6.8 Exponential Growth and Decay

The differential equation whose rate is proportional to the amount, its solution by separating variables, doubling times and half-lives, Newton's law of cooling applied to the temperature difference, and the limits of the model.

Subject: Calculus I · 65 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. Exponential Growth and Decay

Title

Calculus I · Section 6.8

One differential equation, and everything that obeys it

2. What this lesson gives you

Objectives

Section 6.7 built the natural exponential and the natural logarithm as a matched pair. This section spends them. One differential equation says a quantity changes at a rate proportional to how much of it there is, and its solution describes bacteria, bank balances, radioactive atoms and cooling coffee.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 636-645 — learning objectives 6.8.1 to 6.8.4

Everything here comes out of a single sentence: the rate at which something changes is proportional to how much of it there is. That sentence is a differential equation, and this is the first one you will solve. Once it is solved, the rest of the section is the same solution wearing five different costumes.

Watch for the thing that keeps happening. In almost every calculation the starting amount cancels. That is why a half-life can be printed in a table of the elements without anyone asking how large your sample is, and it is worth noticing each time it happens rather than only at the end.

The last part is the most important and the least examinable. The model is a local description. It is superb over the range of data you fitted it to, and it becomes nonsense a long way past that, in a way you will be able to state precisely by the end.

3. Before anything new: guess the function

Warm-up

Discussion prompt

Name a function whose rate of change at every instant is a fixed multiple of its own current value, then check your candidate by differentiating it. What does the fixed multiple turn out to be?

Write something down before reading on, even if you are not sure. If you wrote a power of t, differentiate it and look at what you get: the power drops, so the derivative is not a multiple of the original function. Powers cannot do this.

The exponential can, and it is essentially the only thing that can. Differentiating the exponential of a constant times t brings that constant down in front and leaves the function untouched, so the derivative is the constant times the function. That is the property the whole section rests on, and it is the reason base e rather than base 10 or base 2 is the natural choice here.

Keep the line you just wrote visible in your mind. When the solution appears in a few slides after four lines of algebra, you will already know what it is going to be, and the algebra will be a confirmation rather than a surprise.

4. The equation of proportional change

Section

Part 1

5. Rate proportional to amount

Concept

Many systems change at a rate set by how much of them there is. More bacteria means more divisions per minute, more radioactive atoms means more decays per second, more money in the account means more interest per year.

\[ \frac{dy}{dt} = ky \]

Figure (svg): A slope field over two hours and heights up to 2400 in which the little segments are almost flat near the bottom of the picture and steeply tilted near the top, with three solution curves starting at heights 60, 130 and 200 threading through them.

The segments say only one thing: the higher you are, the faster you rise. Every curve that follows them is an exponential, and the starting height picks which one.

The slope field draws the assumption directly. Every short segment is as steep as the height it sits at, times the constant, so the field is nearly flat along the bottom and steeply tilted along the top. Choose a starting height and follow the segments: you get one of the three curves, and all three have the same shape.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 636-637 — Rule: Exponential Growth Model

Spend a moment on the picture rather than the equation. A slope field draws the derivative at many points at once: each little segment shows how steeply a solution passing through that point would be rising. Here the segments near the bottom are almost flat and the ones near the top are steep, because the rule says the steepness is proportional to the height.

Now trace one out with your eye. Start at the left at a height of 130 and keep following whichever segment you are sitting on. You will sweep out the middle curve. Start lower and you get the lower curve. The field does not care where you start, only how high you are at each moment, and that is exactly what makes all three curves the same shape.

This is also the honest way to see what the equation does and does not say. It says nothing about the size of the population, only about how the size and the rate are related. Fixing the starting height is a separate piece of information, and it is what picks one curve out of the infinite family.

6. Separating the variables

Concept

One technique solves this equation, and it is worth doing in full once. Collect each variable on its own side, each with its own differential, and then integrate both sides.

separation of variables — Rearranging a differential equation so that every appearance of one variable stands on one side with its differential, then integrating both sides separately.

\[ \frac{dy}{dt} = ky \;\Longrightarrow\; \frac{dy}{y} = k\,dt \]

The left-hand integral is the one case the power rule could never do, and Section 6.7 supplied its antiderivative: the natural logarithm. That is why a logarithm turns up in every single calculation in this lesson, including the ones about coffee.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 636 — the differential equation of the growth model

This is the only technique in Calculus I for solving a differential equation, and it is worth understanding rather than memorising. The obstacle is that the unknown function appears on both sides, once as a derivative and once by itself. You cannot integrate with respect to t while an unknown function of t is sitting there.

The repair is to treat the derivative as a genuine ratio of differentials and split it. Dividing by y puts every appearance of y on the left, next to dy, and leaves only constants and dt on the right. Now each side can be integrated in its own variable, and neither integral contains the unknown function in an awkward place.

The left-hand integral is the one exception the power rule always had to exclude, and Section 6.7 filled it in for you. Every logarithm you will see in this lesson, including the ones in the cooling problems and the carbon-dating problems, traces back to this one line.

7. Solving the model from nothing

Worked example

Solve the equation once, with a general starting amount. Every application in the section is this solution with numbers in it.

\[ \frac{dy}{dt} = ky, \qquad y(0) = y_0 \]

Separate the variables

Why: Divide by the amount and multiply by the time differential.

\[ \frac{dy}{y} = k\,dt \]

Integrate both sides

Why: The reciprocal integrates to a logarithm; a constant integrates to itself times the variable.

\[ \int \frac{dy}{y} = \int k\,dt \;\Longrightarrow\; \ln|y| = kt + C \]

Exponentiate

Why: Apply the exponential to both sides to undo the logarithm.

\[ |y| = e^{kt+C} = e^{C}e^{kt} \]

Absorb the constant

Why: The exponential of the constant is a positive factor; allowing either sign lets a single letter carry it.

\[ y = Ae^{kt} \]

Impose the initial condition

Why: Substitute time zero, where the exponential equals one.

\[ y(0) = Ae^{0} = A = y_0 \]

\[ y = y_0e^{kt} \]

Check both requirements

Why: Differentiating returns the constant times the same expression, so the equation holds, and substituting time zero returns the starting amount, so the condition holds. A solution has to pass both tests.

\[ y' = ky_0e^{kt} = ky, \qquad y(0) = y_0e^{0} = y_0 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 636-637 — Rule: Exponential Growth Model

Do not let the shortness of this fool you: these five lines generate every answer in the section. Work through them once with a pencil now, and then you will be able to write the solution down instantly whenever you meet it, which is what the examples assume.

Two lines deserve care. The exponentiation step turns an additive constant into a multiplicative one, because the exponential of a sum is a product; that is where most mistakes on this page come from. And the absorbing step quietly drops the absolute value bars, which is legitimate because the new constant is allowed to be negative and because in every application here the quantity is positive anyway.

The check at the end is not decoration. A solution to an initial-value problem has two duties, satisfying the equation and hitting the given starting value, and they fail in different ways. A wrong exponent breaks the first; a wrong coefficient breaks the second. Testing both takes ten seconds.

8. Reading the solution piece by piece

Notation

Annotate

On: \( y = y_0 e^{kt} \)

  • The amount at time zero. It multiplies the whole curve, so doubling the starting amount doubles every value without altering the shape by one bit.
  • The rate constant, measured in units of one over time. Positive gives growth, negative gives decay, and its size sets how fast.
  • Time measured from the instant at which y₀ was recorded. Move that instant and y₀ changes with it; k never does.
  • Base e, not out of taste but out of necessity: it is the base whose exponential is its own derivative, which is what made the differential equation come out clean.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 637 — the growth constant

Click through the annotations one at a time and connect each to the picture two slides back. The starting amount is a vertical stretch: it scales the whole curve without bending it, which is why all the solution curves in the slope field looked identical in shape.

The constant in the exponent is the one that changes the shape, and it carries units. If time is in minutes the constant is per minute, and the quickest way to catch an error in one of these problems is to notice that the units of the constant do not match the units the time is measured in.

The last annotation is the one to take seriously if you have wondered why base e appears everywhere. You could write the same solution with base 2 or base 10 and a different constant, and sometimes people do. Base e is preferred because its exponential is its own derivative, which is precisely what makes the differential equation come out with no leftover factors.

9. Equation 6.27: the check runs backwards too

Concept

The book states the formula first and then finds the differential equation hiding inside it. Run that direction once; it is the shortest possible proof that the exponential is the right answer and not merely an answer.

\[ y = y_0e^{kt} \]

\[ y' = y_0 \cdot k e^{kt} = k\left(y_0 e^{kt}\right) = ky \]

The chain rule brings the constant down in front, and what remains is the original function, unchanged. So the derivative of the solution is the constant times the solution, which is exactly the equation. The book labels this line Equation 6.27 and calls the proportionality the key feature of exponential growth.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 637 — Equation 6.27

The book does something slightly unusual here and it is worth noticing. It states the exponential formula first, as a model, and only then differentiates it to discover the differential equation inside. You have just done it in the other direction, from the equation to the formula, so now you have both halves.

Having both halves matters because it means the correspondence is exact. Every solution of the differential equation is an exponential with some starting value, and every such exponential solves the equation. There are no other solutions hiding anywhere, so when you find one you have found them all.

The one line to be able to reproduce under pressure is the middle one. Differentiating brings the constant down by the chain rule, and the expression that is left is the whole original function, not merely part of it. Say that out loud once and the whole section becomes easier.

10. The step that makes it integrable

Fill the middle

The solution has five moves. Only the first is a genuine idea; supply it.

Fill in the blanks

\fracy___ = ky \;\Longrightarrow\; \frac______} } = k\,dt

Why: Divide both sides by y. That is what puts every appearance of y on the left with dy and every appearance of t on the right with dt, so the two sides can be integrated in their own variables. Without that division the left side still contains y and cannot be integrated with respect to t.

Fill the blank before checking. If you hesitated, ask yourself what has to be true before you are allowed to integrate: each side must contain only its own variable.

The right-hand side is already fine, because the rate constant and dt are all that is there. The left-hand side is not, because dy is divided by dt and multiplied by nothing. Dividing the whole equation by y moves y across to join dy, and the two sides then live in separate variables.

It is worth noticing what you are not allowed to do. You cannot divide by y if y might be zero, and strictly there is a constant solution, the quantity that is zero forever, which this method loses. In every application in this section the quantity is positive, so nothing is harmed, but the exception is real.

11. Trap: the constant added instead of multiplied

Trap

The trap

From the integrated line, a very common next move:

\[ \ln|y| = kt + C \]

\[ y = e^{kt} + C \]

Wrong. The constant was inside the exponent.

The fix

Exponentiating a sum gives a product, never a sum. The constant leaves the exponent as a factor, and the initial condition then names that factor as the starting amount.

\[ e^{kt+C} = e^{C}e^{kt} \]

\[ y = y_0e^{kt} \]

This is the most common single error in solving the equation, and it survives into later courses. The tempting move is to exponentiate the left side properly and then simply carry the constant along unchanged, as though it were outside the exponent.

