Skip to content

Chapter 4 — Exponential and Logarithmic Functions

4.4 - Laws of Logarithms

Each algebraic property of the exponential has, on the other side of the mirror y=xy=x, an exact counterpart in the logarithm. There is no need to discover them: it suffices to read, one by one, those that are already proved.


§4.3 closed by pointing out what was still missing: the possibility of converting a sum inside the argument of the logarithm, or, seen from the other side, a product, into a simpler operation. That possibility requires no new tool. It requires only taking the homomorphism au+v=auava^{u+v}=a^u a^v, §8, section 4.1, and the corollaries derived from it, §12, section 4.1, and reading them with log⁡a\log_a in hand instead of with axa^x.

Proposition. For all x,y>0x,y>0, log⁡a(xy)=log⁡ax+log⁡ay\log_a(xy)=\log_a x+\log_a y.

Proof. Let u=log⁡axu=\log_a x and v=log⁡ayv=\log_a y, so that au=xa^u=x and av=ya^v=y, Definition of §1, section 4.3. By the homomorphism of §8, section 4.1, valid for any pair of reals,

au+v=au⋅av=xya^{u+v}=a^u\cdot a^v=xy

By the Definition of the logarithm, log⁡a(xy)\log_a(xy) is the unique real whose exponent on aa produces xyxy; since u+vu+v is, precisely, that real, log⁡a(xy)=u+v=log⁡ax+log⁡ay\log_a(xy)=u+v=\log_a x+\log_a y. ■\blacksquare


The same procedure, with a subtraction in the exponent instead of an addition, gives the quotient law. It is the content of §2.

Proposition. For all x,y>0x,y>0, log⁡a(xy)=log⁡ax−log⁡ay\log_a\left(\dfrac{x}{y}\right)=\log_a x-\log_a y.

Proof. With u=log⁡axu=\log_a x, v=log⁡ayv=\log_a y as before, Exercise 1 of §12, section 4.1, gives au−v=au/av=x/ya^{u-v}=a^u/a^v=x/y. By the same Definition, log⁡a(x/y)\log_a(x/y) is the unique real whose exponent on aa produces x/yx/y, and that real is u−vu-v. Hence log⁡a(x/y)=u−v=log⁡ax−log⁡ay\log_a(x/y)=u-v=\log_a x-\log_a y. ■\blacksquare


What is missing is the law that translates a power inside the argument into a product outside the logarithm. It is the most general of the three, because the exponent that comes down is not restricted to the rationals: it can be any real, thanks to the corollary (ax)y=axy(a^x)^y=a^{xy} already proved in 4.1. It is the content of §3.

Proposition. For all x>0x>0 and all r∈Rr\in\mathbb{R}, log⁡a(xr)=rlog⁡ax\log_a(x^r)=r\log_a x.

Proof. Let u=log⁡axu=\log_a x, so that au=xa^u=x. By Exercise 2 of §12, section 4.1, (au)r=aur(a^u)^r=a^{ur}, valid for any reals u,ru,r, without restriction to rational exponents,

xr=(au)r=aurx^r=(a^u)^r=a^{ur}

By the Definition of the logarithm, log⁡a(xr)\log_a(x^r) is the unique real whose exponent on aa produces xrx^r, and that real is urur. Hence log⁡a(xr)=ur=rlog⁡ax\log_a(x^r)=ur=r\log_a x. ■\blacksquare


With the three fundamental laws established, a fourth remains that does not translate an algebraic operation but a change of perspective: what relation the logarithms of one same number in two different bases bear to each other. It is the content of §4, and depends only on the power law just proved.

Proposition. For a,b>0a,b>0, a,b≠1a,b\neq1, and x>0x>0,

log⁡ax=log⁡bxlog⁡ba\log_a x=\frac{\log_b x}{\log_b a}

Proof. Let u=log⁡axu=\log_a x, so that au=xa^u=x. Applying log⁡b\log_b to both sides of this equality between positive reals, a well-defined function assigns the same value to equal arguments, log⁡b(au)=log⁡bx\log_b(a^u)=\log_b x. By the power law of §3, with base bb, argument aa and exponent uu, log⁡b(au)=u⋅log⁡ba\log_b(a^u)=u\cdot\log_b a. Combining,

u⋅log⁡ba=log⁡bxu\cdot\log_b a=\log_b x

It remains to justify that we may divide by log⁡ba\log_b a. Since a≠1a\neq1, the injectivity of log⁡b\log_b, §4, section 4.3, together with log⁡b1=0\log_b1=0, §5, section 4.3, guarantees log⁡ba≠0\log_b a\neq0: if it were log⁡ba=0=log⁡b1\log_b a=0=\log_b1, injectivity would force a=1a=1, contrary to the hypothesis. Dividing,

u=log⁡bxlog⁡bau=\frac{\log_b x}{\log_b a}

and since u=log⁡axu=\log_a x by definition, the formula is proved. ■\blacksquare


With the four laws available, it is time to exercise their combined handling before the close of the section. That is the content of §5.

Exercise 1. Express log⁡a(x2yz3)\log_a\left(\dfrac{x^2\sqrt{y}}{z^3}\right) as a combination of log⁡ax\log_a x, log⁡ay\log_a y, log⁡az\log_a z, for x,y,z>0x,y,z>0.

