Exponential and Logarithmic Functions (Exponential- und Logarithmusfunktionen)

40 min · 12 drills

In this topic the five claims usually sit under a process: a balance earning interest, a revenue that grows, a quantity that decays, or an equation with the unknown in the exponent. Whether a claim holds is nearly always settled by one of a few questions. Is something multiplied or added? Does a quantity depend on the starting value (Anfangswert) or only on the factor? Which logarithm rule actually exists, and which one only looks like a rule?

The formula sheet (Formelsammlung) offers exactly two lines on this topic, the derivatives (ln⁡x)′=1x and (ex)′=ex. It holds no rule for logarithms or powers, no formula for doubling time or half-life and no compound interest formula. The sum formula for the geometric series, which is on the sheet as well, belongs to financial mathematics and is of no help here. Everything else in this article is therefore yours to bring along. Values of logarithms that a task needs, such as ln⁡2, are printed in the task; you should know ln⁡1=0 and ln⁡e=1 yourself.

The task format works as the article on logic explains: a claim is correct if it holds in every admissible case, and a single case in which it fails refutes it. For arithmetic rules such a case is especially cheap. Plug in numbers for which both sides can be worked out in your head.

Exponential growth multiplies by the same factor at every step

Exponential growth (exponentielles Wachstum)

A quantity grows or decreases exponentially if it is multiplied by the same factor a in every unit of time, where a>0 and a≠1. After t units of time it equals

N(t)=N0⋅at

Here N0=N(0) is the starting value (Anfangswert), because a0=1.

Growth factor (Wachstumsfaktor)

The factor a belongs to exactly one unit of time. In this article p stands for a percentage written as a decimal: 25% is p=0.25. If the quantity rises by p per unit of time, then a=1+p; if it falls by p, then a=1−p. A rise of 25% thus gives a=1.25, a loss of 15% gives a=0.85.

FactorBehaviourExample
a>1growtha=1.25: plus 25% per unit of time
0<a<1decrease, also: decay (Zerfall)a=0.85: minus 15% per unit of time

A quantity of 500 that rises by 125 in the first year can go on growing in two ways. Linearly, another 125 are added in every further year. Exponentially with a=1.25, every year 25% of the quantity present at that moment are added. The values are rounded to two decimal places.

Year tlinear 500+125texponential 500⋅1.25tIncrease in the year up to t (exponential)
0500500
1625625125
2750781.25156.25
3875976.56195.31
410001220.70244.14

In the linear column the quantity rises by the same amount every year. In the exponential column every value is 1.25 times the previous one, and dividing a value by its predecessor always gives a:

N(t+1)N(t)=afor every t

The increases in the last column, on the other hand, get larger. The increase in one year is N(t)⋅(a−1), a fixed share of a quantity that is itself growing. For 0<a<1 the term a−1 is negative: the loss is a fixed share of a quantity that is shrinking, and so it gets smaller from year to year.

Your turn1 of 12

A stock of raw material of 800 tonnes decreases by 25% every year. By how many tonnes does it decrease in the third year?

Percentages multiply, they do not add

Because every unit of time contributes its own factor, the total change is the product of the individual factors. Three calculations follow from this.

  • Several equal changes. If a price falls three times by 10% each, the factor left is 0.93=0.729. The price has dropped by 27.1%, not by 30%, because each cut acts on a price that is already lower.
  • There and back. Plus 10% followed by minus 10% gives 1.1⋅0.9=0.99, a loss of 1%. After plus 25% it takes a cut of 20% to get back to the starting value, since 1.25⋅0.8=1.
  • Shorter periods. Earning 3% per quarter gives more than 12% per year: 1.034≈1.1255, about 12.55%. The surplus is interest on the interest of the earlier quarters.
Compound interest (Zinseszins)

Compound interest arises when interest is credited to the capital and from then on earns interest itself. At the interest rate p per period a capital K0 grows in n periods to Kn=K0⋅(1+p)n. This is exponential growth with the factor a=1+p.