The defence is a rule you already know from Section 6.7: the exponential of a sum is the product of the exponentials. The constant is inside the exponent when you exponentiate, so it comes out as a factor, and a factor is exactly what you want, because the initial condition then identifies it as the starting amount. If the constant stayed additive there would be nowhere sensible for the starting amount to live.

There is also a quick test. Differentiate the wrong answer. You get the constant times an exponential, which is not the constant times the whole wrong answer, because of the stray added term. It fails the equation, and that is how you catch it on your own.

12. What the rate constant actually says

Intuition

The constant is not a speed. It is a proportion. Divide the rate of change by the amount present and the amount cancels out, leaving the same number whatever the size of the system.

\[ \frac{1}{y}\frac{dy}{dt} = k \]

So a constant of 0.02 per minute is naturally read as two percent of the colony added every minute, whether the colony is two hundred cells or two million. That is the whole content of the proportionality assumption, and it is also the assumption that eventually fails.

what is changingkread it as
the bacteria of Table 6.10.02 per minuteabout 2% of the colony added each minute
carbon-14 in a bone−0.000121 per yearabout 0.0121% of the atoms lost each year
money at 5%, continuous0.05 per yearabout 5% of the balance added each year

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 637 — the growth constant

Getting this right changes how the rest of the section reads. The constant is not how fast the quantity is growing; it is how fast the quantity is growing relative to its own size. Divide the rate by the amount and the amount cancels, which is the whole assumption in one line.

That is why it is fair to translate a constant of 0.02 per minute into about two percent per minute. When the colony is small, two percent of it is a small absolute increase; when the colony is huge, two percent of it is huge. The proportion stays fixed while the absolute rate grows, and the curve steepens for exactly that reason.

Read the table across. The middle column is the same kind of number in all three rows, even though one is about cells, one about atoms and one about money. If you can state a system's rate constant, you have said everything the model knows about it.

13. Slide the rate constant

Tweak it

Parameter explorer

The curve is the exponential with your chosen rate constant, starting from a height of 1 at the left edge. Slide the constant from its largest positive value down through zero to its most negative. Where is the curve steepest, what happens exactly at zero, and how do the positive and negative settings mirror one another?

\[ y = e^{{k}t} \]

  • k — from -0.05 to 0.05: rate constant k

Start at the largest positive value and slide down slowly. The curve is steepest at the right-hand end, always, because it is tallest there and the rate is proportional to the height. Halving the constant does not merely lower the curve; it stretches the whole shape out along the time axis, so the same heights arrive twice as late.

That stretching is the doubling-time formula seen before it is written down. At a constant of 0.05 the curve reaches twice its starting height after about 13.9 units of time; at 0.025 it takes about 27.7, and at 0.0125 about 55.5. The time is the logarithm of 2 divided by the constant, so the two are inversely proportional.

Now pass through zero. At exactly zero the curve is a flat line at 1: nothing is being added, so nothing changes. Push further and the curve flips into a decay, and the time to halve is the same as the time to double was for a constant of the same size. That symmetry is why one formula serves both, a point the book makes explicitly two definitions from now.

14. Two hundred bacteria, and Table 6.1

Picture it

Figure (svg): The bacteria curve rising from 200 at time zero to 2205 after two hours, with the twelve counts of Table 6.1 drawn as dots that lie exactly on it, and a dashed line at 400 marking the first doubling.

Twelve measured counts, ten minutes apart, all on one curve. Between t = 30 and t = 40 the colony passes 400, which is the doubling the dashed line marks.

The book's own example. A colony of two hundred with a constant of 0.02 per minute, measured every ten minutes for two hours. The counts in the table are not data: they are the model evaluated and rounded to whole cells, which is the book being careful, because a real colony contains a whole number of bacteria and the model does not.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 637-638 — Figure 6.79 and Table 6.1

This is the book's own worked picture, and the dots are the source of the numbers in Table 6.1. Notice that they sit exactly on the curve, because they were computed from it rather than measured in a laboratory. That is a modelling picture, not a data picture.

The sentence the book adds beside the table is more important than the table. The model takes non-integer values, and real colonies contain whole cells. So the curve is an approximation to something inherently lumpy, and reading an answer of 80,685.76 bacteria to two decimal places is meaningless. Round, and say what you rounded.

Look at the shape. Over the first half hour the curve is nearly straight and you could believe it was linear; by two hours it is visibly bending upward. Exponential growth always looks harmless at the beginning, and that is a theme this lesson comes back to in its last part.

15. Example 6.42: five hours of bacteria

Worked example

The colony above, followed well past the table. How many after five hours, and when does it reach one hundred thousand?

\[ f(t) = 200e^{0.02t}, \quad t \text{ in minutes} \]

Convert the time to minutes

Why: The model is written in minutes, so five hours must be converted before anything is substituted.

\[ t = 5 \cdot 60 = 300 \]

Substitute

Why: Put 300 in place of t; the exponent collapses to a whole number.

\[ f(300) = 200e^{0.02(300)} = 200e^{6} \]

Evaluate

Why: The exponential of 6 is about 403.43.

\[ 200(403.4288) \approx 80{,}686 \]

Set the model equal to the target

Why: Now the reverse question: the count is known and the time is not.

\[ 100{,}000 = 200e^{0.02t} \]

Divide by the starting count

Why: This is the move that leaves the exponential alone on one side.

\[ 500 = e^{0.02t} \]

Take logarithms

Why: The natural logarithm brings the exponent down to ground level.

\[ \ln 500 = 0.02t \]

Divide by the rate constant

Why: The logarithm of 500 is about 6.2146.

\[ t = \frac{\ln 500}{0.02} \approx 310.73 \text{ minutes} \]

Check against the doubling time

Why: The colony doubles every 34.66 minutes, so 310.73 minutes is 8.97 doublings; two raised to that power, times the initial 200, is one hundred thousand. The two routes agree.

\[ \frac{310.73}{34.657} \approx 8.966, \qquad 200 \cdot 2^{8.966} \approx 100{,}000 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 638 — Example 6.42

The example asks two questions that are inverse to each other, and it is worth separating them in your head. The first gives a time and wants an amount, so it is a substitution and takes one line. The second gives an amount and wants a time, so the unknown is trapped in an exponent and a logarithm is the only way to get it out.

The unit conversion at the start earns its own step because it is where marks are lost. The model was written with the time in minutes, so five hours must become 300 before anything else happens. If you substitute 5 you get 221 bacteria, a plausible-looking wrong answer that nothing later in the calculation will flag.

The check at the end is the pattern to copy through the whole lesson. Convert the answer into doublings and see whether it makes sense: 310 minutes is nine doublings, and doubling 200 nine times gives about a hundred thousand. Two independent routes to the same number is a much stronger check than re-reading your algebra.

16. Checkpoint 6.42: a faster culture

Worked example

The same two questions for a culture that starts larger and grows two and a half times faster.

\[ f(t) = 500e^{0.05t}, \quad t \text{ in minutes} \]

Convert four hours to minutes

Why: Again the model is in minutes.

\[ t = 4 \cdot 60 = 240 \]

Substitute and evaluate

Why: The exponential of 12 is about 162,755.

\[ f(240) = 500e^{12} \approx 8.14 \times 10^{7} \]

Set the model equal to one hundred million

Why: The second question, in the same shape as before.

\[ 100{,}000{,}000 = 500e^{0.05t} \]

Divide by the starting count

Why: Leaving the exponential by itself.

\[ 200{,}000 = e^{0.05t} \]

Take logarithms

Why: The logarithm of two hundred thousand is about 12.206.

\[ \ln 200{,}000 = 0.05t \approx 12.206 \]

Divide by the rate constant

Why: And the time falls out.

\[ t = \frac{12.206}{0.05} \approx 244.1 \text{ minutes} \]

Check by counting doublings

Why: This culture doubles every 13.86 minutes, so 244.1 minutes is 17.61 doublings, which multiplies 500 by about two hundred thousand, giving one hundred million.

\[ \frac{\ln 2}{0.05} \approx 13.863, \quad \frac{244.1}{13.863} \approx 17.61, \quad 500\cdot 2^{17.61} \approx 10^{8} \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 638 — Checkpoint 6.42

Same two questions, different numbers, and you should be able to run it without looking at the previous slide. Convert the hours, substitute, then reverse the process for the second part.

The one thing that catches people is the size of the answer. Eighty-one million bacteria from a starting five hundred sounds wrong until you count doublings: this culture doubles every fourteen minutes or so, and four hours is about seventeen doublings, which multiplies by well over a hundred thousand. Exponentials get large faster than intuition expects, every time.

Compare the two cultures. The second has a rate constant two and a half times larger and reaches its far bigger target in a shorter time. Rate constants are where all the difference lives, which is why the next part of the lesson is about converting them into something you can feel, namely a doubling time.

17. Money, compounded continuously

Section

Part 2

18. Compound the interest more and more often

Concept

Simple interest is paid once, at the end. Compounding pays part of the year's interest early, so the remainder of the year earns interest on that interest as well. The more often the bank does it, the more you finish with.

\[ A = P\left(1 + \frac{r}{n}\right)^{nt} \]

Figure (svg): Five dots on stems showing the year-end balance on one thousand at two percent when the interest is compounded 1, 2, 3, 12 and 365 times a year; the dots climb steeply at first and then crowd against a dashed line at the continuous value 1020.2013.

The first extra compounding is worth ten cents; going from daily to continuous is worth less than a tenth of a cent. The dashed ceiling is what the limit buys.

The gains shrink fast. Going from once a year to twice adds ten cents on a thousand; going from daily to continuous adds less than a tenth of a cent. The dots are climbing toward a ceiling, and the next slide identifies it exactly.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 638-639 — simple interest and compound interest

The idea here is worth more than the formula. Each compounding pays you part of the year's interest early, and that early payment then earns interest itself for the rest of the year. More compoundings mean more of your interest is working for longer.

Look at how quickly the picture flattens. The step from annual to semi-annual is worth ten cents on a thousand; every step after that is worth less, and the step from daily to continuous is worth under a tenth of a cent. The five positions along the bottom are not to scale, which flatters the right-hand side: the gap from twelve compoundings to three hundred and sixty-five is far bigger than the gap from one to two, and it buys far less.

The fact that there is a ceiling at all is not obvious. You are raising a number slightly larger than one to a larger and larger power, and those two effects fight each other. The next slide shows that the fight ends in a draw at a specific number, and that number is e.

