Solve and interpret logarithmic models

Solve equations/inequalities, assess contextual solutions and read semi-log displays.

What to practice

  • Solve equations/inequalities, assess contextual solutions and read semi-log displays.

Unit 2 ends with three decisions that a calculator cannot make for you. When you solve an exponential or logarithmic equation, you decide which candidates the original statement actually allows. When you build a logarithmic model, you decide what the input and output measure and where the model makes sense. When you read a semi-log plot, you decide whether equal vertical spacing means equal differences or equal ratios. This chapter covers topics 2.13 to 2.15 of the AP Precalculus Course and Exam Description. Keep Chapter 11's rules in view: a logarithm needs a positive argument, and its base is positive and not 1.

Keep a restriction record before you solve

The inverse relationship lets you move between forms: log⁡b(m)=n\log_b(m)=n means exactly the same as bn=mb^n=m. Properties of exponents and logarithms let you combine terms. Some of those moves, however, change the set of allowed inputs. Rewriting log⁡b(u)+log⁡b(v)\log_b(u)+\log_b(v) as log⁡b(uv)\log_b(uv) is valid only when u>0u>0 and v>0v>0, yet the combined form also accepts inputs that make both factors negative. Rewriting 2log⁡b(x)2\log_b(x) as log⁡b(x2)\log_b(x^2) accepts negative values of xx. Use three steps every time:

  1. Record the restrictions from the original statement.
  2. Solve the transformed relation.
  3. Test every candidate in the original statement, not in the transformed one.

A candidate that satisfies the transformed relation but breaks a restriction is an extraneous solution. The framework also asks you to reject results that the context rules out, such as a negative time.

Worked example 12.1: reject a candidate that leaves the domain

Solve log⁡2(x)+log⁡2(x−6)=4\log_2(x)+\log_2(x-6)=4.

Restrictions. The arguments must be positive: x>0x>0 and x−6>0x-6>0, so x>6x>6.

Transform. The product property gives log⁡2(x(x−6))=4\log_2(x(x-6))=4, so x(x−6)=24=16x(x-6)=2^4=16. Then x2−6x−16=0x^2-6x-16=0, which factors as (x−8)(x+2)=0(x-8)(x+2)=0. The candidates are 8 and −2-2.

Check in the original. For x=8x=8: log⁡2(8)+log⁡2(2)=3+1=4\log_2(8)+\log_2(2)=3+1=4. For x=−2x=-2, log⁡2(−2)\log_2(-2) is undefined, so −2-2 is rejected. The solution is x=8x=8 only.

Figure 12.1 shows why. The left side exists only to the right of x=6x=6, and there its graph meets the line y=4y=4 once. The candidate −2-2 belongs to the polynomial equation, not to the original function.

Graph on axes labeled x and y. A dashed vertical line at x equals 6 marks the boundary of the domain, and the region with x less than or equal to 6 is hatched and labeled not in the domain. A solid curve, labeled y equals log base 2 of x plus log base 2 of x minus 6, starts near the dashed line far below the x-axis and rises to the right. A dotted horizontal line is labeled y equals 4. The curve and the line meet at one marked point, (8, 4). A cross on the x-axis at x equals negative 2 is labeled candidate negative 2, rejected.

Figure 12.1. The original equation has one solution; the second candidate comes from the transformed polynomial and lies outside the domain.

Worked example 12.2: use logarithms when the bases differ

Solve 5⋅2x=3x+15\cdot 2^x=3^{x+1} exactly. Both sides are positive for every real xx, so the equation adds no restriction. Take the natural logarithm of both sides:

ln⁡(5)+xln⁡(2)=(x+1)ln⁡(3)\ln(5)+x\ln(2)=(x+1)\ln(3)

Collect the xx terms: x(ln⁡(2)−ln⁡(3))=ln⁡(3)−ln⁡(5)x(\ln(2)-\ln(3))=\ln(3)-\ln(5), so

x=ln⁡(3)−ln⁡(5)ln⁡(2)−ln⁡(3)=ln⁡(5/3)ln⁡(3/2)≈1.260x=\frac{\ln(3)-\ln(5)}{\ln(2)-\ln(3)}=\frac{\ln(5/3)}{\ln(3/2)}\approx 1.260

Check: both sides are approximately 11.974 at this value. Any base would work; the natural logarithm keeps the algebra tidy. The same idea rewrites an exponential in a new base, bx=c(log⁡cb)xb^x=c^{(\log_c b)x}. For example, a quantity that doubles every 5 hours, Q(t)=Q0⋅2t/5Q(t)=Q_0\cdot 2^{t/5}, can also be written Q(t)=Q0e(ln⁡2/5)tQ(t)=Q_0e^{(\ln 2/5)t}, so its continuous rate constant is ln⁡(2)/5≈0.139\ln(2)/5\approx 0.139 per hour.