The exponent counts units of time, and it may also be a fraction. If a quantity grows by the factor 1.4641 over four years, the factor for a single year is 1.46411/4=1.1, so plus 10% per year. A quarter of the total increase, 46.41%4≈11.60%, would be too much, because the compounding of the four years is contained in it.

factor for k units of time=ak

Your turn2 of 12

A price rises twice in a row by 10% each time. By what percentage is it then above the starting price?

The graph stays above the x-axis and passes through the point (0 | 1)

Exponential function (Exponentialfunktion)

A function f(x)=ax with a base a>0 and a≠1 is called an exponential function. The variable sits in the exponent, not in the base; that is what separates it from a power function (Potenzfunktion) such as x3.

Asymptote (Asymptote)

A line is called an asymptote of a graph if the graph comes arbitrarily close to it. This can happen for very large or very small x (horizontal asymptote) or when x approaches a fixed point (vertical asymptote). An example of the second case is ln⁡x for x close to 0: there the graph approaches the y-axis.

The sketch shows 1.5x and (23)x. The second base is the reciprocal of the first, so the second graph is the mirror image of the first in the y-axis.

The rising curve 1.5ˣ (growth) and the dashed falling curve (2/3)ˣ (decay) are mirror images in the y-axis and meet at (0 | 1). Both approach the x-axis without ever reaching it: it is an asymptote. The rising curve 1.5ˣ (growth) and the dashed falling curve (2/3)ˣ (decay) are mirror images in the y-axis and meet at (0 | 1). Both approach the x-axis without ever reaching it: it is an asymptote.
  • Always positive. For every real x, ax>0. That is why ax has no zero (Nullstelle), and a quantity N0⋅at with a positive starting value stays above zero for all time.
  • The point (0∣1). Because a0=1, the point (0∣1) lies on every such graph; there it meets the y-axis, the vertical axis (senkrechte Achse). For N0⋅ax it is the point (0∣N0).
  • The x-axis as asymptote. The x-axis is the horizontal axis (waagrechte Achse). For a>1 the graph comes arbitrarily close to it towards the left, for 0<a<1 towards the right.
  • Strictly monotonic (streng monoton). For a>1 the function ax increases, for 0<a<1 it decreases. Every value is therefore taken at most once, and au=av implies u=v.
  • Range (Wertebereich). ax takes exactly the positive numbers as values.

Two changes to the term look alike and do different things. 1.5x−2 lowers the graph by 2; the asymptote becomes the line y=−2, and unlike 1.5x this graph has a zero. 1.5x−2 shifts it 2 to the right. Since 1.5x−2=1.5x2.25, all function values are divided by 2.25 in the process, and the asymptote stays the x-axis.

Your turn3 of 12

Which line is an asymptote of the graph of f(x)=2⋅0.5x+3?

Doubling time and half-life depend only on the factor

Doubling time (Verdopplungszeit)

The doubling time is the span of time after which an exponentially growing quantity is twice as large as at the start of that span.

Half-life (Halbwertszeit)

The half-life is the span of time after which half of an exponentially decreasing quantity is left.

We are looking for a span T with N(t+T)=2⋅N(t). Inserting the formula gives N0⋅at⋅aT on the left and 2⋅N0⋅at on the right. Dividing both sides by N0⋅at leaves aT=2. Neither the starting value nor the moment t appears in this condition; only a decides it. The same holds for the half-life, with aT=12.

The starting value determines the height of the curve instead: N(t) is proportional to N0. A quantity with three times the starting value lies three times as high at any instant and takes exactly as long to double.

If the half-life T is known, the decrease is most conveniently written like this:

N(t)=N0⋅0.5 t/T

As a check, insert multiples of T: at t=T the factor is 0.5, at t=3T it is 0.53=0.125. In general the same construction works for any factor q:

N(t)=N0⋅q t/T⟹after every span T the quantity is multiplied by q

An example: a radioactive sample has an activity of 640 becquerels and a half-life of 12 minutes.

