Skip to content

Chapter 4 — Exponential and Logarithmic Functions

4.6 - Applications of Exponential and Logarithmic Functions

The financial question that opened §4.2, capital that grows by interest, was not a pedagogical pretext for arriving at ee: it is, in itself, the first particular case of a much more general phenomenon, which this section sets out to characterize precisely before applying it.


Up to here, the chapter constructed axa^x and log⁡ax\log_a x as self-sufficient mathematical objects, without yet asking why they should appear when modeling a phenomenon of the world. This section closes that circle: it identifies, with a proof, not with an analogy, what exact property of a process obliges it to be described with an exponential function, and applies that characterization to three families of phenomena: financial growth, decay, and magnitudes measured on logarithmic scales.

§1. The functional characterization of proportional growth

Section titled “§1. The functional characterization of proportional growth”

When it is said that a magnitude “grows proportionally to itself”, one immediately thinks of an instantaneous rate of change, the terrain of the derivative, which this book does not have. But there is an entirely algebraic version of the same idea, which mentions no instantaneous rate: that the ratio of growth between two instants depends only on how much time has elapsed, not on at what moment one started to measure. This version can indeed be characterized precisely, and with the tools already constructed.

Proposition (characterization of proportional growth). Let Q:R→(0,∞)Q:\mathbb{R}\to(0,\infty) be such that, for all t,h∈Rt,h\in\mathbb{R},

Q(t+h)=Q(t)⋅Q(h)Q(0)Q(t+h)=Q(t)\cdot\frac{Q(h)}{Q(0)}

that is, the multiplicative factor separating Q(t+h)Q(t+h) from Q(t)Q(t) depends only on hh, and such that QQ is continuous at 00 in the sense of the Lemma of §12, section 4.1. Then there exists a>0a>0 such that

Q(t)=Q(0)⋅atfor all t∈R,with a=Q(1)Q(0)Q(t)=Q(0)\cdot a^t \qquad \text{for all } t\in\mathbb{R}, \qquad \text{with } a=\frac{Q(1)}{Q(0)}

Proof. Let R(t):=Q(t)/Q(0)R(t):=Q(t)/Q(0), well defined since Q(0)>0Q(0)>0; from the hypothesis itself with t=h=0t=h=0 it follows Q(0)=Q(0)2/Q(0)Q(0)=Q(0)^2/Q(0), consistent but not informative; that Q(0)>0Q(0)>0 is part of the hypothesis Q:R→(0,∞)Q:\mathbb{R}\to(0,\infty). Then R(0)=1R(0)=1, and

R(t+h)=Q(t+h)Q(0)=Q(t)Q(0)⋅Q(h)Q(0)=R(t)R(h)R(t+h)=\frac{Q(t+h)}{Q(0)}=\frac{Q(t)}{Q(0)}\cdot\frac{Q(h)}{Q(0)}=R(t)R(h)

so that RR is a homomorphism R→(0,∞)\mathbb{R}\to(0,\infty), continuous at 00, because QQ is, and dividing by the constant Q(0)Q(0) does not alter continuity at 00. Let a:=R(1)=Q(1)/Q(0)>0a:=R(1)=Q(1)/Q(0)>0, and let fa(t)=atf_a(t)=a^t be the extension already constructed in §6 or §7, section 4.1, or, if a=1a=1, the constant function f1≡1f_1\equiv1, a trivial homomorphism and continuous at 00 without need of additional proof. In any case, faf_a is a homomorphism continuous at 00, §8 and Lemma of §12, section 4.1. By iterated homomorphism, R(n)=an=fa(n)R(n)=a^n=f_a(n) for all n∈Zn\in\mathbb{Z}, and by the same root-uniqueness argument as in Exercise 2 of §12, section 4.1, applied here to RR instead of to (ax)y(a^x)^y, R(m/n)=am/n=fa(m/n)R(m/n)=a^{m/n}=f_a(m/n) for every rational m/nm/n. By the Uniqueness Lemma for homomorphisms continuous at 00, §12, section 4.1, R=faR=f_a on all of R\mathbb{R}. Hence Q(t)=Q(0)R(t)=Q(0)atQ(t)=Q(0)R(t)=Q(0)a^t. ■\blacksquare


With the model justified in general, its first concrete application corresponds, the one that motivated, already in §4.2, the very construction of ee: compound interest, and what happens in the limit of ever more frequent compounding.

