first glance turns out to be the one that actually holds together.

If you want to work through the full proofs of Theorems 1 through 6 — including the striking fact that eˣ eventually outgrows every polynomial, and the pathological function e^(−1/x²) that is infinitely differentiable yet flat as a pancake at the origin — Chapter 18 of Spivak's Calculus, 4th Edition is the place to see it done in full, with every step justified.

Further reading: How to self-study Spivak's Calculus · Table of contents · Index · Calculus, 4th Edition · Answer Book


Had you ever seen log defined as an integral before encountering it this way — or did you, like most of us, learn it as "the inverse of an exponent" first? Tell us in the comments.

Back to blog

Leave a comment

Please note, comments need to be approved before they are published.