Time in minutes012243648
Activity in becquerels6403201608040
Decrease in the preceding 12 min3201608040

Here the decrease per span halves along with the quantity as well. The formula also holds between the columns. After 4 minutes, a third of the half-life, the factor is 0.51/3≈0.794; about 508 becquerels are left, not the 533.3 that an even decrease by a third of 320 would give. Three such thirds together have to give the factor 0.5, and indeed 0.7943≈0.5.

Your turn4 of 12

A quantity halves every 6 years. What share of the starting value is still there after 24 years?

Ever more frequent compounding leads to the number e and the continuous growth rate

Nominal interest rate (Nominalzinssatz)

The nominal annual interest rate p is the agreed rate that, with m interest dates per year, is split into m equal parts: each date earns pm.

Effective interest rate (Effektivzinssatz)

The effective annual interest rate is the growth the capital actually experiences in one year. With m interest dates the annual factor is (1+pm)m, so the effective rate is (1+pm)m−1.

For a nominal annual interest rate of 12%, that is p=0.12, this gives:

CompoundingRate per dateAnnual factorEffective annual rate
annually, m=112%1.1212%
semi-annually, m=26%1.062=1.123612.36%
quarterly, m=43%1.034≈1.125512.55%
monthly, m=121%1.0112≈1.126812.68%
continuouslye0.12≈1.127512.75%

As soon as interest is credited more than once a year, the effective rate lies above the nominal one, and it rises with m. It does not rise beyond every bound, though: the annual factors approach a fixed value. The same phenomenon defines the number e.

Euler's number (Eulersche Zahl)

The number e≈2.71828 is the value that (1+1m)m approaches as m grows. For m=1 the result is 2, for m=12 about 2.613, for m=1000 about 2.7169.

(1+pm)m  ⟶  epfor m→∞

Continuous compounding (stetige Verzinsung) is this limiting case: interest is credited at every instant, and after t years the capital equals K0⋅ept. For every finite m the annual factor stays below ep. The effective annual rate under continuous compounding is ep−1, so about 12.75% for p=0.12.

Many models write the quantity with base e, as N(t)=N0⋅ekt. This is not a new type of growth but the same form N0⋅at in different notation, since ekt=(ek)t. The number k is called the continuous growth rate (stetige Wachstumsrate); the factor per unit of time is a=ek, and conversely k=ln⁡a, with the natural logarithm from the next section.

N0⋅ekt=N0⋅atwitha=ek,k=ln⁡a
  • Sign. k>0 means growth and k<0 decrease, because ek>1 holds exactly for k>0.
  • The rate is not the percentage. For e0.1t the factor per year is e0.1≈1.1052. The quantity therefore grows by about 10.52% per year, not by 10%.
  • Rate and percentage are close only for small k. e0.01−1≈0.01005 is almost 0.01, but e1−1≈1.72 is far from 1.

Your turn5 of 12

For a quantity, N(t)=N0⋅e0.2t, t in years. By what percentage does it grow in one year?

The logarithm reverses raising to a power

Logarithm (Logarithmus)

For a base b>0 with b≠1 and a number x>0, log⁡bx is the exponent to which b must be raised to give x:

log⁡bx=y  ⟺  by=x

Thus log⁡381=4, because 34=81, and log⁡100.01=−2, because 10−2=0.01.

Natural and common logarithm (natürlicher und dekadischer Logarithmus)

ln⁡x is the logarithm to base e, the natural logarithm. lg⁡x stands for the logarithm with base 10, the common one. We have ln⁡1=0 and ln⁡e=1, and likewise lg⁡1=0 and lg⁡10=1.

Inverse function (Umkehrfunktion)

Two functions are inverses of each other if each undoes what the other does. For ex and ln⁡x this means:

ln⁡(ex)=x for every x,eln⁡x=x for x>0

The condition x>0 appears only with the second equation. ln⁡x exists only for positive x, while ex has a value at every real number x, and a positive one.

