Skip to content

Chapter 4 — Exponential and Logarithmic Functions

4.3 - Logarithmic Functions

Every injective function keeps, in its own behavior, the possibility of undoing itself. The exponential has already shown itself to be one; all that remains is to claim that possibility.


§4.1 proved, for f(x)=axf(x)=a^x with a>0, a≠1a>0,\ a\neq1, much more than was needed at the time: not only that ff is strictly monotone (§9, section 4.1), from which its injectivity follows, but also that its range is exactly (0,∞)(0,\infty) (§10, section 4.1), with nothing missing. Those two pieces, taken together, are exactly what §2.7 required to guarantee the existence of an inverse function: injectivity for the uniqueness of the preimage, surjectivity, onto the correct codomain, for its existence. Nothing new needs to be proved to obtain the inverse; all that is needed is to recognize that the conditions are already given.

§1. Bijectivity of aˣ and existence of the inverse

Section titled “§1. Bijectivity of aˣ and existence of the inverse”

Fix a>0a>0, a≠1a\neq1, and consider f:R→(0,∞)f:\mathbb{R}\to(0,\infty), f(x)=axf(x)=a^x, with the codomain restricted exactly to its range, as already proved.

Proposition. ff is bijective.

Proof. Injectivity: Corollary of §9, section 4.1. Surjectivity: since the codomain was fixed equal to the range, Rec⁡(f)=(0,∞)\operatorname{Rec}(f)=(0,\infty), §10, section 4.1, guarantees that every element of the codomain is attained. ■\blacksquare

By §2.7 §3, every bijective function admits an inverse, unique, characterized by the identities f−1(f(x))=xf^{-1}(f(x))=x and f(f−1(y))=yf(f^{-1}(y))=y. Applied here:

Definition. The inverse function of f(x)=axf(x)=a^x is called the logarithmic function with base aa, and is denoted log⁡a\log_a. That is: for y>0y>0, log⁡ay\log_a y is the unique x∈Rx\in\mathbb{R} such that ax=ya^x=y.


With log⁡a\log_a defined as an inverse, it is time to make explicit the two identities that characterize it completely, and which in practice are the most direct computational tool. That is the content of §2.

By the general characterization of every inverse function, §2.7 §3, applied here to f(x)=axf(x)=a^x and f−1=log⁡af^{-1}=\log_a, we have, without additional proof, the two identities:

log⁡a(ax)=xfor all x∈Ralog⁡ay=yfor all y>0\log_a(a^x)=x \quad \text{for all } x\in\mathbb{R} \qquad\qquad a^{\log_a y}=y \quad \text{for all } y>0

The first says that applying the logarithm after the exponential returns the original exponent; the second, that applying the exponential after the logarithm returns the original number. Neither is a new property of the logarithm: it is, literally, the definition of inverse, §2.7 §3, specialized to this pair of functions.


With the function and its characteristic identities defined, it is time to specify its sets of definition and arrival, which, since log⁡a\log_a is an inverse, require no new computation. That is the content of §3.

Proposition. Dom⁡(log⁡a)=(0,∞)\operatorname{Dom}(\log_a)=(0,\infty) and Rec⁡(log⁡a)=R\operatorname{Rec}(\log_a)=\mathbb{R}.

Proof. By §2.7 §4, for every bijective function ff, Dom⁡(f−1)=Img⁡(f)\operatorname{Dom}(f^{-1})=\operatorname{Img}(f) and Img⁡(f−1)=Dom⁡(f)\operatorname{Img}(f^{-1})=\operatorname{Dom}(f); domain and image are exchanged completely. Applied to f(x)=axf(x)=a^x, with Dom⁡(f)=R\operatorname{Dom}(f)=\mathbb{R} and Img⁡(f)=(0,∞)\operatorname{Img}(f)=(0,\infty), §10, section 4.1, we obtain immediately Dom⁡(log⁡a)=(0,∞)\operatorname{Dom}(\log_a)=(0,\infty) and Rec⁡(log⁡a)=R\operatorname{Rec}(\log_a)=\mathbb{R}. ■\blacksquare

It is worth pausing on what this means in practice: log⁡ay\log_a y is not defined for y≤0y\le 0, since neither zero nor any negative is ever attained by axa^x, according to the range of the exponential itself, while, in the other direction, every real number, however large or small, is the logarithm of some y>0y>0.


Domain and range were resolved by direct inheritance. Monotonicity, on the other hand, requires its own proof, since §2.7 did not prove it in general for every inverse, and is the content of §4.

Proposition. If a>1a>1, log⁡a\log_a is strictly increasing on (0,∞)(0,\infty). If 0<a<10<a<1, log⁡a\log_a is strictly decreasing on (0,∞)(0,\infty).

Proof. Let y1,y2∈(0,∞)y_1,y_2\in(0,\infty) with y1<y2y_1<y_2, and let u1=log⁡ay1u_1=\log_a y_1, u2=log⁡ay2u_2=\log_a y_2, so that au1=y1a^{u_1}=y_1 and au2=y2a^{u_2}=y_2, by the definition of §1.