19. From n compoundings a year to the exponential

Worked example

Let the number of compoundings grow without bound and watch the formula turn into an exponential.

\[ A = P\left(1 + \frac{r}{n}\right)^{nt} \]

Substitute a new counter

Why: Put n equal to r times m, so that the fraction inside the bracket becomes one over m.

\[ n = rm \;\Longrightarrow\; A = P\left(1 + \frac{1}{m}\right)^{rmt} \]

Regroup the exponent

Why: Write the power as an m-th power, itself raised to rt.

\[ A = P\left[\left(1 + \frac{1}{m}\right)^{m}\right]^{rt} \]

Take the limit

Why: As n grows without bound so does m, and the bracket is the classical limit that defines the number e.

\[ \lim_{m\to\infty}\left(1+\frac1m\right)^{m} = e \]

Read off the balance

Why: Only the bracket depended on m, so the limit passes straight through.

\[ \text{Balance} = Pe^{rt} \]

Check against the book's arithmetic

Why: One thousand at two percent for one year gives 1020.00 compounded annually, 1020.10 twice a year, 1020.13 three times, 1020.18 monthly and 1020.2005 daily, closing on the continuous value from below exactly as the picture shows.

\[ 1000\left(1+\tfrac{0.02}{365}\right)^{365} = 1020.2005 \;<\; 1000e^{0.02} = 1020.2013 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 639 — the derivation of continuous compounding

This derivation is where the number e earns its place in a finance course. The substitution in the second line looks like a trick, and it is, but it is the natural trick: you want the inside of the bracket to be one plus one over something, because that is the shape of the limit that defines e.

The regrouping in the third line is the real move, and it is only the exponent law that says a product in an exponent can be read as one power raised to another. Once the bracket is isolated as an m-th power, the rest of the exponent just rides along outside and is untouched by the limit.

The check with the book's numbers is worth doing yourself. Daily compounding gives 1020.2005 and the continuous formula gives 1020.2013: they agree to five figures. That is why banks quote continuous rates even though no bank credits interest continuously. The formula is simpler and the error is smaller than a cent.

20. Example 6.43: what a million costs at twenty-five

Worked example

A student of 25 wants one million at 65, from an account paying five percent compounded continuously. How much has to go in today?

\[ 1{,}000{,}000 = Pe^{0.05(40)} \]

Count the years

Why: From 25 to 65 is forty years, and the exponent is the rate times the years.

\[ rt = 0.05(40) = 2 \]

Divide by the growth factor

Why: The deposit is the target divided by the exponential of 2.

\[ P = \frac{1{,}000{,}000}{e^{2}} = 1{,}000{,}000\,e^{-2} \]

Evaluate

Why: The exponential of minus 2 is about 0.135335.

\[ P \approx 135{,}335.28 \]

Repeat at six percent

Why: Only the exponent changes, from 2 to 2.4.

\[ P = 1{,}000{,}000\,e^{-2.4} \approx 90{,}717.95 \]

Check how much the extra point bought

Why: The ratio of the two deposits is the exponential of minus 0.4, about 0.670, so one extra percentage point over forty years cuts the required deposit by a third, which matches the book's remark about roughly two-thirds.

\[ \frac{e^{-2.4}}{e^{-2}} = e^{-0.4} \approx 0.6703 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 639 — Example 6.43

The equation is the same one you have been solving, run in reverse: instead of a starting amount growing to an unknown total, you have a known total and want the starting amount. Algebraically it is a single division, and the only care needed is with the exponent.

Count the years carefully. The exponent is the rate times the time, so at five percent over forty years it is exactly 2, which is why the arithmetic comes out so cleanly. Getting 65 minus 25 wrong is the only way to fail this problem.

The comparison at the end is the part to remember. One extra percentage point of interest over forty years cuts the required deposit by a third. The reason is that the deposit depends on the exponential of the rate times the time, so a change in the rate is magnified by the length of the investment. Over one year the same extra point would be worth about one percent.

21. Checkpoint 6.43: ten years of waiting

Worked example

Same target, same two rates, but she starts at 35 instead of 25.

\[ 1{,}000{,}000 = Pe^{r(30)} \]

Count the years

Why: From 35 to 65 is thirty, not forty.

\[ t = 65 - 35 = 30 \]

Solve at five percent

Why: The exponent is 1.5 now.

\[ P = 10^{6}e^{-1.5} \approx 223{,}130.16 \]

Solve at six percent

Why: The exponent is 1.8.

\[ P = 10^{6}e^{-1.8} \approx 165{,}298.89 \]

Check the price of the delay

Why: At five percent the required deposit rose from 135,335.28 to 223,130.16, and the ratio is the exponential of 0.5, exactly the ten years of growth she gave away.

\[ \frac{223{,}130.16}{135{,}335.28} = e^{0.5} \approx 1.6487 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 639 — Checkpoint 6.43

Run this yourself before reading the steps. Nothing changes except the number of years, from forty to thirty, so you are substituting one number into a calculation you have already done twice.

The check is the interesting part. At five percent the required deposit goes from about 135,000 to about 223,000: waiting ten years costs you an extra eighty-eight thousand dollars up front. And the ratio between them is the exponential of 0.5, which is precisely the growth factor of the ten years she gave away.

That last observation generalises, and it is the single most useful thing about continuous compounding. Delaying by a fixed period always multiplies the required deposit by the same factor, whatever the target and whenever you start. The cost of waiting is proportional, not additive.

22. Exercise 349: which offer pays more?

Prediction

\[ 500e^{0.025} \quad \text{against} \quad 500(1 + 0.03) \]

Predict first

Five hundred is offered at three percent credited once a year, or at two and a half percent compounded continuously. Which is worth more after one year?

  • Three percent, once a year
  • Two and a half percent, continuously
  • They come out equal
  • It depends on the starting amount

Correct: Three percent, once a year

Why: Three percent once a year multiplies by 1.03, giving 515. Two and a half percent compounded continuously multiplies by the exponential of 0.025, about 1.02532, giving 512.66. Continuous compounding is worth only about a twentieth of a percentage point of extra rate here, far less than the half point given away, so the larger stated rate still wins. The answer also cannot depend on the starting amount, since both offers are multiplications by a fixed factor.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 643 — Exercise 349

Commit to an answer before revealing, because the instinct here is strong and it is wrong. Continuous compounding sounds like it should be a big advantage, so a slightly lower rate compounded continuously feels like it might win.

Work out what continuous compounding is actually worth. At these rates it adds roughly half of the square of the rate, which for two and a half percent is about three hundredths of a percentage point. You gave away half a percentage point to get it, so the trade is a bad one by a wide margin.

The useful rule to carry away is that at ordinary interest rates the compounding convention matters far less than the rate itself. Continuous compounding at a given rate is worth about the same as annual compounding at that rate plus a small fraction of a point. Exercise 363 asks the same question the other way round: the continuous rate that matches nine percent annually is about 8.6 percent.

23. Doubling time and half-life

Section

Part 3

24. The time to double does not depend on the amount

Concept

Ask when a growing quantity reaches twice its starting value. The starting value appears on both sides and cancels, and what survives mentions only the rate constant.

\[ 2y_0 = y_0e^{kt} \]

\[ 2 = e^{kt} \]

\[ \ln 2 = kt \]

\[ t = \frac{\ln 2}{k} \]

doubling time — The time a growing exponential quantity takes to double. It depends on the rate constant alone, not on how much is present when you start counting.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 640 — Definition: doubling time

Follow the cancellation in the first two lines, because it is the whole idea. You ask for twice the starting amount, the starting amount appears on both sides, and it divides out. What is left is an equation about the factor 2 and nothing else.

So the answer cannot possibly mention how much you started with. A colony of two hundred and a colony of two million, growing with the same constant, take exactly the same time to double. The book makes the point with bacteria going from a hundred to two hundred, and from ten thousand to twenty thousand, in the same number of minutes.

This is also the model's sharpest testable prediction and its easiest falsification. Measure the doubling time at two very different population sizes. If the two measurements agree, the exponential model is doing well; if the doubling time lengthens as the population grows, you are watching the model fail, which is the subject of the last part of this lesson.

25. Reading the doubling-time formula

Notation

Annotate

On: \( T_2 = \frac{\ln 2}{k} \)

  • The factor you asked about. Replace it by 3 and the same three lines give the tripling time, ln 3 over k; replace it by 10 and you get the tenfold time.
  • The only quantity that varies. Doubling time and rate constant are reciprocal: double k and the doubling time halves.
  • The starting amount is gone. It cancelled when 2y₀ = y₀e^(kt) was divided through, and that cancellation is the only reason half-lives can be printed in tables.
  • k is per unit time, so ln 2 divided by k is a time. If k is per minute, the doubling time comes out in minutes.

\[ T_3 = \frac{\ln 3}{k} \approx 1.585\,T_2 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 640 — Definition: doubling time

Work through the annotations and pay particular attention to the first one. Nothing about the derivation was special to the factor 2. Ask when a quantity triples and the identical three lines give the logarithm of 3 over the constant; ask about a tenfold increase and you get the logarithm of 10 over the constant.

The line under the annotations makes that concrete. Tripling takes about 1.585 doubling times, always, for every exponentially growing thing in the universe, because the logarithm of 3 divided by the logarithm of 2 is 1.585. That number does not depend on the rate constant at all.

The third annotation is the one to notice. The absence of the starting amount is not an accident of this particular formula; it is a consequence of the cancellation you just watched. Whenever a question is about a ratio rather than an amount, expect the starting value to disappear.

26. Equal intervals, equal factors

Picture it

Figure (svg): The same growth curve with four dashed vertical lines 34.7 minutes apart and dashed horizontal lines at 200, 400, 800, 1600 and 3200; the curve crosses each horizontal line exactly at a vertical line, so each equal time step multiplies the count by two.

Equal steps along the bottom, equal multiplications up the side. A constant ratio, not a constant difference, is what makes a curve exponential.

The dashed vertical lines are equally spaced, 34.66 minutes apart. The dashed horizontal lines are not equally spaced: they are 400, 800, 1600, 3200, each twice the one below. The curve meets them in pairs, and that is the signature of an exponential. A straight line would add the same amount in each interval; this adds the same factor.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 639-640 — the constancy of the doubling time

Look at the two families of dashed lines separately. The vertical ones are equally spaced, about 34.7 minutes apart. The horizontal ones are not equally spaced at all: they are 400, 800, 1600 and 3200, each twice the one below, so they crowd together at the bottom and spread out at the top.

Now notice that the curve passes through the corners. Equal steps along the bottom correspond to equal multiplications up the side. That is the geometric definition of an exponential and the cleanest way to recognise one from a graph or a table: look at the ratios of successive values, not the differences.

If you are ever unsure whether some data is exponential, this is the test to run. Take the values at equally spaced times and divide each by the one before. A constant difference means a straight line; a constant ratio means an exponential.

27. Example 6.44: when may the friends fish?

Worked example

A pond is stocked with five hundred fish. After six months there are a thousand. The owner opens the pond when the population reaches ten thousand.

\[ y = 500e^{kt}, \qquad y(6) = 1000 \]

Read off the doubling time

Why: Five hundred became a thousand in six months, which is one doubling.

\[ 6 = \frac{\ln 2}{k} \]

Solve for the rate constant

Why: Rearranged, the constant is the logarithm of 2 over six.

\[ k = \frac{\ln 2}{6} \approx 0.11552 \text{ per month} \]

Write the model

Why: Starting amount in front, fitted constant in the exponent.

\[ y = 500e^{(\ln 2/6)t} \]

Set it equal to ten thousand

Why: The fishing question, written as an equation.

\[ 10{,}000 = 500e^{(\ln 2/6)t} \]

Divide by the stocked number

Why: Twenty times the original population.

\[ 20 = e^{(\ln 2/6)t} \]

Take logarithms

Why: Bringing the exponent down as usual.

\[ \ln 20 = \frac{\ln 2}{6}\,t \]

Solve for the time

Why: Multiply by six and divide by the logarithm of 2.

\[ t = \frac{6\ln 20}{\ln 2} \approx 25.93 \text{ months} \]

Check by counting doublings

Why: Twenty times is between four doublings, which is sixteen, and five, which is thirty-two; 25.93 months is 4.32 doubling times, and two to that power is twenty. A little over two years, as the book says.

\[ \frac{25.93}{6} \approx 4.322, \qquad 2^{4.322} \approx 20 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 640 — Example 6.44

This example is set up slightly differently from Example 6.42, and the difference is the point of it. You are not handed the rate constant; you are handed a doubling time, and you have to convert. Read the first two steps as one idea: the population went from 500 to 1000, which is one doubling, so six months is the doubling time.