A function and its inverse swap the roles of x and y. Their graphs are therefore mirror images in the line y=x, the bisector of the first quadrant (Winkelhalbierende, also called erste Mediane): the point (0∣1) on the graph of ex becomes the point (1∣0) on the graph of ln⁡x.

Graphs of eˣ and ln x, mirrored in the dashed line y = x: the point (0 | 1) on eˣ becomes (1 | 0) on ln x. eˣ lies entirely above the line, ln x entirely below it and drops steeply near the y-axis. Graphs of eˣ and ln x, mirrored in the dashed line y = x: the point (0 | 1) on eˣ becomes (1 | 0) on ln x. eˣ lies entirely above the line, ln x entirely below it and drops steeply near the y-axis.

Almost everything asked about ln⁡x follows from this reflection.

  • Domain (Definitionsbereich). Positive numbers only. The values of ex are exactly the positive numbers, and they form the domain of ln⁡. ln⁡0 and ln⁡(−3) do not exist, because no power of e is zero or negative.
  • Zero and sign. ln⁡x=0 holds exactly at x=1. Between 0 and 1, ln⁡x is negative, because a number below 1 calls for a negative exponent, for instance ln⁡0.5≈−0.693. For x>1, ln⁡x is positive.
  • No upper limit, but very flat. ln⁡x is strictly increasing and takes every real value; the value 10, for instance, belongs to x=e10, about 22,000. The curve becomes very flat: ln⁡1000≈6.9.
  • On both sides of the mirror line. The graph of ln⁡x lies entirely below the line y=x, that of ex entirely above it: ln⁡x<x<ex for x>0.
  • The y-axis as asymptote. For x close to zero, ln⁡x falls below every bound.

The two derivatives on the formula sheet fit this picture. (ex)′=ex means: at every point the slope equals the function value. (ln⁡x)′=1x is positive for x>0, confirming the increase, and gets smaller as x grows, which is why the graph of ln⁡x keeps flattening.

Your turn6 of 12

Which statement about ln⁡x and ex is true?

Powers with the same base are combined through their exponents

The power rules hold for every positive base, not only for e. Here they are written for e and arbitrary real exponents a and b, with a check for a=3 and b=2:

RuleCheck with a=3, b=2
ea⋅eb=ea+be3⋅e2=e5
eaeb=ea−be3e2=e1
(ea)b=ea⋅b(e3)2=e3⋅e3=e6
e−a=1eae−2=1e2

Three common mistakes look like rules and are not:

Common mistakeCounter-check
ea⋅eb=ea+eba=b=1: left e2≈7.39, right 2e≈5.44
(ea)b=eaba=3, b=2: left e6, right e9
e−a=−eaa=1: left 1e≈0.37, right −e≈−2.72

A check with concrete numbers settles such a claim in seconds. Pick numbers for which the two sides do not agree by coincidence. With a=b=2, for example, ab=a⋅b=4, and the second mistake would look right. A check that works out proves nothing; one that does not work out refutes.

For a sum eu+ev there is no power rule; it stays as it is. And because ex is strictly increasing, eu=ev always implies u=v.

Your turn7 of 12

Which expression equals (ex)3ex−2 for every real number x?

Logarithms turn products into sums, and the change of base follows from that

Because the logarithm counts exponents, it translates the power rules: multiplying becomes adding, dividing becomes subtracting, raising to a power becomes multiplying. For positive numbers u, v and every real number r:

ln⁡(u⋅v)=ln⁡u+ln⁡v ln⁡(uv)=ln⁡u−ln⁡v ln⁡(ur)=r⋅ln⁡u

The same rules hold for lg⁡ and for every logarithm log⁡b. With ln⁡3≈1.0986 and ln⁡7≈1.9459 you get, for instance, ln⁡21=ln⁡3+ln⁡7≈3.0445, ln⁡73=ln⁡7−ln⁡3≈0.8473 and ln⁡49=2ln⁡7≈3.8918. The quotient rule also gives ln⁡1u=−ln⁡u: reciprocals have opposite logarithms.

