Defining the natural logarithm as an integral, proving its algebraic properties from that definition, defining e and the exponential as its inverse, giving irrational exponents a meaning, and repairing the circularity of the earlier treatment.
Subject: Calculus I · 71 slides · symbolic lesson
Open the interactive version of this deck
Title
Calculus I · Section 6.7
Building the logarithm out of an integral, and the exponential out of the logarithm
Objectives
Chapters 1 and 3 used logarithms and exponentials freely. This section goes back and builds both from scratch, using nothing but the integral of Chapter 5, and repairs a gap that has been open since the first week of the course.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, pp. 626-635 — learning objectives 6.7.1 to 6.7.7
This section is unusual, and it helps to know why before you start. Everything in it is about functions you have been using since the first weeks of the course: the natural logarithm, the number e, the exponential, and powers with any base at all. Nothing new is being introduced. What is new is that all of it is going to be built properly, from the one thing Chapter 5 defined completely, which is the definite integral.
The reason is a gap. Chapter 1 defined the logarithm as the inverse of the exponential, and the exponential as a base raised to a power. But raising a base to a power only means something when the power is a whole number or a fraction. Nothing in the course so far says what it means to raise two to the power of the square root of two, and yet every graph of an exponential you have drawn treats that as settled.
So the order is reversed here. The logarithm comes first, as an integral. From it come the number e, the exponential, and finally a meaning for every real exponent. Read the list of outcomes as a construction sequence rather than a set of topics: each one is only possible because of the one before it.
Warm-up
Discussion prompt
Two cubed is three twos multiplied together. Two to the power one half is a square root. Two to the power three quarters is a fourth root cubed. Now write down, as carefully as you can, what two to the power of the square root of two means. Do not look it up.
Write something down before you reveal the answer, even if what you write is that you are not sure. The point of the question is not to catch you out; it is to make the gap visible, because once you have seen it you cannot unsee it, and the rest of the lesson becomes obviously worth doing.
Three meanings are already on the table. A whole-number power is repeated multiplication. A power with denominator two is a square root, and more generally a power whose denominator is n is an nth root. A power like three quarters combines the two. Every one of these is a recipe, and every recipe needs a fraction.
The square root of two is not a fraction. There is no recipe. If your answer was something like the value the fractional powers are heading towards, you are ahead of the section, because that is the right instinct, but it still needs a definition that says which value, and proves there is one. This lesson supplies it on the very last part.
Section
Part 1
Concept
Figure (svg): The values of two raised to a power, drawn as separate dots at exponents that are quarters, from minus two to two and a half; a dashed vertical line sits at the square root of two, where there is no dot.
Exponents that are whole numbers come from multiplying. Exponents that are fractions come from roots. Together those give a value at every rational exponent, and the dots in the picture are all of them that fit on the page.
The square root of two is not a fraction. There is no dot there, and no rule so far produces one.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 626 — the opening paragraphs
Look carefully at the picture. Those dots are not a sampling of a curve; they are the whole of what has been defined. At every exponent that can be written as a fraction there is a value, and the dots drawn here are just the ones whose denominator is four. Fill in every fraction and the dots become dense, which is exactly why it is so easy to believe there is a curve.
Dense is not the same as complete. Between any two of those dots there are still uncountably many exponents with nothing attached to them, and the square root of two, marked by the dashed line, is one of them. No amount of multiplying and no root produces a value there.
This matters for calculus in particular. To differentiate a function you need it defined on an interval, not on a scattering of points, and a cloud of dots has no derivative anywhere. So every derivative of an exponential you have computed has been resting on an object that was not yet there.
Intuition
Figure (svg): Two diagrams. On the left, two boxes labelled e to the x and ln x with arrows running both ways between them, a loop. On the right, six boxes in a chain: the area under one over t, then ln x, then e, then e to the x, then a to the x, then log base a, with single arrows.
The usual route defines the logarithm as the inverse of the exponential. But the exponential itself is only defined at rational exponents, so its graph is a cloud of dots with gaps, and a cloud of dots has no inverse function.
The way out is to build the logarithm first, from something already complete: an integral. Everything else is then constructed from it, one proved step at a time.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 626 — for purposes of this section, assume we have not yet defined…
Trace the left diagram with your finger and notice that you never get out. To know the logarithm you must first have the exponential, and to have the exponential for all real inputs you need, among other things, a meaning for irrational exponents, which nobody has supplied. Going round the loop faster does not help.
The escape is to find something already complete and start there. The definite integral is complete: Chapter 5 defined it for any continuous function on any closed interval, with no reference whatever to exponentials. That is the foothold.
So read the right-hand chain as the plan for the lesson. Each box is built from the one before it and from earlier chapters only. By the time you reach the last box, an irrational exponent has a meaning, and the loop on the left has been straightened out into a line.
Concept
Every power of t has an antiderivative given by one rule, and that rule breaks in exactly one place.
\[ \int t^{n}\,dt = \frac{t^{n+1}}{n+1} + C, \qquad n \ne -1 \]
At the exponent minus one the formula divides by zero. Yet the function itself is perfectly well behaved for positive t, so an antiderivative must exist. The Fundamental Theorem hands one over without any formula at all.
\[ F(x) = \int_1^{x}\frac{dt}{t} \quad\Longrightarrow\quad F'(x) = \frac1x \]
That function is the missing antiderivative. The whole section consists of discovering that it is the natural logarithm.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 626 — recall the power rule for integrals
You have known about this exception since you first met the power rule for integrals. Add one to the exponent, divide by the new exponent: perfect, except when adding one gives zero. Most courses mention the exception, write down that the answer is a logarithm, and move on. This section refuses to do that, because writing down a logarithm is exactly the thing not yet available.
Instead, ask what the exception is really telling you. The function one over t is continuous on any interval of positive numbers, so Part 1 of the Fundamental Theorem guarantees that the area function starting at one is differentiable and that its derivative is the integrand. An antiderivative therefore exists; it simply is not any algebraic combination of powers.
That is the moment to name a new function. You have a perfectly definite object, defined by an integral, with a known derivative, and no existing formula for it. Naming it is not a trick; it is the same move that gave you the sine and cosine, and the rest of the section is finding out what this new function can do.
Notation
Annotate
On: \( \ln x = \int_{1}^{x}\frac{1}{t}\,dt, \qquad x > 0 \)
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 627 — Definition, equation 6.24
Step through the annotations slowly, because from here on this is what the symbol means. In particular, you may no longer justify anything by saying that the logarithm is the exponent you raise e to. That statement will become a theorem near the end of the lesson, and until then it is not available.
The choice of one as the lower limit is worth a second of thought. It is the only choice that makes the logarithm of one come out as zero, and the algebra proved in the middle of the lesson depends on that. Had you started the integral at two, every property would pick up an additive constant and the function would turn products into sums plus a fixed error.
The last annotation is the one that costs marks on a test. The letter inside the integral is a dummy variable; it is used up by the integration. Writing the same letter for the upper limit and the integrand produces nonsense, and the book uses t inside and x outside precisely to keep them apart.
Picture it
Figure (svg): Two copies of the curve y equals one over t. On the left the region from t equals one to t equals three is shaded and labelled the natural logarithm of x. On the right the region from t equals four tenths to t equals one is shaded in red and labelled minus the natural logarithm of x.
For an input larger than one, sweep right from one and the accumulated area is the logarithm. For an input between zero and one, the limits are in the opposite order, so the integral is the negative of the area from the input up to one.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 627 — Figure 6.75
The two panels are the same definition, not two definitions. On the left the input sits to the right of one, the limits are in the usual order, and the integral is a genuine area, so the logarithm is positive there.
On the right the input sits between zero and one. Now the upper limit is to the left of the lower limit, and by the convention from Section 5.2 that reverses the sign. The shaded region still has a positive area; the integral is minus that area. So the logarithm is negative for inputs below one, which matches everything you already believed about logarithms, and here it is a consequence of the definition rather than a rule to remember.
Notice also what neither panel allows. Pushing the input to zero or below would mean integrating across the place where the curve blows up, and that is why the domain stops at zero.