From there it is the same calculation as before. Write the model, set it equal to the target, divide out the stocked number, take a logarithm. The one thing worth pausing on is that the 500 cancels, so the answer would have been the same for a pond stocked with any number, as long as the doubling time were still six months.

The check is the one to practise. Twenty times the original is between four and five doublings, and 25.93 months is 4.32 doubling times. If your answer had come out as 12 months you would see immediately that two doublings only quadruples the population, nowhere near twenty times.

28. Checkpoint 6.44: a slower pond

Worked example

The same pond and the same target, but the thousandth fish arrives only after nine months.

\[ y = 500e^{kt}, \qquad y(9) = 1000 \]

Read the new doubling time

Why: Nine months per doubling instead of six.

\[ k = \frac{\ln 2}{9} \approx 0.077016 \text{ per month} \]

Set the model equal to ten thousand

Why: The same question in the same shape.

\[ 10{,}000 = 500e^{(\ln 2/9)t} \]

Divide by the stocked number

Why: Again a factor of twenty.

\[ 20 = e^{(\ln 2/9)t} \]

Take logarithms and solve

Why: Multiply by nine, divide by the logarithm of 2.

\[ t = \frac{9\ln 20}{\ln 2} \approx 38.90 \text{ months} \]

Check by scaling the first answer

Why: Nothing in the calculation changed except the doubling time, and the answer is proportional to it: nine sixths of 25.93 is 38.90, so a pond that doubles half again as slowly takes half again as long.

\[ \frac{9}{6}\cdot 25.93 = 1.5 \cdot 25.93 = 38.90 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 640 — Checkpoint 6.44

Try this before reading it. Everything is the same except the doubling time, which is now nine months instead of six, and you should be able to see the answer coming before you compute it.

The reason is in the check. The time to reach any given multiple of the starting population is the doubling time multiplied by the number of doublings needed, and the number of doublings has not changed: it is still 4.32, because the target is still twenty times the start. So the answer scales in proportion to the doubling time.

That scaling is worth internalising as a shortcut. If you double all the doubling times in a problem, every answer about how long doubles too. It is the fastest way to sanity-check a calculation of this kind, and it means you rarely have to redo the whole thing when one parameter changes.

29. Decay: the same equation with a minus sign

Concept

A shrinking quantity obeys the same model with a negative constant. Tradition keeps the letter positive and writes the minus sign into the exponent, so that a decay constant can be quoted as a positive number.

\[ y = y_0e^{-kt}, \qquad k > 0 \]

\[ y' = -ky_0e^{-kt} = -ky \]

Figure (svg): A curve falling from 2000 at time zero, steeply at first and then flattening, passing 1000 at 34.7 minutes and 500 at 69.3 minutes, marked with dots and dashed guide lines, and approaching but never touching the horizontal axis.

The decay curve is the growth curve read right to left. Each equal time step multiplies what is left by one half, so the quantity shrinks forever without ever arriving at zero.

Nothing new needs proving. Every result derived for growth holds with the constant replaced by its negative, and the curve is the growth curve read from right to left. Notice that it never touches the axis: an exponential reaches zero only in the limit, never at any finite time.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 640-641 — Rule: Exponential Decay Model and Figure 6.80

There is genuinely nothing new to learn here, and it is worth saying so plainly. Decay is growth with a negative constant. The convention of keeping the letter positive and putting a minus sign in the exponent exists so that a decay constant can be quoted as a positive number in a table.

Look at the figure and compare it with the growth curve a few slides back. It is the same curve read from right to left. The steepest part is at the beginning, where there is most material to decay, and the curve flattens as the material runs out, which is the proportionality assumption acting in reverse.

The detail to notice is that the curve never meets the axis. The exponential of any real number is strictly positive, so some material always remains, however long you wait. This matters practically in radiation safety and it matters on exams, where the claim that a sample is gone after a few half-lives is a standard wrong answer.

30. Half-life is the same formula over again

Concept

Ask when a decaying quantity falls to half its starting value. The same cancellation happens, and after two sign changes the same expression comes out.

\[ \frac{y_0}{2} = y_0e^{-kt} \]

\[ \frac{1}{2} = e^{-kt} \]

\[ -\ln 2 = -kt \;\Longrightarrow\; t = \frac{\ln 2}{k} \]

half-life — The time a decaying exponential quantity takes to fall to half of what is present. Like the doubling time, it depends on the rate constant alone.

The book stops to point this out, and it is worth stopping over. One expression, two names, and the only thing that decides which name applies is the sign of the constant in the exponent.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 642 — Definition: half-life

Watch the two sign changes in the third line. The logarithm of one half is the negative of the logarithm of 2, and the exponent already carries a minus sign, so the two negatives cancel and you are left with exactly the doubling-time expression.

The book stops to draw attention to this and so should you. One expression serves both purposes, and the only thing that decides which name it earns is whether the constant in the exponent is positive or negative. If you can remember the doubling time you already know the half-life.

Underneath the coincidence there is a reason, and it is the same cancellation as before. Both questions ask about a ratio, so the starting amount divides away and the ratios themselves, 2 and one half, are reciprocals. Taking logarithms turns a reciprocal pair into the same number with opposite signs.

31. Example 6.46: dating an artifact

Worked example

Carbon-14 decays with a half-life of 5730 years. Start with one hundred grams: how much survives fifty years, and how old is an artifact whose hundred grams have fallen to ten?

\[ \text{half-life} = 5730 \text{ years} \]

Convert the half-life into a rate constant

Why: The half-life formula, read backwards.

\[ 5730 = \frac{\ln 2}{k} \;\Longrightarrow\; k = \frac{\ln 2}{5730} \approx 1.2097\times 10^{-4} \]

Write the model

Why: One hundred grams in front, the fitted constant in the exponent.

\[ y = 100e^{-(\ln 2/5730)t} \]

Substitute fifty years

Why: Fifty years is under one percent of a half-life, so very little should be gone.

\[ y = 100e^{-(\ln 2/5730)(50)} \approx 99.40 \text{ g} \]

Set the model equal to ten grams

Why: Now the dating question.

\[ 10 = 100e^{-(\ln 2/5730)t} \]

Divide by the original amount

Why: One tenth of the carbon remains.

\[ \frac{1}{10} = e^{-(\ln 2/5730)t} \]

Take logarithms

Why: Both sides pick up a minus sign, which then cancels.

\[ -\ln 10 = -\frac{\ln 2}{5730}\,t \]

Solve for the age

Why: Multiply by 5730 and divide by the logarithm of 2.

\[ t = \frac{5730\ln 10}{\ln 2} \approx 19{,}035 \text{ years} \]

Figure (svg): The fraction of carbon-14 still present, plotted against thousands of years, falling through one half at 5730 years, one quarter at 11,460 and one eighth at 17,190; a dashed horizontal line at one tenth meets the curve at 19,000 years.

Three half-lives leave one eighth, so one tenth is a little further along: 3.32 half-lives, which is the 19,000 years the dating gives.

Check by counting half-lives

Why: Three half-lives leave an eighth and four leave a sixteenth, so a tenth sits between them; 19,035 years is 3.32 half-lives and two to that power is ten. The book rounds to the nearest hundred: about 19,000 years.

\[ \frac{19{,}035}{5730} \approx 3.322, \qquad 2^{-3.322} \approx 0.100 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 642-643 — Example 6.46

This is the model application of exponential decay and the steps are worth knowing by heart. The half-life is given, so the first job is always to convert it into a rate constant, using the half-life formula read backwards.

The two questions are again inverse to each other. Fifty years is a tiny fraction of a half-life, so almost nothing should be gone, and 99.40 grams out of 100 confirms that. The dating question is the reverse and needs a logarithm, and the arithmetic is clean because a tenth is a nice ratio.

The check is the one that protects you. Count half-lives in your head before trusting the calculator: a tenth lies between an eighth, which is three half-lives, and a sixteenth, which is four. So the answer has to lie between about 17,000 and 23,000 years, and 19,035 does. Carbon dating runs out of usefulness at around ten half-lives, which is why nobody carbon-dates a dinosaur.

32. Checkpoint 6.46: five hundred years, and a lighter bone

Worked example

The same isotope and the same starting hundred grams, with a longer wait and a heavier survivor.

\[ y = 100e^{-(\ln 2/5730)t} \]

Substitute five hundred years

Why: Still under a tenth of a half-life.

\[ y = 100e^{-(\ln 2/5730)(500)} \approx 94.13 \text{ g} \]

Set the model equal to twenty grams

Why: The second question: one fifth remains.

\[ 20 = 100e^{-(\ln 2/5730)t} \]

Divide by the original amount

Why: Leaving the exponential alone.

\[ \frac{1}{5} = e^{-(\ln 2/5730)t} \]

Take logarithms

Why: Both sides negative, so both minus signs go.

\[ \ln 5 = \frac{\ln 2}{5730}\,t \]

Solve for the age

Why: About 13,300 years, to the nearest hundred.

\[ t = \frac{5730\ln 5}{\ln 2} \approx 13{,}305 \text{ years} \]

Check against a whole number of half-lives

Why: Two half-lives, 11,460 years, would leave a quarter, and a fifth is less than a quarter, so the bone must be older than 11,460 years. It is, by about 1845.

\[ 11{,}460 \text{ yr} \to \tfrac14 = 0.25 > 0.20, \qquad 13{,}305 > 11{,}460 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 643 — Checkpoint 6.46

The same two questions with easier numbers. Do them before reading, and round the age to the nearest hundred years as the book asks, which gives 13,300.

The check is the one to practise because it needs no calculator at all. A fifth is less than a quarter, and a quarter is exactly two half-lives, so the bone must be older than two half-lives, which is 11,460 years. Any answer under that is wrong for a reason you can state in one sentence.

Notice also how insensitive the age is to measurement error at this end of the scale. Getting 20 grams wrong by a gram moves the age by only a couple of hundred years, because the logarithm is flat there. Nearer the limit of the method, at a gram or two remaining, the same absolute error moves the answer by thousands of years.