Inequalities follow the direction of the function

Exponential and logarithmic functions are one-to-one. Each is increasing when its base is greater than 1 and decreasing when its base is between 0 and 1. Applying an increasing function to both sides of an inequality keeps its direction; applying a decreasing function reverses it. Dividing by a negative number such as ln⁡(0.8)\ln(0.8) reverses it for the same reason.

Worked example 12.3: a decreasing base in a constructed context

A filter removes 20% of a contaminant on each pass, starting from 40 milligrams, so C(t)=40(0.8)tC(t)=40(0.8)^t after tt passes. For which passes is C(t)<10C(t)<10?

Divide by 40: (0.8)t<0.25(0.8)^t<0.25. Take natural logarithms: tln⁡(0.8)<ln⁡(0.25)t\ln(0.8)<\ln(0.25). Because ln⁡(0.8)<0\ln(0.8)<0, dividing reverses the inequality: t>ln⁡(0.25)/ln⁡(0.8)≈6.213t>\ln(0.25)/\ln(0.8)\approx 6.213. Passes are whole numbers, so the amount is below 10 milligrams from the 7th pass on. Check: C(6)≈10.49C(6)\approx 10.49 and C(7)≈8.39C(7)\approx 8.39.

A logarithmic inequality also carries its domain. For ln⁡(2x−1)≤3\ln(2x-1)\leq 3, the argument requires 2x−1>02x-1>0, and because ln⁡\ln is increasing, 2x−1≤e32x-1\leq e^3. Together, 12<x≤e3+12≈10.543\frac{1}{2}<x\leq\frac{e^3+1}{2}\approx 10.543. Writing only x≤10.543x\leq 10.543 includes inputs where the logarithm is undefined.

Try it 12.1 (partly done). Solve log⁡3(x+1)−log⁡3(x−1)=1\log_3(x+1)-\log_3(x-1)=1. The restriction is x>1x>1. The quotient property gives log⁡3(x+1x−1)=1\log_3\left(\frac{x+1}{x-1}\right)=1. Finish the solution and check it.

Check your work. The fraction equals 313^1, so x+1=3x−3x+1=3x-3 and x=2x=2. Since 2>12>1, test it: log⁡3(3)−log⁡3(1)=1−0=1\log_3(3)-\log_3(1)=1-0=1.

Undo the mapping to build an inverse

Functions such as f(x)=abx−h+kf(x)=ab^{x-h}+k and f(x)=alog⁡b(x−h)+kf(x)=a\log_b(x-h)+k are transformations of the basic exponential and logarithmic functions. To invert one, list the operations applied to the input, then undo them in reverse order.

Worked example 12.4: invert a transformed exponential

Let f(x)=3⋅2x+1−4f(x)=3\cdot 2^{x+1}-4. The input is increased by 1, used as an exponent of 2, multiplied by 3 and decreased by 4. Reverse: add 4, divide by 3, take log⁡2\log_2, subtract 1.

f−1(x)=log⁡2(x+43)−1f^{-1}(x)=\log_2\left(\frac{x+4}{3}\right)-1

Because 3⋅2x+1>03\cdot 2^{x+1}>0, the range of ff is y>−4y>-4, so the domain of f−1f^{-1} is x>−4x>-4. Check: f(1)=3⋅4−4=8f(1)=3\cdot 4-4=8 and f−1(8)=log⁡2(4)−1=1f^{-1}(8)=\log_2(4)-1=1. Undoing in the wrong order, as in log⁡2(x/3+4)−1\log_2(x/3+4)-1, fails the same test: it gives about 1.737, not 1.

A logarithm reverses the same way. Solving y=2ln⁡(x−3)+1y=2\ln(x-3)+1 for xx gives x=3+e(y−1)/2x=3+e^{(y-1)/2}, so g(x)=2ln⁡(x−3)+1g(x)=2\ln(x-3)+1 has inverse g−1(x)=3+e(x−1)/2g^{-1}(x)=3+e^{(x-1)/2} for every real xx. Its outputs, all greater than 3, match the domain of gg.