Worked example
Two facts, each one line, and both proved from Chapter 5 only.
\[ \ln x = \int_1^{x}\frac{dt}{t} \]
Put x equal to 1
Why: The two limits of integration coincide.
\[ \ln 1 = \int_1^{1}\frac{dt}{t} \]
Apply the equal-limits property
Why: An integral over an interval of zero width is zero.
\[ \ln 1 = 0 \]
Differentiate using Part 1 of the Fundamental Theorem
Why: The derivative of an area function is the integrand evaluated at the upper limit.
\[ \frac{d}{dx}\int_1^{x}\frac{dt}{t} = \frac1x \]
State the derivative
Why: This is Theorem 6.15.
\[ \frac{d}{dx}\ln x = \frac1x, \qquad x>0 \]
Read off two consequences
Why: The derivative is positive on the whole domain, so the function is strictly increasing; a differentiable function is continuous, which is Theorem 6.16.
\[ \frac1x > 0 \;\text{for}\; x>0 \;\Longrightarrow\; \ln \text{ is increasing and continuous} \]
Check against what was assumed before
Why: Section 3.9 asserted this derivative and used it for three chapters. Here it is a one-line consequence instead of an assumption, and it agrees: the numerical slope of the area function at x equal to 4 is 0.25, and one over four is 0.25.
\[ F'(4) = \frac14 = 0.25 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 627 — Theorems 6.15 and 6.16
This is a short slide with a large payoff, so read it as a demonstration of method rather than as two facts to memorise. Both results come out in one line each, and neither uses anything later than Chapter 5.
The first result is almost free: an integral whose two limits agree is zero, because the region has no width. The second is the reason the definition was chosen the way it was. Part 1 of the Fundamental Theorem says that differentiating an area function with a variable upper limit simply gives back the integrand evaluated there, so the derivative of this new function is one over the input, with no work at all.
Pause on the final check. Section 3.9 told you that the derivative of the natural logarithm is one over x, and you have used that hundreds of times. It was never proved there; it was quoted, and its proof depended on facts about exponentials that were themselves unproved. Here the same statement falls out in a single line from a definition that assumes nothing. That reversal, from assumption to consequence, is the whole section in miniature.
Concept
Figure (svg): The graph of the natural logarithm from just above zero to five: it plunges below minus four near zero, crosses the horizontal axis at one, and rises slowly to about 1.6 at five.
The graph is a single rising curve with no breaks. Three properties matter later, and each comes from the definition.
| property | where it comes from |
|---|---|
| strictly increasing | the derivative one over x is positive for every positive x |
| continuous | differentiable functions are continuous (Theorem 6.16) |
| grows without bound | adding a fixed area over each doubling of x, forever |
| falls without bound near zero | the same areas counted negatively as x is halved |
Taken together these say the logarithm takes every real value exactly once, which is precisely the hypothesis an inverse function needs.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 628 — Figure 6.76
Everything in the table is read off the definition, and each row will be used later, so it is worth seeing why each is true rather than accepting the picture.
Increasing comes from the derivative: one over x is positive for every positive x, so by Section 4.5 the function rises everywhere on its domain. Continuity is free, because a function that has a derivative has to be continuous. The unbounded behaviour is the one that takes an extra thought: each time you double the input you add the same fixed area, the area from one to two, so doubling repeatedly pushes the value up without any ceiling. Halving repeatedly subtracts the same amount each time and drives it down without a floor.
Put the four rows together and you have a continuous strictly increasing function whose values run over the entire real line. That is precisely the hypothesis needed to guarantee an inverse function that accepts every real number as input, and an exponential that accepts every real exponent is exactly what the lesson is trying to produce.
Fill the middle
Fill in the blanks
\ln x = \int_1}^___\frac______
Why: The lower limit is 1. That choice forces the logarithm of one to be zero, because an integral whose limits agree is zero. Starting anywhere else would add a constant to the whole function and break every algebraic property proved later.
Type your answer before checking. This looks trivial, and it is, but the reason behind it is not, and the reason is the thing to take away.
Starting at one is what makes the logarithm of one equal zero, because an integral over an interval of zero width is zero. If you had started at any other positive number, the whole function would shift up or down by a constant. It would still have derivative one over x, and it would still be increasing, but it would no longer turn products into sums, because the proof of that property uses the value at one in an essential way.
So the lower limit is not a matter of taste. It is the choice that makes the algebra come out right, and you will see exactly where it is used two parts from now.
Prediction
\[ \ln\tfrac12 = \int_1^{1/2}\frac{dt}{t} \]
Predict first
Using only the definition, what is the sign of this number?
Correct: Negative
Why: The upper limit sits to the LEFT of the lower limit, so the integral is minus the integral from one half up to 1. That second integral is a genuine positive area under a positive curve, so the answer is negative. Its value is minus 0.69314. Notice that nothing about exponentials was needed, and that the definition covers inputs below one without any special case.
Commit to an option before revealing. The temptation here is to answer from what you remember about logarithms rather than from the definition in front of you, and the whole point of this part is to practise working from the definition.
Look at the limits. The integral runs from one down to one half, so the upper limit is to the left of the lower limit. By the reversal convention this equals minus the integral from one half up to one, and that second integral is a real positive area under a positive curve. So the answer is negative, as the right-hand panel of the earlier figure showed.
If you chose undefined, you were probably thinking of the restriction that the input must be positive. One half is positive; it is simply less than one. The definition handles inputs on both sides of one without any special case, which is one of its quiet advantages.
Section
Part 2
Concept
Once the derivative is known, every logarithm with something inside it is a chain-rule problem. Write the inside as u and the pattern is always the same.
\[ \frac{d}{dx}\ln u = \frac{1}{u}\cdot\frac{du}{dx} = \frac{u'}{u} \]
Say it out loud as the derivative of the inside over the inside. The two places students lose marks are forgetting the inside derivative altogether, and differentiating the inside twice.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 628 — the chain rule applied to Example 6.35
This slide is short because the rule is short, but it is the workhorse of the next few slides, so say it to yourself in words until it sticks: the derivative of the inside, over the inside.
The reason the pattern is so clean is that the outer derivative is one over the input, so the outer factor contributes nothing but a denominator. Every other function of a function leaves something more complicated behind.
Two errors are worth naming in advance. The first is writing one over the inside and stopping, which forgets the chain rule entirely. The second is differentiating the inside and then also differentiating the copy of it in the denominator, which comes from half remembering the quotient rule. Neither happens if you write down the inside as a separate letter first, which is what the next slide does.
Worked example
Calculate both derivatives.
\[ \text{(a) } \frac{d}{dx}\ln\left(5x^{3}-2\right) \qquad \text{(b) } \frac{d}{dx}\left(\ln(3x)\right)^{2} \]
Name the inside in (a)
Why: The logarithm wraps a cubic.
\[ u = 5x^{3}-2, \qquad u' = 15x^{2} \]
Divide the inside derivative by the inside
Why: The chain rule on the logarithm.
\[ \frac{d}{dx}\ln\left(5x^{3}-2\right) = \frac{15x^{2}}{5x^{3}-2} \]
Name the outside in (b)
Why: Here the logarithm is inside a square, so the outer function is the squaring.
\[ y = u^{2}, \qquad u = \ln(3x) \]
Differentiate the outside, then the inside
Why: Twice u, times the derivative of the logarithm of three x.
\[ \frac{dy}{dx} = 2\ln(3x)\cdot\frac{3}{3x} \]
Simplify the inside factor
Why: The threes cancel, which is why the base of the multiple never survives.
\[ \frac{d}{dx}\left(\ln(3x)\right)^{2} = \frac{2\ln(3x)}{x} \]
Check both at x equal to 2
Why: Numerically the slope of the first expression at 2 is 1.578947, and the formula gives sixty over thirty-eight. The slope of the second at 2 is 1.791759, and the formula gives the natural logarithm of six.
\[ \frac{15(4)}{5(8)-2} = \frac{60}{38} = 1.578947, \qquad \frac{2\ln 6}{2} = \ln 6 = 1.791759 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 628 — Example 6.35
Part (a) is the pattern applied directly. Name the inside, differentiate it, write the result over the inside. There is nothing to simplify afterwards, and there is no point trying.