33. Exercise 353: from tenfold to a hundredfold

Worked example

A culture increases by a factor of ten in ten hours. How long does a hundredfold increase take?

\[ 10y_0 = y_0e^{10k} \]

Cancel the starting amount

Why: It divides out of both sides, as always.

\[ 10 = e^{10k} \]

Solve for the rate constant

Why: Take logarithms and divide by ten.

\[ k = \frac{\ln 10}{10} \approx 0.23026 \text{ per hour} \]

Set up the hundredfold question

Why: The same equation with a different factor.

\[ 100 = e^{kt} \]

Take logarithms

Why: And substitute the fitted constant.

\[ \ln 100 = kt = \frac{\ln 10}{10}\,t \]

Use that the logarithm of 100 is twice the logarithm of 10

Why: The logarithms cancel completely and no calculator is needed.

\[ t = \frac{10\ln 100}{\ln 10} = \frac{10(2\ln 10)}{\ln 10} = 20 \text{ hours} \]

Check without arithmetic at all

Why: Multiplying by a hundred is multiplying by ten twice, and each tenfold step takes the same ten hours whatever the amount present, so the answer had to be exactly twenty. The independence from the amount is doing the work.

\[ 100 = 10 \cdot 10 \;\Longrightarrow\; t = 10 + 10 = 20 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 644 — Exercise 353

Try this one by reasoning before you try it by algebra, because the reasoning is faster and it is the point of the exercise. Multiplying by a hundred is multiplying by ten and then by ten again.

Each tenfold step takes the same amount of time, whatever the amount you start that step with. This is the amount-independence of the doubling time, in the version with ten instead of two. So the answer is ten hours plus ten hours, exactly twenty, and no logarithm is needed.

The algebra confirms it in a pleasing way. The logarithm of 100 is twice the logarithm of 10, so the logarithms cancel completely and the answer is exact rather than approximate. Whenever an exponential problem asks about a power of the factor you were given, expect the logarithms to cancel like this and look for it.

34. Every question about time is one logarithm

Concept

Look back at the last five calculations. Bacteria, fish, carbon and money all led to the same three lines, because in each of them the unknown had climbed into the exponent and only one tool brings it back down.

\[ y_0 \cdot F = y_0e^{kt} \;\Longrightarrow\; F = e^{kt} \]

\[ \ln F = kt \]

\[ t = \frac{\ln F}{k} \]

The letter F is whatever factor the question asks about: 2 for a doubling, one half for a half-life, 20 for the pond, one tenth for the artifact, a hundred for the culture. The starting amount never reaches the answer and the rate constant always divides it. Recognising a how-long question and reaching straight for this shape removes most of the work.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 640-643 — the doubling-time and half-life derivations

Before reading the lines, flick back through the last five worked examples and notice how similar the middles were. Every one of them divided out a starting amount and then took exactly one logarithm. That is not a coincidence about those problems; it is the shape of the question.

The reason is structural. If the unknown is a time, it sits in the exponent, and the logarithm is the only operation that brings an exponent down to ground level. So the pattern is forced: divide by the leading coefficient so that the exponential stands alone, log both sides, divide by the rate constant.

Two consequences are worth stating. Because the factor is all that matters, the starting amount never appears in the answer to a how-long question, which is why doubling times and half-lives are properties of the substance rather than the sample. And because the rate constant divides, every such time is inversely proportional to it: a system that grows twice as fast takes half as long to do anything.

35. Complete the rate-constant table

Comparison

Comparison matrix

what is changingrate constantdoubling time or half-life
the bacteria of Table 6.10.02 per minute34.7 minutes
the fish in the pond0.1155 per month6 months
carbon-14 in a bone0.000121 per year5730 years
the coffee's excess over the room0.0835 per minute8.3 minutes
money at 5%, continuous0.05 per year13.9 years

Fill in every blank before revealing, and use the same two conversions throughout: the characteristic time is the logarithm of 2 divided by the rate constant, and the rate constant is the logarithm of 2 divided by the characteristic time.

The two middle rows go from a time to a constant, and the others go the other way. The row about the coffee is the one to inspect afterwards, because the constant there belongs to the excess of the coffee over the room and not to the coffee's temperature. Its 8.3 minutes means the excess halves every 8.3 minutes, which is a very useful way to think about the cooling problems in the next part.

Look down the last column when you are done. These characteristic times span from minutes to thousands of years, and yet they all come from the same one-line formula. That is the payoff of a model with a single parameter.

36. Which quantity is the exponential one?

Sorting

Sort into buckets

For each system, decide whether the quantity named is itself exponential, becomes exponential only after a constant is subtracted, or is not exponential at all.

exponential as it stands
bacteria in fresh, unlimited broth; the number of undecayed carbon-14 atoms; a balance earning continuous interest; how much hotter the coffee is than the room
exponential after subtracting a constant
the temperature of a cooling cup of coffee
not exponential
fish filling a small pond, over ten years; the human population of the Earth, over 300 years
yes
In each of these the rate really is proportional to the amount: more cells divide more often, more atoms decay more often, more money earns more interest, and a larger excess over the room sheds heat faster.
shift
The coffee's temperature heads for 70, not for 0, so it is not proportional to itself. Subtract the room's temperature and the excess that remains is a plain decaying exponential.
no
Both of these run into a limit. A small pond has finite food and space, and the Earth's population has been slowing for decades; the proportionality holds early and then fails.

Sort all seven before checking. The question is not whether the situation is roughly exponential in a loose sense; it is whether the specific quantity named has a rate proportional to itself.

The middle bucket has exactly one member, and it is the one the whole of Part 4 is about. The coffee's temperature is heading for 70, not for zero, so it cannot be proportional to itself: at 70 it would still be falling if it were. Subtract the room and the excess that remains does head for zero, and it is a plain decaying exponential.

The last bucket is about limits. A small pond has finite food and space, and the Earth's population growth rate peaked in the late 1960s and has been falling ever since. Both look exponential over a short stretch of data, which is exactly the trap the final part of the lesson takes apart.

37. Trap: two half-lives and nothing is left

Trap

The trap

Half-life of eight days, so after sixteen days:

\[ \tfrac12 + \tfrac12 = 1 \]

\[ \text{all gone?} \]

Wrong. Halvings multiply; they do not add.

The fix

Each half-life multiplies what remains by one half, so two of them leave a quarter, three leave an eighth, and ten leave about a thousandth. No finite number of halvings ever reaches nothing.

\[ \left(\tfrac12\right)^{2} = \tfrac14 \]

\[ y_0e^{-kt} > 0 \text{ for every } t \]

This error comes from treating a half-life as a fixed amount removed rather than a fixed proportion removed. If a half-life took away half of the original sample each time, two of them would indeed empty it.

But each half-life halves whatever is there at the start of that interval, not whatever was there at the beginning of time. So the amounts removed shrink: half, then a quarter, then an eighth, and the totals left are a half, a quarter, an eighth. Ten half-lives leave about a thousandth, and no finite number leaves nothing.

The defence is the formula itself. The exponential of any real number is strictly positive, so the product of a positive starting amount and a positive exponential is positive at every finite time. Decay approaches zero; it does not arrive.

38. Match each system to its constant

Matching

Match the pairs

  • a. coffee at 200 cooling in a 70 degree room
  • b. carbon-14 in a bone
  • c. money at 5%, compounded continuously
  • d. a culture that grows tenfold in 10 hours
  • w. k = 0.0835 per minute; the excess halves every 8.3 minutes
  • x. k = 0.000121 per year; half-life 5730 years
  • y. k = 0.05 per year; the balance doubles in 13.9 years
  • z. k = 0.2303 per hour; the count doubles every 3.0 hours

Why: Every one of these constants came from the same two lines: divide out the starting amount, then take one logarithm. Read them the other way and each gives a characteristic time, ln 2 over the constant, which is the doubling time when the constant is positive and the half-life when it is negative. Notice that the coffee's constant belongs to its excess over the room, not to its temperature.

Each pairing can be made two ways, from the constant or from the characteristic time, and you should check that both agree. The characteristic time is always the logarithm of 2 divided by the constant.

The coffee row is the one that carries the lesson. Its constant belongs to the excess over the room, not to the temperature, which is why the phrasing on the right-hand side says the excess halves every 8.3 minutes. If you paired it by remembering that coffee cools fast, you got the right answer for the wrong reason.

Notice how wide the range is. The same equation with the same shape of solution describes something that halves in eight minutes and something that halves in five thousand years. The rate constant is doing all the work, and that is why fitting it correctly from data is most of the labour in a real problem.

39. Newton's law of cooling

Section

Part 4

40. The difference cools, not the temperature

Concept

A hot object loses heat at a rate set by how much hotter it is than its surroundings, not by how hot it is. A cup already at room temperature is not cooling at all, however warm the room happens to be.

\[ T' = -k\left(T - T_a\right) \]

Figure (svg): Three cooling curves starting at 200, 150 and 110 degrees, all bending down toward a dashed horizontal line at 70 degrees which none of them crosses; the hottest curve is the steepest at the start.

Same room, same cup, same rate constant: only the starting excess differs. The hottest cup loses heat fastest, and all three flatten out on the room's temperature.

Three cups poured at different temperatures into the same room. The hottest falls fastest at the start, all three flatten out on the room's temperature, and none of them crosses it. That last fact is the one the equation is built to guarantee: when the difference is zero the rate is zero.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 641 — Newton's law of cooling

The physical statement comes first and it is easy to believe once you say it out loud. Heat flows because of a temperature difference, so the rate at which the cup loses heat depends on how much hotter it is than the air, not on where zero happens to sit on your thermometer.

The figure makes the same point three times. Three cups, one room, one rate constant, and only the starting temperature differs. The hottest cup falls fastest at the start because it has the largest excess. All three flatten onto the room's temperature, and none crosses it, because when the difference reaches zero the equation says the rate is zero.

There is one more check hiding in the statement. If the rate depended on the temperature itself rather than the difference, then converting from Fahrenheit to Celsius would change the physics, because zero sits in a different place on the two scales. A law that depends on your choice of thermometer is not a law.

41. One change of variable fixes it

Concept

The cooling equation is not yet in the model's shape, because the extra constant on the right spoils the proportionality. One substitution removes it, and the book's phrasing is worth copying: we want the derivative proportional to the function.

\[ y(t) = T(t) - T_a \]

\[ y'(t) = T'(t) - 0 = T'(t) \]

\[ y' = -ky \]

\[ T - T_a = \left(T_0 - T_a\right)e^{-kt} \]

The ambient temperature is a constant, so it differentiates away and leaves the excess obeying the plain decay model. Solve for the excess, then add the room back at the very end. Every cooling problem in this section is that sentence.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 641 — the change of variables

This is a small idea with a large payoff, and it is the first substitution in a differential equation you have met. The equation is nearly the model already; the only thing spoiling it is the constant being subtracted on the right.

So name that combination as a new function. Because the ambient temperature is a constant, differentiating the new function gives exactly the derivative of the old one, with nothing left over. The equation in the new variable is the plain decay model, whose solution you already have, and undoing the substitution at the end gives the cooling formula.

The habit generalises far beyond this section. When an equation is almost in a solved form except for a constant offset, shift the variable to absorb the offset. You will meet the same move in Calculus II when you solve equations with an equilibrium value, and the equilibrium is always the thing you subtract.