Common mistakeCheck with u=7, v=3What holds instead
ln⁡(u+v)=ln⁡u+ln⁡vleft ln⁡10≈2.303, right ≈3.045there is no rule for ln⁡(u+v)
ln⁡u⋅ln⁡v=ln⁡(uv)left ≈2.138, right ≈3.045ln⁡(uv) is the sum ln⁡u+ln⁡v
ln⁡(u2)=(ln⁡u)2left ≈3.892, right ≈3.787the exponent becomes a factor: 2ln⁡u

A logarithm to another base is computed via ln⁡. If y=log⁡bx, then by=x. Taking logarithms of both sides gives y⋅ln⁡b=ln⁡x, and this yields the change of base (Basiswechsel):

log⁡bx=ln⁡xln⁡b

An example: log⁡37=ln⁡7ln⁡3≈1.94591.0986≈1.771. That fits, because 31=3<7<9=32.

Your turn8 of 12

Which equation holds for all positive numbers u and v?

Exponential equations are solved by taking logarithms and checked by bracketing

When the unknown sits in the exponent, two routes lead to the solution.

  1. Make the bases equal. If the right-hand side can be written as a power of the same base, compare the exponents. 52x=125=53 becomes 2x=3 and hence x=1.5. This is allowed because an exponential function takes every value only once.
  2. Take logarithms. Otherwise apply ln⁡ to both sides and bring the exponent to the front with the power rule: ax=c becomes x⋅ln⁡a=ln⁡c.
ax=c  ⟹  x=ln⁡cln⁡a(a>0, a≠1, c>0)

For 3x=50 with ln⁡50≈3.912 and ln⁡3≈1.0986 this gives x≈3.561.

  • Bracketing. Since 33=27<50<81=34 and 3x increases, the solution lies between 3 and 4. This check needs no logarithms and is often enough for the decision.
  • Solvable or not. ax=c has exactly one real solution for every c>0 and none for c≤0, because ax is always positive.
  • Multiplying the right-hand side. Replacing c by k⋅c shifts the solution by ln⁡kln⁡a, that is by a constant. For 10x=c, for example, the factor 100 shifts the solution by 2, the factor 110 by −1.
  • Logarithmic equations are solved the other way round, by raising the base to both sides. From ln⁡x=−2 it follows that x=e−2≈0.135, from lg⁡x=−1 that x=10−1=0.1.

Doubling time and half-life come out the same way. aT=2 becomes T=ln⁡2ln⁡a; aT=12 becomes T=−ln⁡2ln⁡a, which is positive because ln⁡a<0 for a decrease. T comes out in the unit of time that a belongs to: if a is the factor of a quarter, T counts quarters, and only T4 is in years. For a quantity growing by 50% per year, a=1.5, and with ln⁡1.5≈0.4055 this gives T≈0.69310.4055≈1.71 years. Bracketing confirms it, 1.51=1.5<2<2.25=1.52. If the question is in which year the doubling falls, count: the first year runs from t=0 to t=1, so the moment t≈1.71 lies in the second year.

Your turn9 of 12

Between which two whole numbers does the solution of 5x=70 lie?

Your turn10 of 12

The equation 3x=c with c>0 has the solution x0. What is the solution of 3x=9c?

The higher growth factor catches up with any lead

Firm A sells 1000 units today and raises its sales by 20% per year; firm B sells 7000 units and raises them by 4% per year. Today A adds 0.2⋅1000=200 units and B adds 0.04⋅7000=280 units: at the moment B grows more in absolute terms, although its percentage is smaller. The two are level when

1000⋅1.2t=7000⋅1.04t  ⟺  (1.21.04)t=7  ⟺  t=ln⁡7ln⁡1.2−ln⁡1.04≈1.94590.1431≈13.60

So A overtakes B in the fourteenth year, and after that the gap keeps widening, because the ratio (1.21.04)t grows without bound.