Part (b) is genuinely different, and it catches people out. The logarithm is not on the outside here; the squaring is. So the first move is the outer derivative, twice the inside, and only then the derivative of the logarithm, which itself needs a chain rule of its own. Two layers, worked from the outside in.
The small surprise in part (b) is that the multiple of three vanishes. Three over three x is one over x, so the answer looks exactly as it would if the logarithm had contained only x. That is not an accident; the logarithm of three times x differs from the logarithm of x by a constant, and constants have no derivative. Seeing that in advance saves a line, and the numerical check at the end confirms both answers independently.
Step zero
\[ \text{(a) } \frac{d}{dx}\ln\left(2x^{2}+x\right) \qquad \text{(b) } \frac{d}{dx}\left(\ln\left(x^{3}\right)\right)^{2} \]
Discussion prompt
Before differentiating either one, write down the outside function and the inside function. Then, and only then, apply the chain rule.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 628 — Checkpoint 6.35
Write both answers before revealing anything. The instruction to name the outside and the inside first is not busywork: almost every chain-rule error in this topic comes from starting to differentiate before deciding which function is wrapped around which.
Part (a) is a one-layer problem and should take you about ten seconds. Part (b) has two layers, exactly like part (b) of the worked example: a square on the outside, a logarithm in the middle, and a cube on the inside.
One extra step makes part (b) much tidier. The logarithm of x cubed is three times the logarithm of x, by the property this lesson proves in the next part, so the derivative of that inside function is three over x rather than something involving a cube. If you got six over x times the logarithm of x cubed, you are right, and simplifying the logarithm turns your six into eighteen.
Concept
Figure (svg): The natural logarithm of the absolute value of x, drawn as two mirror-image branches: one for negative x falling steeply near zero, one for positive x, meeting the horizontal axis at minus one and at one.
The logarithm is only defined for positive inputs, which makes it useless as an antiderivative on an interval where the variable is negative. Wrapping it in an absolute value repairs that, and the derivative formula survives intact.
\[ \frac{d}{dx}\ln|x| = \frac1x \quad\text{for every } x \ne 0 \]
On the negative side the function is the logarithm of minus x, whose derivative is minus one over minus x, which is one over x again.
Theorem 6.17 — The natural logarithm of the absolute value is an antiderivative of the reciprocal: the integral of du over u is the logarithm of the absolute value of u, plus a constant.
\[ \int\frac{1}{u}\,du = \ln|u| + C \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 628 — Theorem 6.17
This is the slide that repairs a practical problem. Suppose you need the integral of one over x on an interval sitting entirely to the left of zero. The natural logarithm cannot help, because it refuses negative inputs, yet the integral obviously exists: the function is continuous there.
The fix is to use the logarithm of the absolute value, which is the original curve on the right and its mirror image on the left. Check the derivative on the mirrored branch and you will see why it works: the outside contributes one over minus x and the inside contributes minus one, and the two minus signs cancel to leave one over x again. The single formula therefore covers both sides.
From here on, write the absolute value every time you integrate a reciprocal. It costs two keystrokes and it is the difference between an answer that is right on every interval and one that is right only half the time.
Worked example
Calculate the integral.
\[ \int\frac{x}{x^{2}+4}\,dx \]
Choose the substitution
Why: The numerator is half the derivative of the denominator, which is the signal for a logarithm.
\[ u = x^{2}+4, \qquad du = 2x\,dx \]
Rewrite the integral in u
Why: Half of du replaces x dx.
\[ \int\frac{x}{x^{2}+4}\,dx = \frac12\int\frac{1}{u}\,du \]
Apply Theorem 6.17
Why: The antiderivative of the reciprocal.
\[ \frac12\int\frac{1}{u}\,du = \frac12\ln|u| + C \]
Substitute back
Why: The quantity x squared plus four is never negative, so the absolute value can be dropped here.
\[ \int\frac{x}{x^{2}+4}\,dx = \frac12\ln\left(x^{2}+4\right) + C \]
Check by differentiating the answer
Why: Differentiate half the logarithm of x squared plus four: the chain rule gives two x over twice the quantity, which is the integrand. At x equal to 3 both sides are 0.230769.
\[ \frac{d}{dx}\left[\tfrac12\ln\left(x^{2}+4\right)\right] = \frac{2x}{2\left(x^{2}+4\right)} = \frac{x}{x^{2}+4} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, pp. 628-629 — Example 6.36
The signal to look for is a fraction whose numerator is, up to a constant, the derivative of its denominator. Here the denominator differentiates to two x and the numerator is x, so the numerator is exactly half of what the substitution wants; that half rides out in front and stays there.
After the substitution the integral is the simplest one in the section, and Theorem 6.17 finishes it. The only judgement call is whether to keep the absolute value: here the denominator is a sum of a square and four, which is never negative, so the bars can be dropped. Whenever you are not certain that the inside stays positive, keep them.
The check is the one you should run on every integral for the rest of your life. Differentiate your answer and see whether the integrand comes back. It takes ten seconds, it catches sign errors and missing constants, and here it also confirms the factor of one half.
Prediction
\[ \int\frac{x^{2}}{x^{3}+6}\,dx \]
Predict first
Which of these is the antiderivative?
Correct: (1/3) ln|x³ + 6| + C
Why: Let u be x cubed plus six, so du is three x squared dx. The numerator supplies only x squared dx, which is one third of du, and that one third rides out in front. Differentiating the answer returns three x squared over three times the quantity, which is the integrand; at x equal to 2 both come to 0.285714. The last option is the wrong instinct of dividing an antiderivative of the top by the bottom.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 629 — Checkpoint 6.36
Choose an option before revealing. Every one of the four is a plausible-looking answer, and three of them come from a specific error worth recognising.
The correct route is the same substitution as the worked example. The denominator differentiates to three x squared and the numerator supplies x squared, so the numerator is a third of what is needed and a third comes out in front. Getting three instead of a third means you multiplied where you should have divided, which is the single most common slip in substitution.
The last option deserves a comment. It divides an antiderivative of the numerator by the denominator, as though integration had a quotient rule. It does not, and no rearrangement of the quotient rule for derivatives gives one. If you were drawn to it, differentiate it and watch how little it resembles the integrand.
Trap
The antiderivative of the reciprocal, written the way most people first learn it:
\[ \int\frac{du}{u} = \ln u + C \]
Wrong as soon as u is negative.
\[ \int_{-4}^{-2}\frac{du}{u} = \ln(-2) - \ln(-4)? \]
The natural logarithm is only defined for positive inputs, so that last line is meaningless. With the absolute value the same integral is a real number.
\[ \int\frac{du}{u} = \ln|u| + C \]
\[ \ln 2 - \ln 4 = -\ln 2 \]
The mirrored branch works because the two minus signs cancel when it is differentiated.
\[ \frac{d}{du}\ln(-u) = \frac{-1}{-u} = \frac1u \]
Most textbooks before this one write the antiderivative of one over u without bars, and for positive u nothing goes wrong. The damage shows up the first time an integral sits on the negative side of zero, and then it is not a small error: the expression on the page is undefined, so there is nothing to evaluate.
The repaired version costs nothing and works everywhere except at zero itself, which no interval of integration may contain anyway. Practise reading the bars as a reminder that the logarithm has a domain, rather than as decoration.
One thing the bars do not do is let you integrate across zero. The function one over u has an infinite discontinuity there, so an integral whose interval contains zero is improper and needs the machinery of Section 3.7, not an absolute value.
Section
Part 3
Intuition
Nothing yet says this integral behaves like a logarithm. The algebra has to be proved, and the proof rests on one feature of the reciprocal that no other power has.
Stretch an interval by a factor and it gets that many times wider. Under the reciprocal curve it also sits that many times lower. The two effects cancel exactly, so the enclosed area is unchanged.
\[ \int_{a}^{ab}\frac{dt}{t} = \int_{1}^{b}\frac{du}{u} \]
Read in words: the cost of multiplying by a factor does not depend on where you start. That is what turns multiplication on the inside into addition on the outside.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 629 — the proof of Theorem 6.18(ii)
Everything proved so far has been about calculus. Nothing yet says this function behaves like a logarithm at all, and the name has been used on trust. This part earns it.