Case a>1a>1. Suppose, for contradiction, u1≥u2u_1\ge u_2. Since f(x)=axf(x)=a^x is strictly increasing for a>1a>1, §9, section 4.1, in particular it is nondecreasing: u1≥u2u_1\ge u_2 implies au1≥au2a^{u_1}\ge a^{u_2}, that is, y1≥y2y_1\ge y_2, contradicting y1<y2y_1<y_2. Hence u1<u2u_1<u_2, that is, log⁡ay1<log⁡ay2\log_a y_1<\log_a y_2.

Case 0<a<10<a<1. Suppose, for contradiction, u1≤u2u_1\le u_2. Since f(x)=axf(x)=a^x is strictly decreasing for 0<a<10<a<1, §9, section 4.1, in particular it is nonincreasing: u1≤u2u_1\le u_2 implies au1≥au2a^{u_1}\ge a^{u_2}, that is, y1≥y2y_1\ge y_2, contradicting y1<y2y_1<y_2. Hence u1>u2u_1>u_2, that is, log⁡ay1>log⁡ay2\log_a y_1>\log_a y_2. ■\blacksquare


With domain, range and monotonicity already established, what remains are the distinguished values of the logarithm, its behavior at the ends of the domain, its graph, and the block of exercises and close.

Proposition. log⁡a1=0\log_a 1=0 and log⁡aa=1\log_a a=1.

Proof. By the rational consistency of §6 (or §7) of section 4.1, a0=1a^0=1. By the Definition of §1, log⁡a1\log_a 1 is the unique real xx with ax=1a^x=1; since a0=1a^0=1, that unique real is x=0x=0. Likewise, a1=aa^1=a, a trivial particular case of exponentiation, so that log⁡aa\log_a a is the unique real with ax=aa^x=a, and that real is x=1x=1. ■\blacksquare


It remains to describe what happens with log⁡a\log_a at the ends of its domain: near y=0y=0, where the domain ends but does not include that point, and when yy grows without bound. That behavior is, exactly, the mirror of §11 of section 4.1, with the roles of domain and range exchanged.

Proposition (case a>1a>1). For every M>0M>0 there exists δ>0\delta>0 such that 0<y<δ  ⟹  log⁡ay<−M0<y<\delta \implies \log_a y<-M. Moreover, for every M>0M>0 there exists N>0N>0 such that y>N  ⟹  log⁡ay>My>N \implies \log_a y>M.

Proof. Let δ=a−M>0\delta=a^{-M}>0. Suppose 0<y<δ0<y<\delta and, for contradiction, log⁡ay≥−M\log_a y\ge -M. Since f(x)=axf(x)=a^x is nondecreasing for a>1a>1, §9, section 4.1, log⁡ay≥−M\log_a y\ge -M implies alog⁡ay≥a−Ma^{\log_a y}\ge a^{-M}, that is, y≥δy\ge\delta, contradicting y<δy<\delta. Hence log⁡ay<−M\log_a y<-M.

Let N=aM>0N=a^M>0. Suppose y>Ny>N and, for contradiction, log⁡ay≤M\log_a y\le M. By the same monotonicity, alog⁡ay≤aMa^{\log_a y}\le a^M, that is, y≤Ny\le N, contradicting y>Ny>N. Hence log⁡ay>M\log_a y>M. ■\blacksquare

The line x=0x=0 is, then, a vertical asymptote of log⁡a\log_a when a>1a>1: as yy approaches 00 from the right, log⁡ay\log_a y decreases without bound. And log⁡a\log_a grows without bound as y→+∞y\to+\infty, with no horizontal asymptote whatsoever.

Proposition (case 0<a<10<a<1). For every M>0M>0 there exists δ>0\delta>0 such that 0<y<δ  ⟹  log⁡ay>M0<y<\delta\implies\log_a y>M. Moreover, for every M>0M>0 there exists N>0N>0 such that y>N  ⟹  log⁡ay<−My>N\implies\log_a y<-M.

Proof. Analogous to the previous one, using that f(x)=axf(x)=a^x is nonincreasing for 0<a<10<a<1, §9, section 4.1: with δ=aM\delta=a^M, if 0<y<δ0<y<\delta and log⁡ay≤M\log_a y\le M, nonincreasing monotonicity gives alog⁡ay≥aM=δa^{\log_a y}\ge a^M=\delta, that is y≥δy\ge\delta, a contradiction; hence log⁡ay>M\log_a y>M. With N=a−MN=a^{-M}, if y>Ny>N and log⁡ay≥−M\log_a y\ge -M, nonincreasing monotonicity gives alog⁡ay≤a−M=Na^{\log_a y}\le a^{-M}=N, that is y≤Ny\le N, a contradiction; hence log⁡ay<−M\log_a y<-M. ■\blacksquare

In this second case, the line x=0x=0 is still a vertical asymptote, but with inverted sign: log⁡ay\log_a y grows without bound as y→0+y\to 0^+, and decreases without bound as y→+∞y\to+\infty, exactly the mirror of the behavior of axa^x with 0<a<10<a<1, which decreased toward 00 instead of growing.