42. Reading the cooling formula

Notation

Annotate

On: \( T = (T_0 - T_a)e^{-kt} + T_a \)

  • The initial excess over the room. This, and not the temperature, is the quantity that decays exponentially.
  • The plain decay factor. It runs from 1 at time zero down toward 0, so the excess shrinks by the same proportion in every equal time step.
  • Add the room back at the end. Forget this term and your model cools the coffee straight through the freezing point.
  • The constant depends on the cup, the liquid and the air, never on how hot the coffee started. Two identical cups in the same room share one k.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 641 — the cooling model

Work through the four annotations and notice that the formula has three constants in it, and that in a typical problem you are given two of them outright. The initial temperature and the room temperature are usually stated, so the only unknown is the rate constant, and one further reading fixes it.

The second annotation is worth dwelling on. The exponential factor runs from 1 down toward 0 and is doing exactly the same job it did for radioactive decay: it multiplies the excess by the same fraction in every equal time step. The excess has a half-life, just like an isotope.

The third annotation is where the marks are. Forgetting to add the room back is the single most common error in this material, because the exponential part of the work feels like the hard part and the addition feels like nothing. Write the final step down every time.

43. What the wrong model predicts

Picture it

Figure (svg): Two curves through the same measured point at 2 minutes and 180 degrees: one levels off at a dashed line at 70 degrees, the other keeps falling toward zero and is below 20 degrees by 45 minutes.

Both curves pass through the single measured point, so one reading cannot tell them apart. Only the physics can: coffee stops at room temperature.

Both curves pass through the single measurement, 180 degrees at two minutes, so one reading cannot tell them apart. The dashed curve is the model fitted to the temperature itself; it heads for zero and keeps going. The solid one is fitted to the excess over the room and levels off where physics says it must.

There is a sharper test still. Redo the wrong fit in Celsius rather than Fahrenheit and it predicts a different final temperature, because zero sits in a different place on the two scales. A model whose answer depends on which thermometer you own is not a model.

Look carefully at the point where the two curves cross. Both pass through the one measurement, 180 degrees at two minutes, so the data alone cannot choose between them. That is the uncomfortable lesson of this picture: a single reading never validates a model.

Follow them to the right and they separate completely. The correct model settles onto the room's temperature and stays there. The wrong one keeps falling and is under 20 degrees by three quarters of an hour, which would mean coffee freezing solid in a warm kitchen.

The sharper test is the one in the second paragraph on the slide. Redo the wrong fit in Celsius and it predicts a different final temperature from the Fahrenheit version, because the two scales put zero in different places. Whenever a model's prediction depends on your choice of units in a way physics does not, the model is wrong, and you can find that out without any data at all.

44. Before you fit a cooling curve

Step zero

\[ 190 \text{ degrees, a } 68 \text{ degree room, } 150 \text{ degrees after } 8 \text{ minutes} \]

Discussion prompt

A bowl of soup is described above. Before fitting anything, write down the two subtractions you must do first, then say which of the three constants in the cooling formula is still unknown and what single equation determines it.

Write your answer down before revealing. Thirty seconds of setting up saves the whole problem here, and the temptation is to start substituting numbers into the exponential immediately.

The two subtractions are the initial excess and the excess at the moment of the second reading. Everything in the calculation from that point on is about those two differences, and the raw temperatures never appear again until the final line, when the room is added back.

Counting the unknowns is the other half of the setup. Three constants, two of them given, so only the rate constant has to be found, and a single further reading determines it. If you find yourself with two unknowns you have almost certainly forgotten that the initial temperature was stated.

45. Example 6.45: when can the coffee be served?

Worked example

Baristas call coffee servable between 155 and 175 degrees Fahrenheit. It is poured at 200 into a 70 degree room and measured at 180 after two minutes.

\[ T = (T_0 - T_a)e^{-kt} + T_a, \qquad T_0 = 200, \; T_a = 70 \]

Substitute the two-minute reading

Why: Everything is known except the rate constant.

\[ 180 = (200-70)e^{-k(2)} + 70 \]

Subtract the room temperature

Why: This is the step that exposes the pure exponential.

\[ 110 = 130e^{-2k} \]

Divide by the initial excess

Why: Leaving the exponential alone on the right.

\[ \frac{11}{13} = e^{-2k} \]

Take logarithms

Why: The logarithm of a quotient is a difference of logarithms.

\[ \ln 11 - \ln 13 = -2k \]

Solve for the rate constant

Why: Negate and halve; the constant is positive because 13 is larger than 11.

\[ k = \frac{\ln 13 - \ln 11}{2} \approx 0.083527 \text{ per minute} \]

Figure (svg): The coffee's cooling curve from 200 degrees, with the band between 155 and 175 degrees shaded, the measured point at 2 minutes and 180 degrees marked, and the two crossings of the band marked at about 2.6 and 5.1 minutes.

The whole problem, drawn: one measured point fixes the curve, and the shaded band is the answer. The coffee is inside it for about two and a half minutes.

Set the model equal to 175

Why: The top of the serving window.

\[ 175 = 130e^{-kt} + 70 \]

Subtract the room temperature

Why: The same move as before, and always the first one.

\[ 105 = 130e^{-kt} \]

Take logarithms and divide by the constant

Why: The first time the coffee is cool enough.

\[ t = \frac{\ln(130/105)}{k} \approx 2.56 \text{ minutes} \]

Repeat at 155 degrees

Why: The bottom of the window, by the identical route.

\[ 85 = 130e^{-kt} \;\Longrightarrow\; t = \frac{\ln(130/85)}{k} \approx 5.09 \text{ minutes} \]

Check the fitted curve against the measurement

Why: Putting two minutes back into the finished model returns exactly the measured 180, so the constant was fitted correctly; rounded to the nearest half minute the window opens at 2.5 minutes and closes at 5.

\[ T(2) = 130e^{-2(0.083527)} + 70 = 130\left(\tfrac{11}{13}\right) + 70 = 180 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, pp. 641-642 — Example 6.45

This is the longest calculation in the lesson, and it is really three short ones: fit the constant, then solve twice. Do not let its length disguise how repetitive it is. The three parts share the same two moves, subtract the room and take a logarithm.

The book carries the constant through in exact form, as a difference of two logarithms over 2, rather than rounding it early. Copy that habit. Rounding the rate constant to two decimal places here shifts the answers by a tenth of a minute or so, which the book's rounding to the nearest half minute would hide, but in a longer problem it will not.

Look at the figure while you read the last three steps. The shaded band is the answer, the curve enters it at about two and a half minutes and leaves it at about five, and the measured point sits on the curve above the band. The check at the end puts two minutes back into the finished model and recovers exactly the 180 that was measured, which is how you know the constant was fitted correctly.

46. Put the room back

Fill the middle

The excess has been solved for. One term still has to be restored before the answer is a temperature.

Fill in the blanks

T - T_a = (T_0 - T_a)e^T_a \;\Longrightarrow\; T = (T_0 - T_a)e^___ + ___

Why: Add the ambient temperature back to both sides. The exponential describes the excess over the room, so on its own it is a temperature difference and not a temperature. Leaving this term out is what produces a model that cools everything toward zero degrees on whichever scale happens to be in use.

This blank is worth one mark on an exam and it is the one most often dropped. Fill it in before checking.

The reason it is dropped is psychological. By the time you reach this line you have done the interesting work, and adding a number back feels like bookkeeping. But the exponential describes a temperature difference, so on its own it is not a temperature at all, and an answer that omits this term is out by the whole room temperature.

A useful habit: at the end of any cooling problem, ask whether your answer is above the room's temperature and below the starting temperature. If it is not, something has gone wrong, and nine times out of ten it is this term.

47. Checkpoint 6.45: a warmer room

Worked example

The same pour into a warmer room, and after two minutes the coffee has only reached 185.

\[ T_0 = 200, \quad T_a = 75, \quad T(2) = 185 \]

Substitute the reading

Why: The initial excess is 125 this time, not 130.

\[ 185 = (200-75)e^{-2k} + 75 \]

Subtract the room temperature

Why: The excess has fallen from 125 to 110.

\[ 110 = 125e^{-2k} \]

Divide by the initial excess

Why: A clean decimal, for once.

\[ 0.88 = e^{-2k} \]

Take logarithms and solve

Why: The logarithm of 0.88 is negative, so the constant is positive.

\[ k = -\frac{\ln 0.88}{2} \approx 0.063917 \text{ per minute} \]

Solve for 175 degrees

Why: An excess of 100 out of the original 125.

\[ 100 = 125e^{-kt} \;\Longrightarrow\; t = \frac{\ln 1.25}{k} \approx 3.49 \]

Solve for 155 degrees

Why: An excess of 80 out of 125.

\[ 80 = 125e^{-kt} \;\Longrightarrow\; t = \frac{\ln(125/80)}{k} \approx 6.98 \]

Check against the cooler room

Why: The warmer room gives a smaller rate constant, 0.0639 against 0.0835, so everything happens later: to the nearest half minute the window now opens at 3.5 minutes instead of 2.5 and closes at 7 instead of 5.

\[ 0.063917 < 0.083527, \qquad 3.5 > 2.5, \quad 7.0 > 5.0 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 642 — Checkpoint 6.45

Run this one yourself. The structure is identical to Example 6.45 and only two numbers have changed: the room is five degrees warmer and the two-minute reading is five degrees higher.

The arithmetic is slightly kinder here, because the excess falls from 125 to 110, which is 0.88 exactly. But notice that the rate constant comes out smaller than in the example, 0.0639 against 0.0835. A warmer room means a smaller temperature difference, which means slower cooling.

That is why every answer moves later: the window opens at three and a half minutes instead of two and a half, and closes at seven instead of five. You could have predicted the direction of both changes before doing any arithmetic, and predicting the direction first is a good discipline for catching sign errors.

48. Trap: cooling toward zero

Trap

The trap

Fitting the temperature itself, with no room in sight:

\[ T = 200e^{-kt} \]

\[ T \to 0 \]

Wrong. Coffee does not approach absolute zero.

The fix

The rate is proportional to the excess over the room, so the excess is what decays. Subtract the ambient temperature before fitting and add it back before reporting; the model then levels off exactly where the previous picture shows it must.

\[ T - T_a = (T_0 - T_a)e^{-kt} \]

\[ T \to T_a \]

This is the same error as the wrong curve you saw earlier in this part, written out as algebra. It happens when you reach for the decay formula by reflex, because the temperature is going down and decay is what going down looks like.

The equation is quite clear about what decays. The rate is proportional to the difference between the object and its surroundings, so that difference is the quantity in the model, and the temperature is that quantity plus a constant. Put the constant in and the curve levels off in the right place.

If you want one sentence to remember: subtract the room before you fit, add the room before you report. Both halves matter, and dropping either one produces an answer that is wrong by exactly the room temperature.