The key observation is easier to feel than to state. Multiplying an interval's endpoints by a fixed factor makes the interval that many times longer. Underneath the reciprocal curve, it also sits at that many times lower a height, because the reciprocal of a scaled input is the scaled reciprocal. Longer and lower by the same factor means the same area.
Say it in words and it sounds like the logarithm law itself: the cost of multiplying by a given factor is the same wherever you start from. Adding up costs along the way turns multiplication of inputs into addition of outputs. The next two slides turn that sentence into a picture and then into a proof.
Picture it
Figure (svg): The curve y equals one over t with two shaded strips: a narrow tall one from t equals one to two, and a wide short one from t equals three to six. A note says both areas equal 0.6931.
The narrow strip runs from one to two, and the wide strip from three to six. The second is three times wider and its curve is three times lower, so the two shaded regions have exactly the same area, and both equal the natural logarithm of two.
Look at the two shaded regions and resist the feeling that the wide one must be bigger. The strip on the left has width one and heights running from one down to one half. The strip on the right has width three, three times as wide, but its heights run from one third down to one sixth, exactly a third of the first strip's heights at the matching places.
Three times as wide and a third as tall gives the same area, and this is the one curve for which that trade is exact. Both regions have area 0.69314, the natural logarithm of two.
Keep this picture in mind while you read the proof on the next slide. The substitution there is nothing but this stretching, written out algebraically.
Worked example
Prove that the logarithm turns a product into a sum, for any two positive numbers.
\[ \ln(ab) = \int_{1}^{ab}\frac{dt}{t} \]
Split the interval at a
Why: Additivity of the definite integral, from Section 5.2.
\[ \int_{1}^{ab}\frac{dt}{t} = \int_{1}^{a}\frac{dt}{t} + \int_{a}^{ab}\frac{dt}{t} \]
Substitute in the second integral only
Why: Set t equal to a times u, so that the interval from a to ab becomes the interval from 1 to b.
\[ t = au, \qquad dt = a\,du \]
Rewrite the integrand
Why: The factor a appears in the numerator and the denominator and cancels.
\[ \frac{dt}{t} = \frac{a\,du}{au} = \frac{du}{u} \]
Change the limits
Why: When t equals a, u is 1; when t equals ab, u is b.
\[ \int_{a}^{ab}\frac{dt}{t} = \int_{1}^{b}\frac{du}{u} = \ln b \]
Assemble
Why: The first piece was already the logarithm of a.
\[ \ln(ab) = \ln a + \ln b \]
Check with numbers
Why: Take a equal to 2 and b equal to 3. The two sides come to the same value, and the middle strip is the one drawn on the previous slide.
\[ \ln 6 = 1.791759, \qquad \ln 2 + \ln 3 = 0.693147 + 1.098612 = 1.791759 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 629 — Theorem 6.18(ii) and its proof
There are only two ideas in this proof. Split the interval at an intermediate point, which Section 5.2 allows for any definite integral, and then stretch the second piece back to standard position.
Watch the stretching step closely. Setting t to a times u makes the differential pick up a factor of a, and the denominator picks up the same factor of a, and they cancel. That cancellation is the whole theorem. It happens because the denominator carries the variable to the power one and the differential carries it to the power one as well, and it happens for no other power, which is what the next slide makes concrete.
Do not skip the change of limits. When t is a, the new variable u is one; when t is a times b, u is b. So the second piece becomes an integral from one to b, and an integral from one is by definition a logarithm. The proof ends because the last expression matches the definition, not because of any algebraic manipulation.
Counterexample
\[ G(x) = \int_1^{x}\frac{dt}{t^{2}} \]
Discussion prompt
The proof above worked because the factor a cancelled. Run the same idea on the next power down: evaluate this new function, then compare G of six with G of two plus G of three, and say in one sentence what the comparison proves.
Do this one with a pen before revealing. It is short, it is entirely mechanical, and the conclusion is one of the more surprising things in the section.
The new area function is easy to evaluate, because the power rule works for the exponent minus two. Then simply test whether it turns a product into a sum, using the smallest interesting numbers, two and three. It does not, and it misses by a clear margin.
Now ask why. Repeat the substitution from the proof on this integrand and the differential contributes one factor of the scale while the squared denominator contributes two, so a leftover factor survives and the second piece is not an area function of the same shape. The conclusion is worth stating out loud: the logarithm's algebra is not a general property of accumulated area. It belongs to the reciprocal alone, which is also the one power the ordinary integration rule cannot handle. Those two facts are the same fact.
Worked example
Prove that a rational exponent inside the logarithm comes out as a multiplier. The tool is two functions with the same derivative.
\[ \ln\left(x^{r}\right) = r\ln x, \qquad r \text{ rational} \]
Differentiate the left-hand side
Why: Chain rule, using the power rule for a rational exponent from Section 3.6.
\[ \frac{d}{dx}\ln\left(x^{r}\right) = \frac{rx^{r-1}}{x^{r}} = \frac{r}{x} \]
Differentiate the right-hand side
Why: A constant times the logarithm.
\[ \frac{d}{dx}\left(r\ln x\right) = \frac{r}{x} \]
Compare
Why: Two functions with the same derivative on an interval differ by a constant, by the corollary to the Mean Value Theorem.
\[ \ln\left(x^{r}\right) = r\ln x + C \]
Evaluate the constant at x equal to 1
Why: Both logarithms of one are zero.
\[ \ln\left(1^{r}\right) = r\ln 1 + C \;\Longrightarrow\; 0 = 0 + C \]
Conclude
Why: The constant is zero, so the two sides agree everywhere on the domain.
\[ \ln\left(x^{r}\right) = r\ln x \]
Check the boundary of the claim
Why: Take x equal to 8 and r equal to one third: the logarithm of two is 0.693147, and a third of the logarithm of eight is 2.079442 divided by three, which is the same. Note that the very first line used the power rule for a rational exponent, so nothing here yet covers an irrational r.
\[ \ln 8^{1/3} = \ln 2 = 0.693147 = \tfrac13(2.079442) \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, pp. 629-630 — Theorem 6.18(iv) and its proof
This proof uses a different tool from the last one, and the tool is worth remembering on its own. If two functions have the same derivative on an interval, they differ by a constant, and you pin the constant down by evaluating both at a convenient point.
Both sides are easy to differentiate. The left side is a chain rule with the power rule inside; the right side is a constant times the logarithm. Both give the constant over x, so the difference of the two sides is constant. Setting the input to one makes both logarithms vanish, so the constant is zero and the two sides agree everywhere.
Now notice the restriction, because it matters enormously later. The very first line differentiated x raised to the power r using the power rule, and the power rule has only been proved for rational exponents. So this theorem, as proved, says nothing about an irrational exponent. The error-analysis slide a little further on is built entirely around forgetting that, and the last part of the lesson is what finally removes the restriction.
Worked example
Use the properties just proved to write this as a single logarithm.
\[ \ln 9 - 2\ln 3 + \ln\left(\frac13\right) \]
Rewrite each argument as a power of three
Why: Nine is three squared and a third is three to the minus one.
\[ \ln\left(3^{2}\right) - 2\ln 3 + \ln\left(3^{-1}\right) \]
Bring every exponent out in front
Why: Theorem 6.18(iv), with the rational exponents two and minus one.
\[ 2\ln 3 - 2\ln 3 - \ln 3 \]
Collect
Why: The first two terms cancel.
\[ \ln 9 - 2\ln 3 + \ln\left(\frac13\right) = -\ln 3 \]
Check numerically
Why: The logarithm of nine is 2.197225, twice the logarithm of three is 2.197225, and the logarithm of a third is minus 1.098612; the total is minus 1.098612, which is minus the logarithm of three.
\[ 2.197225 - 2.197225 - 1.098612 = -1.098612 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 630 — Example 6.37
The strategy for any expression like this is to get every argument onto a single base first. Here everything is a power of three, including the fraction, which is three to the power minus one.
Once that is done, the exponent property empties each term out into a multiple of one logarithm, and what is left is ordinary arithmetic with a common factor. The cancellation is dramatic here, which is the point of the exercise: two terms disappear completely and a single negative logarithm survives.