With domain, range, monotonicity and asymptotics already complete, the graph of log⁡a\log_a requires no new analysis: it is obtained, geometrically, from the already known graph of axa^x. That is the content of §7.

By §2.7 §4, the graph of every inverse function is the reflection of the graph of the original function about the line y=xy=x. Applied here: Gr(log⁡a)\text{Gr}(\log_a) is, without further computation, the reflection of Gr(ax)\text{Gr}(a^x) about that diagonal. Each feature already proved translates geometrically without an additional word: the horizontal asymptote y=0y=0 of axa^x is reflected into the vertical asymptote x=0x=0 of log⁡a\log_a; the point (0,1)(0,1), common to every exponential, is reflected into the point (1,0)(1,0), common to every logarithm, exactly what §5 asserted; and monotonicity is preserved in the same direction, because a reflection about y=xy=x does not invert whether a curve rises or falls from left to right.

Two panels: on the left, y=2^x and y=log_2(x) reflected about y=x, with the points (1,2) and (2,1) marked; on the right, y=(1/2)^x and y=log_(1/2)(x) reflected about y=x, with the points (1, 0.5) and (0.5, 1) marked Figure 4.1 — axa^x and log⁡ax\log_a x, reflected about y=xy=x, for a>1a>1 (left, a=2a=2) and 0<a<10<a<1 (right, a=1/2a=1/2). In both cases, the point (1,a)∈Gr(ax)(1,a)\in\text{Gr}(a^x) corresponds to (a,1)∈Gr(log⁡ax)(a,1)\in\text{Gr}(\log_a x), the same type of correspondence as Figure 2.5 of §2.7 §4, applied here to a pair of curves instead of to two lines.


What remains, to close the section, is to exercise the handling of the inversion identities and of the monotonicity just proved, and a brief reflection on what the logarithm makes possible for the next section. That is the content of §8.

Exercise 1. Compute log⁡28\log_2 8.

Solution. 8=238=2^3, so that, by the identity log⁡a(ax)=x\log_a(a^x)=x of §2 with a=2, x=3a=2,\ x=3, log⁡28=log⁡2(23)=3\log_2 8=\log_2(2^3)=3.

Exercise 2. Solve log⁡3x=4\log_3 x=4.

Solution. By the Definition of §1, log⁡3x=4\log_3 x=4 means 34=x3^4=x. Hence x=81x=81.

Exercise 3. Solve the inequality log⁡2x<3\log_2 x<3, specifying the domain.

Solution. The domain requires x>0x>0, §3. Since 2>12>1, log⁡2\log_2 is strictly increasing (§4), so that log⁡2x<3=log⁡28\log_2 x<3=\log_2 8, Exercise 1, is equivalent, by strict monotonicity, to x<8x<8. Combining with the domain, the solution is 0<x<80<x<8.

Exercise 4. Simplify alog⁡a5+log⁡a2a^{\log_a 5+\log_a 2}.

Solution. By the homomorphism of §8, section 4.1, valid for any pair of reals in the exponent, and log⁡a5,log⁡a2\log_a5,\log_a2 are such, alog⁡a5+log⁡a2=alog⁡a5⋅alog⁡a2a^{\log_a5+\log_a2}=a^{\log_a5}\cdot a^{\log_a2}. By the inversion identity of §2, each factor simplifies separately: alog⁡a5=5a^{\log_a5}=5 and alog⁡a2=2a^{\log_a2}=2. Hence alog⁡a5+log⁡a2=5⋅2=10a^{\log_a5+\log_a2}=5\cdot2=10.

It is worth noting what was not used: at no moment was it necessary to write log⁡a5+log⁡a2\log_a5+\log_a2 as the logarithm of a product; that equivalence, log⁡a(xy)=log⁡ax+log⁡ay\log_a(xy)=\log_a x+\log_a y, is precisely the first law that section 4.4 will prove. What this exercise resolves is the opposite direction: what happens when the sum of two logarithms appears in the exponent, not inside another logarithm.

With this, the logarithm is completely constructed from a single idea, that the exponential, being bijective, can be undone, and from a single source of new results, monotonicity and asymptotics, both obtained by contradiction from what the exponential already guaranteed. There was, in this section, no construction from scratch: everything rested on §2.7, written chapters ago without yet thinking of the exponential, and on §4.1, written this very week thinking exactly of this moment. That both fit together without friction is no coincidence: it is the kind of economy the book has pursued since the first chapter, and which only becomes visible when a distant piece fits, without adjustments, into a place that did not know it was awaited.


With section 4.3 closed, §4.4 takes up the homomorphism of the exponential, §8 of section 4.1, to translate it, term by term, into the laws of logarithms: product into sum, quotient into difference, power into product.