§2. Compound interest and continuous compounding

Section titled “§2. Compound interest and continuous compounding”

A capital PP invested at an annual rate rr, compounded nn times per year, satisfies exactly the hypothesis of the Proposition of §1: in each subinterval of 1/n1/n years, the capital is multiplied by the same factor (1+r/n)(1+r/n), regardless of at what moment of the year that subinterval begins. After tt years, that is, ntnt compoundings, the accumulated capital is

An(t)=P(1+rn)ntA_n(t)=P\left(1+\frac{r}{n}\right)^{nt}

a formula already completely justified by what was constructed in 4.1: the base 1+r/n>01+r/n>0 and the exponent ntnt can be any real, without restriction.

The question that opened §4.2, what happens when nn grows without bound, has, for the particular case P=1P=1, r=1r=1, t=1t=1, an answer already completely proved: An(1)=(1+1/n)n=anA_n(1)=(1+1/n)^n=a_n, and an→ea_n\to e in the rigorous sense of §4 of that section. For a general rate rr, the same phenomenon occurs, ever more frequent compounding stabilizes at a limit value, and that value is ere^r:

lim⁡n→∞(1+rn)n=er\lim_{n\to\infty}\left(1+\frac{r}{n}\right)^n=e^r

A complete proof of this fact for arbitrary rr repeats, step by step, the technique of the two squeezing sequences of §4.2, with an(r)=(1+r/n)na_n(r)=(1+r/n)^n and a sequence analogous to bnb_n, and is not repeated here for economy: the reader who followed §4.2 has all the tools to reconstruct it. What does correspond to be established, as a working model for the rest of this section, is the resulting formula:

Continuous compounding model. A(t)=PertA(t)=Pe^{rt}.


The same model, with a∈(0,1)a\in(0,1) instead of a>1a>1, describes processes that decrease instead of grow, the most studied case of which is radioactive decay. It is the content of §3.

A radioactive substance loses mass in such a way that it satisfies, empirically, the same hypothesis as §1: the fraction of mass lost in an interval depends only on its duration. By the Proposition of §1, with 0<a<10<a<1,

M(t)=M0⋅atM(t)=M_0\cdot a^t

Definition. The half-life TT of a substance is the time such that M(T)=M0/2M(T)=M_0/2.

Example. A substance with half-life T=10T=10 years satisfies M0a10=M0/2M_0 a^{10}=M_0/2, that is, a10=1/2a^{10}=1/2, whence a=(1/2)1/10a=(1/2)^{1/10}, using the corollary (ax)y=axy(a^x)^y=a^{xy}, section 4.1 §12, with exponent 1/101/10. The complete model is M(t)=M0(12)t/10M(t)=M_0\left(\dfrac12\right)^{t/10}: every 1010 years, whichever they are, the mass is reduced to exactly half, a direct consequence of the Proposition of §1, not a coincidence of the example.

Worked exercise. How long does it take that same substance to be reduced to one tenth of its initial mass?

Solution. We seek tt with M(t)=M0/10M(t)=M_0/10, that is, (12)t/10=110\left(\dfrac12\right)^{t/10}=\dfrac1{10}. By the Principle of direct application, §1, section 4.5, t10=log⁡1/2110\dfrac{t}{10}=\log_{1/2}\dfrac1{10}, whence, by change of base to base 1010, §4, section 4.4, log⁡1/2(1/10)=log⁡10(1/10)log⁡10(1/2)=−1log⁡101−log⁡102=−1−log⁡102=1log⁡102≈3.32\log_{1/2}(1/10)=\dfrac{\log_{10}(1/10)}{\log_{10}(1/2)}=\dfrac{-1}{\log_{10}1-\log_{10}2}=\dfrac{-1}{-\log_{10}2}=\dfrac1{\log_{10}2}\approx3.32. Hence t≈10×3.32=33.2t\approx10\times3.32=33.2 years.


The last type of application does not model a process that changes in time, but a way of measuring magnitudes whose range of possible values is too wide for an ordinary additive scale. It is the content of §4.