Do the numerical check even though the algebra looked safe. Sign errors on the term with the fraction are common, because it is easy to read the logarithm of one third as positive. The numbers settle it immediately.
Fill the middle
Fill in the blanks
\ln 8 - \ln 2 - \ln\left(\frac14\right) = \ln 16
Why: The first two terms give the logarithm of four by the quotient property. The last term is minus the logarithm of one quarter, which is plus the logarithm of four. Two logarithms of four add to the logarithm of sixteen, and numerically 2.079442 minus 0.693147 plus 1.386294 is 2.772589, which is the natural logarithm of sixteen.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 630 — Checkpoint 6.37
Work this out fully before typing an answer, and be careful with the two minus signs, which is where nearly every wrong answer comes from.
The first two terms combine by the quotient property into the logarithm of four. The third term is subtracted, and what is being subtracted is the logarithm of one quarter, which is itself negative; subtracting a negative adds. So the third term contributes plus the logarithm of four, and the two together give the logarithm of sixteen.
If you answered two, you probably divided by four instead of multiplying. If you answered one, you treated the last minus sign as though it belonged to the argument rather than to the term. Both are worth catching now, because the same two slips reappear whenever logarithms are combined under exam pressure.
Error analysis
Annotate
On: \( \ln\left(2^{\sqrt2}\right) = \sqrt2\,\ln 2 \;\Longrightarrow\; 2^{\sqrt2} = e^{\sqrt2\,\ln 2} \)
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 630 — note that we can extend this property to irrational values later
Read the argument on its own first. Nothing in it is computationally wrong, and if you met it in a solution you would very likely accept it. That is what makes it worth studying.
The flaw is one of order. The property that brings an exponent out in front was proved only for rational exponents, and the exponent here is the square root of two. Worse, the left-hand side of the argument writes down a symbol that has no meaning yet; the section has not said what two to an irrational power is. You cannot take the logarithm of an object that does not exist.
The repair is elegant and it is what the book actually does. Rather than deriving the right-hand side from the left, declare the right-hand side to be the definition of the left. Then the property is true by construction, and because it already held at rational exponents, the new definition agrees with the old meaning wherever the old meaning existed. Extending a definition so that a desired theorem becomes true is a standard move in mathematics, and this is a clean example of it.
Sorting
Sort into buckets
Each statement is used somewhere in this section. Which are built here, and which are imported from earlier chapters?
Sort all seven before revealing. This is a check on the architecture of the section rather than on any calculation, and if the sorting comes easily then you have understood what the rebuild is for.
The three borrowed items share a feature: none of them mentions a logarithm or an exponential. The power rule for integrals, the Fundamental Theorem and the Intermediate Value Theorem are results about integrals, derivatives and continuous functions, established long before this chapter, so using them here creates no circularity at all.
The four built items are in dependency order. The derivative has to come first, because the proof that the function is increasing, and therefore one-to-one, needs it. The algebra comes next, because the existence of e uses the doubling property. Only then can the exponential appear, and only then can an irrational power be given a meaning. Rearranging any two of them would reintroduce exactly the loop the section exists to remove.
Section
Part 4
Concept
Figure (svg): The curve y equals one over t with the region from t equals one to t equals 2.718 shaded in yellow and labelled area equals one exactly; a dashed line marks the right-hand edge at e.
the number e — The real number whose natural logarithm equals one; equivalently, the point at which the accumulated area under the reciprocal, measured from one, first reaches one.
\[ \ln e = 1 \]
Notice what this definition does not do. It does not write e as a decimal, and it does not write it as a limit. It names e by a property, and the next slide shows that exactly one number has that property.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 630 — Definition of the number e
This is a definition by property, which may be an unfamiliar style. Instead of saying what e equals, it says what e does: it is the input at which this particular accumulated area reaches one.
The picture makes that concrete. Start at one, sweep right under the reciprocal curve, and keep a running total of the area. That total starts at zero, rises, and passes through every positive value exactly once. The place where it passes through one is e.
Nothing in that description mentions exponentials, compound interest or a decimal expansion, and that is deliberate. The whole lesson depends on e being available before the exponential exists, so e must be defined without one. The next slide checks that this description really does pick out exactly one number.
Estimation
\[ \int_1^{b}\frac{dt}{t} = 1 \]
Predict first
Sweep right from t equal to 1 under the curve one over t. How far must you go before the area first reaches one?
Correct: About 2.7
Why: Most people guess low, because the curve starts at height one and the first unit of width alone contributes 0.693. After that the curve has dropped below one half and progress slows sharply. At b equal to 2.2 the area is only 0.788, and at 2.7 it is 0.993, just short. At 3.6 you have overshot to 1.281. The exact crossing is at 2.718281828, the number this part is about.
Guess before you reveal, and try to guess from the shape of the curve rather than from what you remember about e. Most people guess too small.
Here is why. The very first unit of width, from one to two, is under the tallest part of the curve, and it contributes 0.693 all by itself. That feels like most of the job. But by the time you reach two, the curve has fallen to a height of one half and it keeps falling, so the remaining 0.307 of area needs a lot more horizontal distance than you would expect.
The numbers make the point. At 2.2 you have 0.788, at 2.7 you have 0.993, and you are still short. The crossing is at 2.718281828, and recognising it as the familiar number is the small reward for having built the definition honestly.
Worked example
The definition names a number by a property, so the property has to be shown to pick out exactly one number.
Bound the logarithm of two from above
Why: On the interval from 1 to 2 the reciprocal is below one, and the interval has width one.
\[ \ln 2 = \int_1^{2}\frac{dt}{t} < 1\cdot 1 = 1 \]
Bound the logarithm of two from below
Why: On the same interval the reciprocal stays above one half.
\[ \ln 2 > \tfrac12\cdot 1 = \tfrac12 \]
Double it using the product property
Why: The logarithm of four is twice the logarithm of two, which is therefore above one.
\[ \ln 4 = 2\ln 2 > 1 \]
Apply the Intermediate Value Theorem
Why: The logarithm is continuous on the interval from 2 to 4, below one at the left end and above one at the right end.
\[ \exists\; c \in (2,4): \;\ln c = 1 \]
Rule out a second such number
Why: The logarithm is strictly increasing, so it takes the value one at most once.
\[ x_1 < x_2 \;\Longrightarrow\; \ln x_1 < \ln x_2 \]
Check the bracket numerically
Why: The logarithm of 2.7 is 0.993252, just under one, and the logarithm of 2.72 is 1.000632, just over; so e lies between 2.7 and 2.72, and to ten places it is 2.7182818285.
\[ \ln 2.7 = 0.993252 < 1 < 1.000632 = \ln 2.72 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 630 — the proof that such a number exists and is unique is left to you
A definition by property is only legitimate if exactly one object has the property, so this slide is not optional bookkeeping; it is what makes the previous slide mean anything.
Existence comes from the Intermediate Value Theorem, and the only work is producing one input where the logarithm is below one and another where it is above. Below is easy: on the interval from one to two the curve never reaches height one, so the area there is less than the area of a unit square. Above takes one extra idea, and the idea is to reuse the product property you have just proved. The logarithm of two is at least one half, by a rectangle sitting under the curve, so the logarithm of four is at least one, by doubling.
Uniqueness is the shorter half. A strictly increasing function cannot take the same value twice, so at most one input has logarithm one. Existence plus uniqueness gives exactly one, and the number is named e. The numerical bracket at the end is not part of the proof, but it is a reassuring sanity check that the number the argument produced is the one you expected.
Intuition
Chapter 1 met e as the number that compound interest approaches when the compounding gets finer and finer, and Section 4.8 met it again in a limit. Neither of those is the definition used here, yet they agree.
\[ \lim_{n\to\infty}\left(1+\frac1n\right)^{n} = e \]
The bridge is the derivative. Take the logarithm of the expression inside the limit, use the property proved a few slides ago to bring the exponent out, and what remains is a difference quotient for the derivative of the logarithm at one.
\[ \ln\left(1+\tfrac1n\right)^{n} = \frac{\ln(1+h) - \ln 1}{h} \;\longrightarrow\; \frac{d}{dx}\ln x\Big|_{x=1} = 1 \]
So the logarithm of the limit is one, and the number whose logarithm is one is the e defined here. The old story and the new one describe the same point on the line.