Build a logarithmic model from a context or data

A logarithmic model fits a situation in which the input changes proportionally while the output changes by equal amounts: each time the input is multiplied by the same factor, the output rises by the same amount. Exponential models reverse those roles, adding to the input and multiplying the output. You can build a logarithmic model from a proportion and a zero, from two input-output pairs, or from data with a logarithmic regression. The natural logarithm is common in data work.

Worked example 12.5: a proportion, a zero and two points

A constructed sensor scale rises by 2 units each time the concentration cc, in milligrams per liter, is multiplied by 10, and it reads 0 when c=0.5c=0.5. The factor of 10 suggests base 10, and the rise of 2 is the coefficient:

R(c)=2log⁡10(c0.5)=2log⁡10(2c),c>0R(c)=2\log_{10}\left(\frac{c}{0.5}\right)=2\log_{10}(2c),\quad c>0

Check: R(0.5)=0R(0.5)=0, R(5)=2R(5)=2 and R(50)=4R(50)=4. A model can also find an input. If R=5R=5, then log⁡10(2c)=2.5\log_{10}(2c)=2.5 and c=102.5/2≈158.1c=10^{2.5}/2\approx 158.1 milligrams per liter. Whether that concentration is meaningful depends on the range over which the sensor was tested; the formula cannot tell you.

Two input-output pairs also determine a model. If y=a+blog⁡2(x)y=a+b\log_2(x) passes through (4,9)(4,9) and (32,15)(32,15), then 9=a+2b9=a+2b and 15=a+5b15=a+5b. Subtracting gives b=2b=2, and then a=5a=5, so y=5+2log⁡2(x)y=5+2\log_2(x).

Worked example 12.6: fit, interpret and limit a regression

The table is constructed practice data: the number of bird species SS recorded after hh hours of observation at one site.

Hours, hh124816
Species, SS1219253339

Each doubling of hh adds 6 to 8 species, which suggests a logarithmic model. Enter the lists and request the regression S=a+bln⁡(h)S=a+b\ln(h); keep the unrounded values stored. The result is a≈12.000a\approx 12.000 and b≈9.810b\approx 9.810.

Interpret. Because ln⁡(2h)=ln⁡(h)+ln⁡(2)\ln(2h)=\ln(h)+\ln(2), each doubling adds bln⁡(2)≈6.800b\ln(2)\approx 6.800 predicted species. The value of aa is the prediction at h=1h=1, where ln⁡(1)=0\ln(1)=0; the model is undefined at h=0h=0. The residuals, observed minus predicted, are 0.0, 0.2, −0.6-0.6, 0.6 and −0.2-0.2, which are small and show no clear pattern.

Limit. The prediction S(10)≈34.6S(10)\approx 34.6 lies inside the observed interval. Solving S(h)=45S(h)=45 gives h=e(45−a)/b≈28.9h=e^{(45-a)/b}\approx 28.9 hours, outside the observed 1 to 16 hours. That answer assumes the doubling pattern continues, yet a site holds a limited number of species, and the model grows without bound.

Try it 12.2. A constructed model y=a+blog⁡3(x)y=a+b\log_3(x) passes through (3,7)(3,7) and (27,13)(27,13). Find aa and bb, then find xx when y=16y=16.

Check your work. 7=a+b7=a+b and 13=a+3b13=a+3b, so b=3b=3 and a=4a=4. Then 4+3log⁡3(x)=164+3\log_3(x)=16 gives log⁡3(x)=4\log_3(x)=4 and x=81x=81.

Read and build semi-log plots

A semi-log plot scales one axis logarithmically with a base n>1n>1. On a logarithmic vertical axis, equal distances represent equal ratios: the step from 10 to 100 has the same length as the step from 100 to 1,000. Data with exponential behavior therefore appear linear, as in Figure 12.2. The algebra shows why. If y=abxy=ab^x, then

log⁡n(y)=(log⁡nb)x+log⁡n(a)\log_n(y)=(\log_n b)x+\log_n(a)

The slope of the line is log⁡n(b)\log_n(b) and its vertical intercept is log⁡n(a)\log_n(a). Linear tools such as slope, intercept and linear regression now apply, provided you convert back with the same base.

Two side-by-side graphs of the same seven points, labeled constructed practice data, at x equals 0 through 6. In the left panel the vertical axis, y, has ordinary equal spacing from 0 to 60; the points rise slowly and then steeply, following a curve. In the right panel the vertical axis, y, is logarithmically scaled with base 10 and gridlines labeled 1, 10 and 100; the same points lie along a dashed straight line. Brackets on the right panel show two equal vertical distances, each labeled times 1.6 per unit of x.