49. Find the error: a logarithm across a sum

Error analysis

Annotate

On: \( 180 = 130e^{-2k} + 70 \;\Longrightarrow\; \ln 180 = -2k + \ln 70 \)

  • The logarithm of a sum is not the sum of the logarithms. There is no rule at all for ln(a + b); the product rule turns products into sums, never the other way round.
  • Subtract the 70 first, so that the exponential term stands alone and the line reads 110 equals 130 times the exponential. Only an isolated exponential may be logged.
  • The bad line rearranges to k = (ln 70 − ln 180)/2, about −0.472. The constant comes out negative, so the model has the coffee heating up without limit, and it is also about six times too large.
  • Before taking a logarithm, check that one side is a single exponential with nothing added to it and nothing multiplied onto it except a constant you can divide away.

Read the line and try to find the problem before opening the annotations. The numbers are right and the model is right; what is wrong is a single algebraic step.

There is a logarithm rule for products and a rule for quotients and a rule for powers, and no rule at all for sums. The logarithm of a sum cannot be broken up, and there is no identity waiting to be remembered. The only legitimate move is to isolate the exponential first by subtracting the 70, and then take logarithms of a line that has nothing added to it.

The third annotation shows how loudly the answer complains. The rate constant comes out negative, which would mean the coffee gets hotter over time, and about six times too large in size. Wrong algebra usually produces a wrong number that looks plausible; occasionally, as here, it produces one that is absurd, and it is worth training yourself to notice.

50. Exercise 366: the resting turkey

Worked example

A turkey comes out of the oven at 165 degrees into a 70 degree apartment, and ten minutes later it reads 155. What does it read after twenty?

\[ T - 70 = (165 - 70)e^{-kt} = 95e^{-kt} \]

Substitute the ten-minute reading

Why: The excess has fallen from 95 to 85.

\[ 85 = 95e^{-10k} \]

Solve for the rate constant

Why: Divide, log, divide by ten.

\[ k = \frac{1}{10}\ln\frac{95}{85} \approx 0.011123 \text{ per minute} \]

Substitute twenty minutes

Why: Twenty minutes is two of the ten-minute steps, so the decay factor is squared.

\[ T(20) - 70 = 95e^{-20k} = 95\left(\frac{85}{95}\right)^{2} \]

Simplify the square

Why: One factor of 95 cancels.

\[ 95\cdot\frac{85^{2}}{95^{2}} = \frac{7225}{95} \approx 76.05 \]

Add the room temperature back

Why: The last step, and the one most often dropped.

\[ T(20) \approx 76.05 + 70 = 146.05 \text{ degrees} \]

Check with the equal-ratio rule

Why: The excess fell from 95 to 85 in the first ten minutes, a factor of about 0.895. The second ten minutes must multiply by the same factor again, giving an excess of 76.05, not another straight drop of 10 to an excess of 75 and a reading of 145.

\[ 95 \to 85 \to \frac{85^{2}}{95} = 76.05 \;\ne\; 75 \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 644 — Exercise 366

A straight application, but the shortcut in the middle is worth learning. Twenty minutes is two of the ten-minute intervals, so instead of computing the rate constant and exponentiating, you can simply apply the same factor twice.

That works because the excess is multiplied by a fixed factor in every equal time step, which is the equal-intervals-equal-factors idea from Part 3. The excess fell from 95 to 85, a factor of about 0.895, so in the next ten minutes it falls from 85 to about 76, not from 85 to 75.

The check names the wrong answer explicitly, and it is the wrong answer nearly everyone gives first. Cooling by ten degrees in the first ten minutes does not mean cooling by another ten in the second ten. The drops get smaller, because the excess driving them is smaller.

51. Where the model breaks down

Section

Part 5

52. Nothing grows exponentially for long

Concept

The model rests on one assumption: that the growth per unit of stuff stays the same however much stuff there is. Nothing in the physical world satisfies that for long, because food, space, fuel and customers all run out.

\[ \frac{1}{y}\frac{dy}{dt} = k \quad \text{for every value of } y \]

The book raises the same worry in a milder form on page 638, warning that a continuous function is being used to model something that is really a whole number of bacteria. The honest reading of a fitted exponential is local: it describes the stretch of data it was fitted to, and a short way beyond, and nothing further.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 638 — the caution about modelling discrete behaviour continuously

This part of the lesson is not examinable in the usual sense and it is the part a scientist most needs. The exponential model makes a claim about every size the quantity could ever reach, and that claim is false for every real system.

The failure is not subtle and it is not about measurement error. Bacteria need broth, animals need food and territory, investments need someone on the other side of the trade. Each of these puts a ceiling somewhere, and above the ceiling the proportionality assumption is simply not true any more.

So read a fitted exponential the way you would read a tangent line: an excellent local approximation that gets worse the further you go from where it was fitted. The next two slides show exactly how bad the extrapolation can get.

53. Three days of hourly doubling

Worked example

A bacterium that divides once an hour, starting from a single cell. Follow the model literally and see what it claims.

\[ N(t) = 2^{t}, \qquad t \text{ in hours} \]

Substitute three days

Why: Seventy-two hours, so seventy-two doublings.

\[ N(72) = 2^{72} \approx 4.72\times 10^{21} \]

Convert the count to a mass

Why: A bacterium weighs roughly one picogram, which is ten to the minus twelve grams.

\[ 4.72\times 10^{21} \cdot 10^{-12}\,\text{g} = 4.72\times 10^{9}\,\text{g} \approx 4700 \text{ tonnes} \]

Push the model further still

Why: Set the mass equal to the mass of the Earth, about 5.97 times ten to the twenty-seventh grams, and solve for the hour.

\[ 2^{h}\cdot 10^{-12} = 5.97\times 10^{27} \;\Longrightarrow\; h = \frac{\ln\left(5.97\times10^{39}\right)}{\ln 2} \approx 132 \]

Check what that number means

Why: A hundred and thirty-two hours is five and a half days, so the model claims one cell outweighs the planet inside a week. No culture has ever done this, so the proportionality must fail long before then, and it does: the broth runs out.

\[ \frac{132}{24} \approx 5.5 \text{ days} \]

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 638 — why the model must be interpreted in context

The arithmetic here is easy; the point is what it feels like. A single cell, doubling every hour, gives nearly five thousand tonnes of bacteria in three days. Nothing like that has ever happened.

Push it further and the model becomes comic. At about a hundred and thirty-two hours, five and a half days, the colony from one cell outweighs the Earth, and a couple of days after that it outweighs the Sun. The arithmetic is not wrong; the assumption is.

So the argument runs backwards. Because the conclusion is absurd, something in the premises must fail, and the premise that fails is the proportionality. In a real flask the growth rate collapses within a day as the broth is exhausted, which is precisely what the next slide's repair describes.

54. What replaces it: a brake on the rate

Concept

The standard repair keeps the proportionality and multiplies it by a factor that closes as the quantity approaches a ceiling. The ceiling is called the carrying capacity.

\[ \frac{dy}{dt} = ky\left(1 - \frac{y}{M}\right) \]

Figure (svg): Two curves from the same small start and the same early growth rate: one keeps curving upward and leaves the top of the frame, the other bends over and flattens onto a dashed ceiling line at 100.

For the first ten units of time the two curves stay within about two percent of each other. Early data cannot choose between them, which is why an exponential fit says nothing about the far future.

When the quantity is small against the ceiling the bracket is nearly one and the equation is the old one; the two curves drawn here differ by about two percent at ten units of time and are still within five percent at fifteen. As the quantity nears the ceiling the bracket closes toward zero and growth stops. Early data cannot distinguish the two, which is exactly why an exponential fit is worthless as a long extrapolation.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 645 — Exercise 373, on why exponential growth fails to predict the future

The repair is minimal, which is why it is standard. Keep the proportionality and multiply it by a factor that is nearly one when the quantity is small and nearly zero when the quantity approaches the ceiling.

Trace the consequences at the two extremes. For small quantities the bracket is essentially one, so the equation is the exponential model and the curve is indistinguishable from an exponential. As the quantity nears the ceiling the bracket closes to zero, the rate goes to zero, and the curve flattens out on the ceiling.

Now look at the picture and be uncomfortable. At ten units of time the two curves differ by about two percent, and at fifteen they are still within five, which is smaller than the noise in most real data. So data over that range cannot tell you which model is right, and fitting an exponential to it and extrapolating is a decision you have made, not a result the data gave you.

55. The world population table

Real world

Figure (svg): Seven measured world populations by decade from 1950 to 2010 drawn as dots, with the fitted exponential curve running through them and continuing past them to meet a dashed line at ten thousand million at about 82 years after 1950; a separate dot marks the true 2020 figure well below the curve.

The fit tracks the dots from 1950 to 2010 well, then runs ahead of reality: it wants 8255 million in 2020 and the world had about 7795.

Discussion prompt

The table of world population by decade since 1950 is fitted by the curve drawn here. Say what the fit gets right, where it starts to go wrong, and what the sign of the second derivative of the fitted curve commits the model to claiming about the future.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 645 — Exercises 370 to 372

Write down your reading of the graph before revealing the answer. What does the fit do well, and where exactly does it start to come apart?

Over the seven decades of the table it is genuinely good, within a few percent everywhere. The trouble is visible at the right-hand end: the model wants 8255 million for 2020 and the true figure was about 7795 million, an overshoot of nearly half a billion people, and the gap widens every year.

The second derivative is the part worth arguing about. It is positive for all time and growing, so the model does not merely predict growth, it predicts ever-accelerating growth, forever. The real growth rate peaked around 1968 and has been falling since. No amount of curve fitting to data from the 1950s could have discovered that, which is exactly the limitation Exercise 373 is pointing at.

56. When does the fit reach ten billion?

Estimation

\[ 10{,}000 = 2686e^{0.01604t} \]

Predict first

Ten billion is ten thousand million. Using the fitted curve only, in which year does the world population reach it?

  • 2032
  • 2050
  • 2075
  • 2100

Correct: 2032

Why: Divide by 2686 to get 3.723, take logarithms to get 1.3146, and divide by 0.01604 to get t about 82, so the year is 2032. The point of Exercise 373 is that this is not a forecast. The same curve already overshoots 2020 by more than four hundred million, and its second derivative commits it to ever-faster growth, which contradicts the slowdown the data since 1970 actually show. Current projections put ten billion somewhere in the 2050s.

OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay §6.8, p. 645 — Exercise 373

Commit to an option before revealing. The calculation is a single logarithm and you can do it in your head to within a decade if you notice that ten thousand is about 3.7 times 2686.

The fit says 2032, and you should not believe it. The same curve already overshoots 2020 by more than four hundred million, and the error is growing, so its prediction for a date twelve years further on is worse still. Serious projections put ten billion in the 2050s or later, and some of them never reach it.

This is the whole argument of the last part of the lesson, in one number. The exponential model is excellent where it was fitted, indefensible far outside, and nothing in the fit itself warns you which is which. The warning has to come from knowing what the system is and where its limits are.

57. Putting it together

Section

Part 6

58. Pattern: any growth or decay problem, in order

Pattern

Figure (svg): A two-tier flow diagram. The top tier asks what is changing and sends an amount straight to the model while a temperature is first replaced by its excess over the room. The bottom tier asks what is wanted and sends a how-much question to a substitution, a how-long question to one logarithm, and a rate question to a fit from a second reading.