Stewart, Calculus: Early Transcendentals 8e, Appendix G — The Logarithm Defined as an Integral Appendix G, pp. A50-A56 — the same construction, with the limit shown explicitly
It would be strange if this section produced a different number and called it by the same name, so the agreement is worth seeing rather than assuming.
Follow the bridge carefully. Take the logarithm of the compound-interest expression and the exponent comes out in front, leaving n times the logarithm of one plus one over n. Write one over n as a small increment h, and that product is precisely the difference quotient for the logarithm at the input one, because the logarithm of one is zero. As n grows, h shrinks, and the difference quotient converges to the derivative, which is one over one, that is, one.
So the logarithm of the limit is one, and the number whose logarithm is one is the e of this section. Two very different-looking descriptions, the same point on the number line. Notice also which direction the argument runs: the calculus fact about the derivative is what forces the two to agree, and the calculus came from the integral definition.
Trap
A common answer to what e is:
\[ e = 2.71828 \]
Wrong, and not only because it is rounded.
\[ \ln(2.71828) = 0.99999932 \]
No finite decimal can be e, because e is irrational. A decimal is an approximation produced from the definition, never a substitute for it.
The definition is a property, and that is what proofs use.
\[ \ln e = 1 \]
This matters in practice too: the derivative rule about to be derived holds exactly for e and only approximately for 2.71828.
Nobody sets out to define a number by its rounded decimal, and yet a great many written answers do exactly that, particularly when a question asks what e is. The honest answer is one line and it is a property, not a value.
The reason this matters is not pedantry. The derivative rule you are about to derive, that the exponential is its own derivative, is exact for e and false for any decimal approximation. If you replace e by 2.71828 in a derivative you introduce an error in the fourth decimal place, and in a growth model run over many steps that error compounds.
There is also a logical point. Since e is irrational, no finite decimal equals it, so a decimal cannot be a definition of it at all. The digits are computed from the definition afterwards, to whatever precision you happen to need.
Matching
Match the pairs
Why: Five facts, five different tools, and not one of them mentions an exponential. That is the check that the construction is not circular: every result so far rests on Chapters 3 to 5 alone.
Work through all five before checking. The exercise is a summary of the lesson so far, and the interesting column is the right-hand one.
Scan that column and notice what is absent. There is no exponential anywhere in it. Five substantial facts about logarithms have been established using only the additivity of integrals, the Fundamental Theorem, a substitution, the Intermediate Value Theorem and the sign of a derivative.
That absence is the proof that the construction is not circular, and it is the reason the section is allowed to build the exponential next. If any of these five had needed an exponential, the whole thing would collapse back into the loop from Part 1.
Section
Part 5
Picture it
Figure (svg): The curves of the exponential function and the natural logarithm on the same equally scaled axes, with the dashed line y equals x between them; each curve is the mirror image of the other in that line.
The logarithm is strictly increasing and takes every real value, so it is one-to-one with all the real numbers as its range. That entitles it to an inverse, defined for every real input, and the inverse's graph is the mirror image in the line where the two coordinates agree.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 631 — Figure 6.78
Everything needed for this slide was established earlier, so read it as a payoff rather than as new content. The logarithm is strictly increasing, which makes it one-to-one, and its values run over the whole real line, which makes that entire line the domain of the inverse.
Those two facts are exactly the hypotheses an inverse function needs, and they were proved, not assumed. So the inverse exists as a matter of theorem, and it accepts every real number as an input, including irrational ones. That last clause is the one that will finally close the gap from the beginning of the lesson.
The graph is the usual reflection in the line where the two coordinates agree. Check the two marked points against each other: the logarithm passes through the input one with output zero, so the inverse passes through the input zero with output one. Every feature of one curve translates into a feature of the other.
Notation
Annotate
On: \( \exp(\ln x) = x \;(x>0), \qquad \ln(\exp x) = x \;(\text{all } x) \)
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 631 — the paragraph introducing exp
The temporary name exp is deliberate and you should honour it for the next few slides. Calling the inverse e to the x right away would assume that it behaves like a power, and proving that is precisely the work of this part.
Pay attention to the two different ranges of validity. The first law can only be tested at positive inputs, because those are the only inputs the logarithm accepts. The second law holds for every real input without exception, because the logarithm's outputs already cover the whole line.
Both laws are used constantly in what follows, usually in the direction that turns an awkward expression into a plain variable. When a derivation seems to conjure an x out of nowhere, look for one of these two.
Concept
The inverse has been called exp so far. To justify the familiar notation, check it against the old meaning wherever the old meaning existed, which is at rational exponents.
\[ \ln\left(e^{x}\right) = x\ln e = x\cdot 1 = x, \qquad x \text{ rational} \]
The left side is the logarithm of a rational power, so the property proved in Part 3 applies. The result says that the old e to the x and the new exp of x have the same logarithm, and the logarithm is one-to-one, so they are the same number.
For an irrational exponent there was no old meaning, so nothing needs checking: the definition simply supplies one.
Definition (6.25) — For any real number x, e to the x is the number whose natural logarithm is x. Equivalently, e to the x is exp of x for every real x.
\[ e^{\ln x} = x \;(x>0), \qquad \ln\left(e^{x}\right) = x \;(\text{all } x) \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 631 — Definition and equations 6.25, 6.26
There are two halves to justifying the familiar notation, and they are quite different in character.
For a rational exponent there is an old meaning, so agreement has to be checked. Take the logarithm of e raised to a rational power, bring the exponent out using the property from Part 3, and use the defining property of e to reduce the logarithm of e to one. What is left is the exponent itself, which is also the logarithm of exp of that exponent. Two positive numbers with the same logarithm are equal, because the logarithm is one-to-one, so the old power and the new inverse are the same number.
For an irrational exponent there was never an old meaning to conflict with, so nothing needs checking. The definition simply supplies a value where none existed, and it does so consistently, because it agrees with the old meaning everywhere the old meaning made sense. That is what a good extension of a definition looks like.
Worked example
Figure (svg): The curve of the exponential function with two dashed tangent lines, one at height one and one at height about four; at each point the steepness of the tangent matches the height of the curve.
Derive the differentiation formula by implicit differentiation, using only the inverse relation.
Name the function and take logarithms
Why: The second cancellation law.
\[ y = e^{x} \;\Longrightarrow\; \ln y = x \]
Differentiate both sides with respect to x
Why: The left side needs the chain rule, because y depends on x.
\[ \frac{1}{y}\cdot\frac{dy}{dx} = 1 \]
Solve for the derivative
Why: Multiply through by y.
\[ \frac{dy}{dx} = y \]
Put y back
Why: The function equals its own derivative.
\[ \frac{d}{dx}e^{x} = e^{x} \]
Read off the integral
Why: Every derivative formula is an antiderivative formula read backwards.
\[ \int e^{x}\,dx = e^{x} + C \]
Check against the picture
Why: At x equal to 1.4 the height is 4.0552, and the tangent drawn there has slope 4.0552; at zero both are 1. A curve whose slope always equals its height is what the formula asserts.
\[ e^{1.4} = 4.0552 = \frac{d}{dx}e^{x}\Big|_{x=1.4} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 632 — the implicit differentiation and the integration formula
This derivation is three lines long and you should be able to reproduce it from memory, because it is the cleanest example of implicit differentiation in the course.
The trick is to take logarithms first. That converts an object you do not yet know how to differentiate into an equation involving one you do. Differentiating both sides then requires the chain rule on the left, since the variable inside the logarithm is itself a function of x, and that is what produces the reciprocal factor which you then clear.
The picture is the result made visible. At a height of one the tangent has slope one; at a height of 4.0552 the tangent has slope 4.0552. No matter where you stand on this curve, the steepness and the height are the same number, and that property characterises the exponential up to a constant multiple. The integration formula is the same statement read backwards, and it is the reason this function appears in every differential equation you will meet.