An exponential function with a base above 1 eventually overtakes every power function xn, even if the latter stays ahead for a long time. On the way the graphs can intersect more than once. For 2x and x3 a table of values shows where:

x12591011
2x243251210242048
x31812572910001331

Between 1 and 2 the order changes for the first time, between 9 and 10 for the second time; more precisely the intersections lie at about 1.37 and 9.94. After that 2x stays ahead for good. When x moves on by 1, 2x doubles, while x3 grows by the factor (x+1x)3, and that factor is below 2 from x=4 on.

Your turn11 of 12

Firm A sells 2000 units today and raises its sales by 10% per year; firm B sells 5000 units and raises them by 2% per year. In which year does A overtake B? Use ln⁡2.5≈0.916, ln⁡1.1≈0.095 and ln⁡1.02≈0.020.

How to check a claim about growth and logarithms

  1. Insert t=0. This gives the starting value and shows whether you have read the formula correctly.
  2. Factor or amount? In an exponential process the amount added or lost per span changes; only the factor stays the same.
  3. What depends on the starting value? Doubling time, half-life and percentage do not; every single value N(t) and every absolute increase do, and proportionally so.
  4. Whole-number points. Powers with whole exponents are easy to compute and bracket every solution. This checks a solution, the year of an event and the location of intersections.
  5. Check rules with a number. An arithmetic rule that fails for a single unluckily chosen number is false. Use numbers whose logarithms or powers you know, such as 1, e, e2.

Your turn12 of 12

Someone checks the claim “lg⁡(u+v)=lg⁡u+lg⁡v for all u,v>0” with u=v=2 and gets lg⁡4 on both sides. What follows?

Common mistakes

  • Adding percentages. Several changes are multiplied as factors: three times minus 10% is minus 27.1%, and 3% per quarter is more than 12% per year.
  • Taking a fraction of the factor. A quarter of a unit of time goes with a1/4, not with a4.
  • Taking the continuous rate for the percentage. For ekt the factor per unit of time is ek, and ek−1 is close to k only when k is small.
  • Splitting a logarithm that cannot be split. There is no rule for ln⁡(u+v), and ln⁡u⋅ln⁡v is not ln⁡(uv).
  • Bringing the exponent down without a logarithm. ax=c is not a⋅x=c. Bracketing between two whole-number powers exposes the mistake at once.
  • Letting an exponential quantity fall to zero. N0⋅at stays positive when the starting value is positive; the time axis is an asymptote, not a zero.
  • Confusing a shift in the exponent with a shift up or down. ax−c shifts horizontally, ax−c lowers the graph and its asymptote with it.

Key takeaways

  • Exponential means the same factor per unit of time. N(t)=N0⋅at, growth for a>1, decrease for 0<a<1. From step to step the increase changes, the factor does not.
  • Doubling time and half-life depend only on a: T=ln⁡2ln⁡a and T=−ln⁡2ln⁡a respectively, measured in the unit of time of a. A different starting value scales the values, not the times.
  • The graph of ax is positive, passes through (0∣1) and has the x-axis as asymptote.
  • The number e is the limit of (1+1m)m; ekt has the factor ek per unit of time.
  • Logarithm means exponent: ln⁡ is defined only for positive numbers and undoes ex.
  • Equations with the unknown in the exponent are solved by taking logarithms and checked by bracketing between whole-number powers.
ln⁡(uv)=ln⁡u+ln⁡v,ln⁡uv=ln⁡u−ln⁡v,ln⁡ur=rln⁡u,ax=c⇒x=ln⁡cln⁡a

Now practise it

Exponential and Logarithmic Functions (Exponential- und Logarithmusfunktionen)

Trainer · Mathematics

Practise
Next: Derivatives and Optimization (Ableitungen und Extremwertaufgaben)
0/12

Was this article helpful?