Figure 12.2. The same constructed values of y=3(1.6)xy=3(1.6)^x on ordinary axes and on a base-10 logarithmic vertical axis.

Worked example 12.7: from a semi-log line back to an exponential

A semi-log display of constructed data plots ln⁡(y)\ln(y) against xx, and its line passes through (0,3)(0,3) and (8,1)(8,1). The slope is (1−3)/8=−0.25(1-3)/8=-0.25, so ln⁡(y)=3−0.25x\ln(y)=3-0.25x. Exponentiate both sides:

y=e3(e−0.25)x≈20.09(0.779)xy=e^3\left(e^{-0.25}\right)^x\approx 20.09(0.779)^x

Each unit of xx multiplies yy by about 0.779, a decrease of about 22.1%. Check at x=4x=4: the line gives ln⁡(y)=2\ln(y)=2, so y=e2≈7.389y=e^2\approx 7.389, and the model gives e3(e−0.25)4=e2e^3\left(e^{-0.25}\right)^4=e^2 as well. The slope −0.25-0.25 does not mean that yy falls by 0.25 per unit. It is the change in ln⁡(y)\ln(y).

Try it 12.3. A line on a semi-log plot of log⁡10(y)\log_{10}(y) passes through (0,0.5)(0,0.5) and (3,1.4)(3,1.4). Write y=abxy=ab^x.

Check your work. The slope is 0.3, so a=100.5≈3.162a=10^{0.5}\approx 3.162 and b=100.3≈1.995b=10^{0.3}\approx 1.995. Check: 3.162(1.995)3≈25.1≈101.43.162(1.995)^3\approx 25.1\approx 10^{1.4}.

Worked example 12.8: trust the trend for large inputs

The framework adds a useful point: if the logarithms of the outputs trend linear for large inputs, a transformed exponential such as y=abx+ky=ab^x+k can model the data, even before you remove the constant kk. Constructed values of y=5⋅3x+200y=5\cdot 3^x+200 show the idea.

xx02468
yy2052456053,84533,005
log⁡10(y)\log_{10}(y)2.3122.3892.7823.5854.519

The changes in log⁡10(y)\log_{10}(y) over each 2 units are 0.077, 0.393, 0.803 and 0.934. They approach 2log⁡10(3)≈0.9542\log_{10}(3)\approx 0.954, so the plot bends at first and then straightens. A student who demands a constant change across the whole table would reject a model that suits the data.

Error analysis: a check in the wrong place

A student solves log⁡4(x−3)+log⁡4(x+3)=2\log_4(x-3)+\log_4(x+3)=2, writes x2−9=16x^2-9=16, finds x=5x=5 or x=−5x=-5, and confirms both because 25−9=1625-9=16. The check used the transformed equation, which was built to accept both candidates. The original requires x>3x>3, and log⁡4(−8)\log_4(-8) is undefined, so the solution is x=5x=5 only. The repair is a habit: check in the first line you were given.

Retrieval and review

At your next study session, without notes, write the three-step restriction record and use it to solve the error-analysis equation again on a blank page. About a week later, explain in two sentences why equal vertical gaps on a logarithmic axis mean equal ratios, then convert one semi-log line into an exponential model. In Chapter 20's mixed practice, name the restriction before choosing a method. If a logarithm property caused your error, return to Chapter 11 first.

Practice

Ten multiple-choice questions, Questions 181–190, and one free-response task, Written task 13, are printed after this chapter. Questions 181–186 are no-calculator questions. Questions 187–190 and Written task 13 require a graphing calculator in radian mode. Keys, explanations for every choice and a scoring guide with a complete response are under Chapter 12: answers and explanations. For each miss, record the first incorrect step and the restriction or interpretation you overlooked.

Use this chapter companion

Review the lesson, attempt the original practice, then check the explanation. Your printed study guide remains complete without the website.

Independence

Skill by Skill Prep is independent. SAT® and AP® are trademarks registered by the College Board, which is not affiliated with, and does not endorse, this website or book. PSAT/NMSQT® is a registered trademark of the College Board and the National Merit Scholarship Corporation, which are not affiliated with, and do not endorse, this site. Bluebook™ is a trademark of the College Board. ACT® is a trademark of ACT Education Corp., which is not affiliated with and does not endorse this website or book. Desmos is a trademark of Desmos Studio PBC. Practice results describe these questions, not official scores or admission predictions.

Last updated .