Worked example
The exponential was built as an inverse, not as a power, so the laws of exponents have to be proved rather than assumed.
\[ e^{p}e^{q} = e^{p+q}, \qquad p, q \text{ any real numbers} \]
Take the logarithm of the left-hand side
Why: Turn the product into a sum with Theorem 6.18(ii).
\[ \ln\left(e^{p}e^{q}\right) = \ln\left(e^{p}\right) + \ln\left(e^{q}\right) \]
Cancel each logarithm against its exponential
Why: The second cancellation law, valid for every real exponent.
\[ \ln\left(e^{p}\right) + \ln\left(e^{q}\right) = p + q \]
Recognise the right-hand side
Why: By the same cancellation law read backwards.
\[ p + q = \ln\left(e^{p+q}\right) \]
Use one-to-one-ness
Why: Two positive numbers with the same logarithm are equal, because the logarithm is strictly increasing.
\[ \ln\left(e^{p}e^{q}\right) = \ln\left(e^{p+q}\right) \;\Longrightarrow\; e^{p}e^{q} = e^{p+q} \]
Check with irrational exponents
Why: Take p equal to the square root of two and q equal to one third of the square root of two. Multiply the two values and compare with the single exponential of their sum: 4.11325 times 1.602243 is 6.590427, and so is the exponential of 1.885618.
\[ e^{1.414214}e^{0.471405} = 6.590427 = e^{1.885618} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 632 — Theorem 6.19 and its proof
It is easy to feel that this theorem needs no proof, since the laws of exponents have been familiar since school. But the object on the page was defined as an inverse function, not as a repeated product, so nothing about products is automatic. Everything has to be routed through the logarithm.
The pattern of the proof is one you will reuse. To show two positive quantities are equal, show their logarithms are equal, then invoke the fact that the logarithm is one-to-one. Inside, the only tools are the product property from Part 3 and the cancellation law.
The numerical check uses two irrational exponents deliberately. For rational exponents the law was already known, so a check there would prove nothing about the new content. Two irrational exponents is the case the theorem actually adds, and the arithmetic confirms it.
Worked example
Evaluate both derivatives.
\[ \text{(a) } \frac{d}{dt}e^{3t}e^{t^{2}} \qquad \text{(b) } \frac{d}{dx}e^{3x^{2}} \]
Combine the two factors in (a)
Why: The law of exponents just proved turns a product into one exponential.
\[ e^{3t}e^{t^{2}} = e^{3t+t^{2}} \]
Differentiate with the chain rule
Why: The exponential is its own derivative, times the derivative of the exponent.
\[ \frac{d}{dt}e^{3t+t^{2}} = e^{3t+t^{2}}(3+2t) \]
Differentiate (b) the same way
Why: Here the exponent is three x squared, whose derivative is six x.
\[ \frac{d}{dx}e^{3x^{2}} = e^{3x^{2}}\cdot 6x \]
Check both at the value one
Why: The numerical slope of the first at t equal to 1 is 272.9908, and five times e to the fourth is 272.9908. The numerical slope of the second at x equal to 1 is 120.5132, and six times e cubed is 120.5132.
\[ e^{4}(3+2) = 272.9908, \qquad 6e^{3} = 120.5132 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 632 — Example 6.38
Part (a) rewards a moment's thought before differentiating. You could use the product rule on two exponentials, and you would get the right answer eventually, but combining them first with the law of exponents just proved turns it into a single chain rule and halves the work.
After that, both parts are the same problem: the exponential reproduces itself, and the chain rule supplies the derivative of the exponent as a factor. There is no simplification to chase afterwards, so resist the urge to expand anything.
The numerical checks at the end are worth running even on problems this routine, because the commonest error here is not conceptual but clerical: differentiating the exponent wrongly, or forgetting the factor entirely. A single substitution catches both.
Prediction
\[ \frac{d}{dx}\left(\frac{e^{x^{2}}}{e^{5x}}\right) \]
Predict first
Which is the derivative?
Correct: e^(x² − 5x) · (2x − 5)
Why: Subtract the exponents first: the quotient is a single exponential with exponent x squared minus five x. Then the chain rule multiplies by the derivative of that exponent, which is two x minus five. At x equal to 1 the value is minus 0.054947, matching a numerical slope. The last two options come from differentiating the top and bottom separately, which no rule permits.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 632 — Checkpoint 6.38
Pick an option before revealing, and do the simplification first rather than reaching for the quotient rule.
A quotient of two exponentials with the same base is a single exponential whose exponent is the difference, so the whole expression collapses to one exponential before any differentiation happens. Then it is one chain rule, and the factor is the derivative of x squared minus five x.
The two options on the right of the list both come from differentiating the numerator and the denominator separately, which no rule permits. If either looked plausible, that is a sign to write the quotient rule out in full next time rather than improvising it.
Worked example
Evaluate the integral.
\[ \int 2xe^{-x^{2}}\,dx \]
Choose the substitution
Why: The exponent is the natural candidate, and its derivative is already present up to a sign.
\[ u = -x^{2}, \qquad du = -2x\,dx \]
Rewrite the integral
Why: Two x dx is minus du, so a minus sign comes out in front.
\[ \int 2xe^{-x^{2}}\,dx = -\int e^{u}\,du \]
Integrate
Why: The exponential is its own antiderivative.
\[ -\int e^{u}\,du = -e^{u} + C \]
Substitute back
Why: Restore the exponent.
\[ \int 2xe^{-x^{2}}\,dx = -e^{-x^{2}} + C \]
Check by differentiating
Why: The chain rule on the answer returns minus e to the minus x squared times minus two x, which is the integrand; at x equal to 1 both are 0.735759.
\[ \frac{d}{dx}\left(-e^{-x^{2}}\right) = -e^{-x^{2}}(-2x) = 2xe^{-x^{2}} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 633 — Example 6.39
The shape to recognise is an exponential whose exponent differentiates into something already sitting outside it, up to a constant. That is the signal for substitution, and it is by far the most common way an exponential integral is set up in a course or on an exam.
Watch the sign. The exponent differentiates to minus two x, while the integrand offers plus two x, so the substitution introduces a minus sign that stays to the end. Losing it is the single most frequent error on this problem.
Note also that this integral has no elementary antiderivative if the factor of x is removed. The exponential of minus x squared on its own is the bell curve, and it cannot be integrated in closed form at all. The factor of x is exactly what makes the substitution work, so recognising it is not a nicety.
Worked example
Evaluate the integral.
\[ \int\frac{4}{e^{3x}}\,dx \]
Move the exponential to the numerator
Why: A reciprocal exponential is the exponential of the negative.
\[ \frac{4}{e^{3x}} = 4e^{-3x} \]
Substitute
Why: The exponent is minus three x.
\[ u = -3x, \qquad du = -3\,dx \]
Rewrite and integrate
Why: dx is minus a third of du.
\[ 4\int e^{-3x}\,dx = -\frac43\int e^{u}\,du = -\frac43 e^{u} + C \]
Substitute back
Why: Restore the exponent.
\[ \int\frac{4}{e^{3x}}\,dx = -\frac43 e^{-3x} + C \]
Check by differentiating
Why: Differentiating brings down a factor of minus three, which cancels the minus one third and leaves four e to the minus three x; at x equal to 1 both sides are 0.199148.
\[ \frac{d}{dx}\left(-\tfrac43 e^{-3x}\right) = 4e^{-3x} = \frac{4}{e^{3x}} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 633 — Checkpoint 6.39
The first move is the whole problem. An exponential in a denominator is an exponential with the sign of its exponent flipped, and once it is upstairs the integral is a routine substitution.
If you left it downstairs you may have been tempted to treat it as a quotient and look for a rule that does not exist. There is no quotient rule for integrals, and no partial fractions to attempt here either.
The factor of minus one third comes from the substitution, not from anywhere else, so check it by differentiating your answer: the chain rule should produce a minus three that cancels it and leaves exactly the integrand.
Section
Part 6
Concept
The exponential is now defined for every real input, and the logarithm accepts every positive number. Putting them together supplies the meaning that was missing at the start of the lesson.