Two decisions and then arithmetic. The first tier is where cooling problems are won or lost; the second tier decides whether you substitute or take a logarithm.
  1. Name the quantity that actually decays. In a cooling problem it is the excess over the ambient temperature, never the temperature.
  2. Write the model with the value at time zero as the leading coefficient and the rate constant in the exponent.
  3. Fit the rate constant from one further reading: divide out the coefficient, then take one logarithm.
  4. Answer the question asked. How much at a given time is a substitution; how long until a given value is one logarithm.
  5. Undo any shift. Add the ambient temperature back before reporting a temperature.
  6. Sanity-check against the doubling time or half-life, and refuse to extrapolate far past the data.

This is the order to work in for every problem in the section, and the first step is the only one that requires judgement. Everything after it is mechanical.

The first step is also where the marks are lost. Deciding what actually decays is the difference between a cooling problem done right and a cooling problem done wrong, and it is a decision about the physical situation rather than about the algebra. If there is an ambient constant anywhere, the thing that decays is the excess over it.

The second tier of the diagram sorts the questions by what they ask for. A how-much question is a substitution and needs no logarithm at all. A how-long question and a what-rate question both need exactly one logarithm, taken only after the exponential has been left standing alone. If you find yourself taking two logarithms, stop and look for an algebra error.

59. Put a how-long calculation in order

Ranking

Put in order

Order the steps of a complete answer to a question of the form: how long until this quantity reaches a given value?

  1. Decide what decays: the quantity itself, or its excess over an ambient constant
  2. Put the value at time zero in front as the leading coefficient
  3. Use a second reading to fit the rate constant
  4. Set the fitted model equal to the target value
  5. Divide out the leading coefficient so the exponential stands alone
  6. Take logarithms and divide by the rate constant

Why: The first step is the one people skip, and it is the only one that is a judgement rather than a calculation: in a cooling problem the thing that decays is the excess over the room. After that the order is forced, because the rate constant must be known before the model can be set equal to anything, and the exponential must stand alone before a logarithm may be taken.

Drag the steps into order before checking. The only genuinely debatable position is the first one, and the reason it is first is that it is not a calculation: it is a decision about what the model is even about.

After that the order is forced by dependency. You cannot set the model equal to a target until you have a model, and you cannot have a model until the rate constant is fitted. You cannot take a logarithm until the exponential stands alone, which means dividing out the leading coefficient first.

The most common real-world scrambling of this list is doing step five before step four, that is dividing before setting the model equal to anything. It usually still works, but it makes it very easy to divide by the wrong thing when the problem has a shift in it.

60. Check: solving the equation

Check

Check your understanding

Which function satisfies the equation saying the derivative equals three times the function, together with the value 5 at time zero?

  • A. y = 5e³ᵗ (correct)
  • B. y = 3e⁵ᵗ
  • C. y = 5 + e³ᵗ
  • D. y = e³ᵗ⁺⁵

Answer: A

Why: The rate constant belongs in the exponent and the initial value belongs in front. Differentiating 5e³ᵗ gives 15e³ᵗ, which is three times the function, and at time zero the value is 5. Both requirements hold.

Why B tempts people
This has the two constants swapped. Its derivative is five times itself, not three, and its value at time zero is 3.
Why C tempts people
The constant was added rather than absorbed as a factor. Its derivative is 3e³ᵗ, which is not three times the whole function, because the added 5 survives in the function but not in the derivative.
Why D tempts people
The 5 is inside the exponent, so this equals e⁵ times e³ᵗ. The derivative is right but the value at time zero is about 148, not 5.

Both constants in this question have a home, and the whole question is whether you put them in the right ones. The rate constant lives in the exponent and the initial value lives in front.

Test each option against both requirements, the differential equation and the value at time zero. Option B fails both, in mirror image. Option C gets the initial value right by accident and fails the equation, because the added constant does not survive differentiation. Option D is the interesting one: its derivative really is three times itself, so it solves the equation, but its value at time zero is the exponential of 5, about 148.

That last distinction is the point of the question. Solving the differential equation is not the same as solving the initial-value problem. The equation has infinitely many solutions; the initial condition picks exactly one.

61. Check: half-lives

Check

Check your understanding

A radioactive sample has a half-life of 12 days. What fraction of it is left after 30 days?

  • A. about 18% (correct)
  • B. about 25%
  • C. about 40%
  • D. nothing, since 30 days is more than two half-lives

Answer: A

Why: Thirty days is 2.5 half-lives, so the fraction remaining is one half raised to the power 2.5, which is about 0.1768, or roughly 18 percent. It sits between the quarter left after two half-lives and the eighth left after three, as it must.

Why B tempts people
A quarter is what remains after exactly two half-lives, that is 24 days. Six more days of decay must take it lower.
Why C tempts people
Forty percent is close to what is left after 1.3 half-lives, about 16 days. Halving is a good deal faster than this option suggests.
Why D tempts people
Half-lives multiply, they do not add. Two of them leave a quarter, not nothing, and no finite number of them ever empties the sample.

Count the half-lives first. Thirty days divided by twelve is two and a half of them, so the answer must sit between what two half-lives leave and what three half-lives leave.

Two leave a quarter and three leave an eighth, so the answer is between 12.5 and 25 percent, which already eliminates two of the options. One half raised to the power two and a half is about 0.177, so roughly 18 percent, and the fact that fractional half-lives make sense at all is because the model is a continuous exponential and not a sequence of discrete events.

Option D is the error from the trap slide, in a fresh disguise. Half-lives multiply, so no whole number of them ever empties the sample, and a fractional number certainly does not.

62. Check: cooling

Check

Check your understanding

A cup at 190 degrees in a 70 degree room cools to 150 degrees in 10 minutes. What does it read 10 minutes after that?

  • A. about 123 degrees (correct)
  • B. 110 degrees
  • C. about 118 degrees
  • D. about 100 degrees

Answer: A

Why: Work with the excess over the room. It went from 120 to 80 in the first ten minutes, a factor of two thirds, so the next ten minutes multiply it by two thirds again, giving 53.3. Add the room back: about 123 degrees.

Why B tempts people
This subtracts another 40 degrees, treating the cooling as a constant rate. Cooling slows down as the cup approaches the room, so the second drop must be smaller than the first.
Why C tempts people
This applies the ratio to the temperature itself rather than to the excess, giving 190 times the square of 150 over 190. That is the error the whole section is built to prevent.
Why D tempts people
Two thirds of 150 is 100, but the factor applies to the excess over the room, not to the reading on the thermometer.

Do not touch the thermometer readings directly. Convert to excesses over the room first: 120 degrees of excess at the start, 80 degrees after ten minutes.

That is a factor of two thirds in the first ten minutes, so the second ten minutes multiplies by two thirds again, giving an excess of 53.3. Add the room back and the reading is about 123 degrees. Notice that the cup fell 40 degrees in the first interval and only about 27 in the second, because cooling slows as the excess shrinks.

Option C is the one to understand rather than merely avoid. It applies the correct ratio to the wrong quantity, the temperature rather than the excess. It is the same mistake as the wrong curve in Part 4, and it gives an answer close enough to the right one that it will not look obviously wrong.

63. Explain the cancellation

Explain it to yourself

Discussion prompt

In two or three sentences, explain why the doubling time of a growing population and the half-life of a decaying isotope are given by the same expression, and why neither one mentions how much you started with.

Write your two or three sentences before revealing. Being able to say this out loud is worth more than being able to quote the formula, because it tells you when to expect a starting amount to matter and when not to.

The core is that both questions ask about a ratio. Reaching twice as much, or falling to half as much, is a statement about a factor, and factors do not care about scale. So the starting amount appears on both sides of the equation and divides away before any logarithm is taken.

The second half is about why the same expression covers both. The two ratios are reciprocals, so their logarithms are the same number with opposite signs, and the minus sign already built into the decay exponent cancels the other one. That is why the book remarks that it is the same expression as for the doubling time.

64. Exit ticket

Exit ticket

\[ 80 \text{ mg falls to } 50 \text{ mg in } 6 \text{ hours} \]

Discussion prompt

Find the half-life of this isotope, and say how much is left after a full day. Then say, in one sentence, why you did not need to know that the sample weighed 80 milligrams to answer the first question.

This ties the whole lesson together in one problem and you should attempt all three parts before revealing. Fit the constant, convert it to a half-life, then answer the how-much question.

There is a shortcut in the last part worth spotting. A full day is exactly four of the six-hour intervals given, so the remaining fraction is five eighths multiplied by itself four times, and you never need the rate constant at all. Equal intervals, equal factors, one last time.

The final part of the question is the idea to carry out of the lesson. The half-life came from the ratio 50 to 80, and the milligrams cancelled on the first line. That is why an isotope has one half-life rather than one per sample, and it is the same cancellation that made doubling times, tripling times and tenfold times all independent of how much you started with.

65. Recap

Recap

questionwhat you writewhat comes out
rate proportional to amountdy/dt = kyy = y₀eᵏᵗ, fixed by one initial value
how long to multiply by a factorfactor = eᵏᵗt = ln(factor)/k, independent of y₀
doubling time, half-life2 = eᵏᵗ, 1/2 = e⁻ᵏᵗthe same expression, ln 2 over k
money compounded continuouslybalance = Peʳᵗthe limit of n compoundings as n grows
a cooling objectT − Tₐ = (T₀ − Tₐ)e⁻ᵏᵗthe excess decays; add the room back
far-future predictiondy/dt = ky(1 − y/M)the exponential is local, and only local

\[ y = y_0e^{kt}, \qquad T_2 = T_{1/2} = \frac{\ln 2}{k} \]

Chapter 6 ends here. The single equation of this section is the first differential equation of the course, and the technique that solved it, separating the variables, is the doorway into the whole subject: Calculus II opens with integration methods and then returns to differential equations properly.

Stewart, Calculus: Early Transcendentals 8e, §3.8 Exponential Growth and Decay §3.8, pp. 237-244 — the same material in Stewart

Six rows, one equation. Everything in the table is the solution of the proportional-change equation with different questions asked of it, and the one line under the table is worth having by heart: the solution, and the single expression that serves as both doubling time and half-life.

Two habits are more valuable than the formulas. First, in any problem with an ambient constant, find what actually decays before writing anything down. Second, whenever a question is about a factor rather than an amount, expect the starting value to cancel, and be suspicious of an answer in which it has not.

One last thing to carry forward. Every model in this section was a decision to believe a proportionality, and that decision is where the physics or the biology lives, not in the algebra that followed. When you meet a new situation, the question to ask first is whether the rate really is proportional to the amount, and to what amount, and over what range of amounts you are prepared to believe it.

Sources

  1. OpenStax Calculus Volume 1, §6.8 Exponential Growth and Decay — Strang & Herman, OpenStax / Rice University, CC BY-NC-SA 4.0, pp. 636-645
  2. Stewart, Calculus: Early Transcendentals 8e, §3.8 Exponential Growth and Decay — James Stewart, Cengage Learning, 2016, pp. 237-244

Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.

Book on Wyzant · Text (657) 465-8108