Certain magnitudes, the energy released by an earthquake, the concentration of hydrogen ions in a solution, the intensity of a sound, vary, in nature, over several orders of magnitude: an earthquake can release a thousand, a million or a billion times more energy than another. Representing that variation on an ordinary additive scale, where each unit adds a fixed quantity, makes the complete range almost impossible to graph or to communicate precisely. The solution, in each of these cases, is the same: to define the scale as the logarithm of the physical magnitude, so that multiplying the original magnitude by ten, adding a complete order of magnitude, adds only one unit to the scale. It is the product law of §4.4 §1, working in the inverse direction: what was multiplicative in the phenomenon becomes additive in the scale that measures it.

pH scale. pH=−log⁡10[H+]\text{pH}=-\log_{10}[\text{H}^+], where [H+][\text{H}^+] is the concentration of hydrogen ions in moles per liter. A solution with [H+]=10−7[\text{H}^+]=10^{-7} has pH=7\text{pH}=7, neutral; one with [H+]=10−3[\text{H}^+]=10^{-3} has pH=3\text{pH}=3, acidic, and its ion concentration is 10410^4 times greater than that of the neutral solution, although the difference on the scale is only 44 units.

Richter scale (moment magnitude, in its simplified formulation). M=log⁡10(EE0)M=\log_{10}\left(\dfrac{E}{E_0}\right), where EE is the energy released and E0E_0 a fixed reference energy. Each unit of magnitude corresponds to a factor of 1010 in the energy released, a direct consequence of log⁡10\log_{10}, not an additional empirical datum.

Decibel scale. β=10log⁡10(II0)\beta=10\log_{10}\left(\dfrac{I}{I_0}\right), where II is the sound intensity and I0I_0 a reference intensity, the threshold of human hearing. The additional factor 1010 is a convention of the unit, not a different property from the two previous scales.

Worked exercise. An earthquake of magnitude 77 releases how much more energy than one of magnitude 55?

Solution. Let E7,E5E_7,E_5 be the respective energies. By the definition, 7−5=log⁡10(E7/E0)−log⁡10(E5/E0)=log⁡10(E7/E5)7-5=\log_{10}(E_7/E_0)-\log_{10}(E_5/E_0)=\log_{10}(E_7/E_5), quotient law, §4.4 §2. Hence E7/E5=102=100E_7/E_5=10^2=100: the earthquake of magnitude 77 releases one hundred times more energy than that of magnitude 55, not the “forty percent more” that a naive reading of the difference 7−57-5 relative to 55 might suggest.


Exercise 1. A capital of $10001000 is invested at 8%8\% annually, compounded quarterly. How much is accumulated after 55 years?

Solution. n=4n=4 (quarterly), r=0.08r=0.08, t=5t=5. By §2, A=1000(1+0.084)4×5=1000(1.02)20A=1000\left(1+\dfrac{0.08}{4}\right)^{4\times5}=1000(1.02)^{20}.

Exercise 2. A population of bacteria doubles every 33 hours. If initially there are 500500 bacteria, how many are there after 1010 hours?

Solution. By the Proposition of §1, with Q(0)=500Q(0)=500 and aa such that a3=2a^3=2, doubling every 33 hours, that is a=21/3a=2^{1/3}: Q(t)=500⋅(21/3)t=500⋅2t/3Q(t)=500\cdot\left(2^{1/3}\right)^t=500\cdot2^{t/3}. At t=10t=10: Q(10)=500⋅210/3Q(10)=500\cdot2^{10/3}.

Exercise 3. How long does it take a capital to double, at a continuous rate of 6%6\% annually?

Solution. We seek tt with Pe0.06t=2PPe^{0.06t}=2P, that is, e0.06t=2e^{0.06t}=2. By the Principle of direct application, §1, section 4.5, 0.06t=ln⁡20.06t=\ln2, whence t=ln⁡20.06t=\dfrac{\ln2}{0.06}.


The three families of applications of this section, financial growth, decay, measurement scales, are not loose examples chosen for their illustrative value: all three are instances of the same Proposition of §1, applied with different values of aa and, in the case of the scales, read from the side of the logarithm instead of the side of the exponential. That is, in short, the complete achievement of the chapter: not a catalog of formulas to memorize case by case, but a single structure, exponentiation extended to R\mathbb{R} by means of completeness, and its inverse, capable of explaining why that same structure reappears in compound interest, in nuclear physics and in seismology, without any of the three having been designed with the other two in mind.

What remains, to close the chapter with the depth it deserves, the place of ee and of the logarithm in the history of science, and not only in the arithmetic that constructed them, is a more extensive reflection that this section is not the place to develop.


The chapter continues, outside the original syllabus of six sections, with a more extensive epistemological and philosophical close, in the tradition of §2.9 and §3.6, which situates the exponential and the logarithm in their historical context and in their present use in the sciences. That is the content of §4.7.