Solution. By the quotient law (§2), log⁡a(x2yz3)=log⁡a(x2y)−log⁡a(z3)\log_a\left(\dfrac{x^2\sqrt y}{z^3}\right)=\log_a(x^2\sqrt y)-\log_a(z^3). By the power law (§3), log⁡a(z3)=3log⁡az\log_a(z^3)=3\log_a z. By the product law (§1), log⁡a(x2y)=log⁡a(x2)+log⁡a(y1/2)\log_a(x^2\sqrt y)=\log_a(x^2)+\log_a(y^{1/2}), and by the power law again, log⁡a(x2)=2log⁡ax\log_a(x^2)=2\log_a x and log⁡a(y1/2)=12log⁡ay\log_a(y^{1/2})=\tfrac12\log_a y. Combining everything,

log⁡a(x2yz3)=2log⁡ax+12log⁡ay−3log⁡az\log_a\left(\frac{x^2\sqrt y}{z^3}\right)=2\log_a x+\frac12\log_a y-3\log_a z

Exercise 2. Knowing that log⁡102≈0.301\log_{10}2\approx0.301 and log⁡103≈0.477\log_{10}3\approx0.477, compute log⁡1012\log_{10}12.

Solution. 12=22⋅312=2^2\cdot3. By the product and power laws, log⁡1012=log⁡10(22)+log⁡103=2log⁡102+log⁡103≈2(0.301)+0.477=1.079\log_{10}12=\log_{10}(2^2)+\log_{10}3=2\log_{10}2+\log_{10}3\approx2(0.301)+0.477=1.079.

Exercise 3. Compute log⁡432\log_4 32 using change of base to base 22.

Solution. By §4, log⁡432=log⁡232log⁡24\log_4 32=\dfrac{\log_2 32}{\log_2 4}. Since 32=2532=2^5 and 4=224=2^2, the inversion identity of §2, section 4.3, gives log⁡232=5\log_2 32=5 and log⁡24=2\log_2 4=2. Hence log⁡432=52\log_4 32=\dfrac52.

Exercise 4. Solve log⁡a(x+3)+log⁡a(x−3)=log⁡a16\log_a(x+3)+\log_a(x-3)=\log_a 16, specifying the domain.

Solution. The domain requires x+3>0x+3>0 and x−3>0x-3>0 simultaneously, §3, section 4.3, that is, x>3x>3. By the product law (§1), the left side is log⁡a((x+3)(x−3))=log⁡a(x2−9)\log_a\big((x+3)(x-3)\big)=\log_a(x^2-9). The equation reduces to log⁡a(x2−9)=log⁡a16\log_a(x^2-9)=\log_a16, and by the injectivity of log⁡a\log_a, §4, section 4.3, x2−9=16x^2-9=16, whence x2=25x^2=25, that is, x=±5x=\pm5. Combining with the domain x>3x>3, the only valid solution is x=5x=5; x=−5x=-5 is discarded, not because of an algebraic error, but because it falls outside the original domain, a reminder that each step of this kind requires checking the restriction one started with.


With the laws exercised, what remains is to situate what this section achieved within the broader history of why the logarithm came to exist. That is the content of §6, with which the section closes.

What the four laws of this section have in common, seen from sufficient distance, is that they all convert a multiplicative operation, product, quotient, power, into an additive one, sum, difference, product by a scalar. It is not a coincidence among four loose rules: it is a single idea, that of the homomorphism of §8 of section 4.1, seen from the side of the logarithm instead of the side of the exponential. log⁡a\log_a translates the world of the product into the world of the sum, and that translation is exact, reversible and lossless, as reversible as the very relation of inverse that defined the logarithm in §4.3.

That translation is not a modern curiosity. John Napier, when publishing his tables of logarithms in 1614, Mirifici Logarithmorum Canonis Descriptio, was not pursuing an axiomatic construction like this chapter’s, but something far more urgent for his time: to reduce the error-prone labor of multiplying many-digit numbers by hand. With a table of logarithms, multiplying two numbers was reduced to adding their logarithms and looking up, in the same table, which number corresponded to that sum, exactly the law of §1, used in the direction opposite to how it was proved here. That economy of calculation, before any electronic calculator, accelerated for more than three centuries astronomy, navigation and engineering; the slide rule, a standard instrument until well into the twentieth century, is nothing other than this same product law, materialized in a sliding ruler.

What is remarkable, with the distance of this book, is that Napier arrived at that tool without having, or needing, the rigorous construction of section 4.1: he worked with geometric and arithmetic progressions compared term by term, an operational intuition that worked long before the completeness axiom existed to justify it fully. The same distance that separated Euler’s use of irrational axa^x, mentioned in the close of §12, section 4.1, from its rigorous foundation, also separates Napier’s tables from the Proposition of §1 of this section: almost three hundred years of fruitful use before the object used was, as here, completely justified from its foundations.


With the laws of logarithms complete, §4.5 combines them with the injectivity of §4.3 to solve, systematically, equations and inequalities that mix exponentials and logarithms, the type of problem that historically motivated the very invention of the logarithm.