Definition of a general power — For any positive base a and any real number x, a to the x is defined to be e raised to the power x times the natural logarithm of a.
\[ a^{x} = e^{x\ln a}, \qquad a>0 \]
Two things to notice. The base has to be positive, because otherwise its logarithm does not exist; that is why negative bases have no general power. And at a rational exponent this agrees with the old meaning, because the property from Part 3 applies there.
\[ a^{r} = e^{r\ln a} \text{ and } \ln\left(a^{r}\right) = r\ln a \quad\text{agree for rational } r \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 633 — Definition of a to the x
This is the slide the whole lesson has been building towards, so slow down. Both ingredients on the right-hand side are now fully defined: the logarithm accepts any positive base, and the exponential accepts any real input. Their composition therefore makes sense for any positive base and any real exponent whatsoever.
Two conditions are worth noticing. The base must be positive, because otherwise its logarithm does not exist, and that is the honest reason why a negative number raised to an arbitrary real power is undefined. It is not a convention; there is genuinely nothing to write down.
The other is consistency. At a rational exponent the new definition and the old repeated-multiplication meaning agree, because the property from Part 3 applies there. So no previously correct calculation becomes wrong. What has happened is that a partial definition has been extended to a complete one, which is precisely what was promised on the first slide.
Worked example
Figure (svg): Four horizontal bars stacked above one another, each shorter than the one below it, all straddling a dashed vertical line at 2.66514. The bars are labelled with exponent ranges 1.4 to 1.5, 1.41 to 1.42, 1.414 to 1.415 and 1.4142 to 1.4143.
The question from the first slide of the lesson, answered.
Apply the definition
Why: The base is two and the exponent is the square root of two.
\[ 2^{\sqrt2} = e^{\sqrt2\,\ln 2} \]
Evaluate the exponent
Why: The square root of two is 1.41421356 and the logarithm of two is 0.69314718.
\[ \sqrt2\,\ln 2 = 1.41421356 \times 0.69314718 = 0.98025814 \]
Exponentiate
Why: One value of the exponential function, at one real input.
\[ 2^{\sqrt2} = e^{0.98025814} = 2.6651441427 \]
Check against the rational brackets in the picture
Why: Rational exponents on either side of the square root of two give values on either side of the answer, and the brackets close on it: with exponents 1.4142 and 1.4143 the values are 2.66511909 and 2.66530383.
\[ 2.66511909 < 2.6651441427 < 2.66530383 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 633 — the definition applied to an irrational exponent
Compare this slide with the very first question of the lesson. The same expression, which then had no meaning at all, now has a value computed in three lines from objects that were each built rather than assumed.
The arithmetic is unremarkable: multiply the exponent by the logarithm of the base, then exponentiate. What is remarkable is that every step is licensed. The logarithm of two exists because the logarithm is defined for all positive inputs; the exponential accepts 0.98025814 because its domain is the whole real line.
The picture is the sanity check. Rational exponents on either side of the square root of two give values on either side of the answer, and the brackets close in on it. Those brackets are what your intuition was reaching for at the start of the lesson, but a sequence of brackets does not by itself name a number. The definition does, and the brackets then confirm it.
Intuition
Figure (svg): Three rising curves through the point zero comma one: two to the x, e to the x and three to the x. A dashed straight line touches each at that point, steepest for base three and shallowest for base two.
Writing a general base as an exponential says something geometric: every exponential curve is the single curve of e to the x, with the horizontal axis rescaled by the logarithm of the base.
\[ a^{x} = e^{(\ln a)x} \]
So there is really only one exponential function. The base only sets how fast the clock runs, which shows up on the graph as the steepness the curve has where it crosses the vertical axis.
The formula on this slide says something worth pausing over: there is essentially only one exponential function, and every base is that same function with the horizontal axis rescaled.
You can see it in the picture. The three curves have the same shape, and they differ only in how quickly they climb, which is governed by the single constant multiplying x in the exponent. Doubling that constant compresses the curve horizontally by half.
The visible fingerprint of that constant is the slope where each curve crosses the vertical axis. Base two leaves at slope 0.693, base three at slope 1.099, and base e at slope exactly one, which is what singles e out of the family. That observation is the content of the theorem on the next slide, and it explains why e is the natural base rather than merely a convenient one.
Worked example
With the definition in hand, one chain rule settles every base at once.
Start from the definition
Why: Rewrite the general power as an exponential.
\[ \frac{d}{dx}a^{x} = \frac{d}{dx}e^{x\ln a} \]
Differentiate with the chain rule
Why: The exponent is a constant times x, so its derivative is that constant.
\[ \frac{d}{dx}e^{x\ln a} = e^{x\ln a}\ln a \]
Translate back
Why: The exponential is the general power again.
\[ \frac{d}{dx}a^{x} = a^{x}\ln a \]
Read the antiderivative backwards
Why: Divide by the constant to undo the extra factor.
\[ \int a^{x}\,dx = \frac{1}{\ln a}a^{x} + C \]
Check the special case and the picture
Why: Putting a equal to e makes the logarithm one and returns the earlier formula. At x equal to zero the derivative of two to the x is 0.693147 and that of three to the x is 1.098612, the two slopes drawn on the previous slide.
\[ a=e: \;\ln e = 1 \;\Longrightarrow\; \frac{d}{dx}e^{x} = e^{x} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 633 — Theorem 6.20
Notice how little work this takes now. The definition converts a general base into base e, the chain rule differentiates it, and the constant that comes down is the logarithm of the base. One derivation covers every base at once.
The integral is the same statement read in reverse. Since differentiating multiplies by the logarithm of the base, integrating must divide by it, and that is the entire content of the formula. If you ever forget which way round it goes, differentiate your candidate answer and see whether the extra constant cancels or doubles.
Test the formula on the case you already know. When the base is e the logarithm of the base is one, the extra factor disappears, and both formulas reduce to the ones from Part 5. A general rule that does not reduce correctly on a known case is a rule you have written down wrongly.
Concept
Figure (svg): Three logarithm curves through the point one comma zero: base two highest, natural logarithm in the middle, base ten flattest. Each is a fixed multiple of the one beside it.
If the base is positive and not one, the general power is one-to-one and so has an inverse, written as the logarithm to that base. Setting the inverse's output as y and taking natural logarithms turns it into something already known.
\[ y = \log_a x \;\iff\; x = a^{y} \;\Longrightarrow\; \ln x = y\ln a \]
\[ \log_a x = \frac{\ln x}{\ln a} \]
So every logarithm is the natural one divided by a constant, which is why all of them cross the horizontal axis at the same place. Differentiating that quotient gives the general rule.
\[ \frac{d}{dx}\log_a x = \frac{1}{\ln a}\cdot\frac{d}{dx}\ln x = \frac{1}{x\ln a} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 634 — the change of base and Theorem 6.21
The derivation is short, so make sure you follow the first line. Setting the logarithm to base a equal to y is the same as saying that a raised to y gives x; taking natural logarithms of that second equation and using the property from Part 3 gives the logarithm of x as y times the logarithm of a, and solving for y is the change-of-base formula.
Read the formula as a statement about graphs, which is what the picture shows. Dividing by a constant scales the whole curve vertically. That is why every logarithm crosses the horizontal axis at the same input, and why base ten is flatter than base two: dividing by the larger constant squashes it more.
The derivative then needs no new ideas at all. A constant divisor comes straight through the differentiation, so the derivative of any logarithm is one over the input times the logarithm of the base. Remember it in that form rather than memorising a separate rule per base.
Tweak it
Parameter explorer
The curve is the logarithm to base a, drawn as the natural logarithm divided by a constant. Slide the base: where does the curve pass through height one, and what happens to the whole picture as the base grows?
\[ \log_a x = \frac{\ln x}{\ln {a}} \]
Move the slider slowly from one end to the other and watch two things: where the curve crosses the horizontal axis, and where it reaches height one.
The crossing never moves. Every logarithm in the family gives zero at the input one, because the natural logarithm does and dividing zero by a constant leaves zero. The height-one crossing does move, and it lands exactly at the input equal to the base, which is a restatement of the defining property of a logarithm.
Notice the overall flattening as you increase the base. Dividing by a larger constant compresses the curve toward the axis, so a base-ten logarithm grows far more slowly than a base-two one. It is the same curve throughout; only the vertical scale changes.
Worked example
Evaluate both derivatives.
\[ \text{(a) } \frac{d}{dt}\left(4^{t}\cdot 2^{t^{2}}\right) \qquad \text{(b) } \frac{d}{dx}\log_{8}\left(7x^{2}+4\right) \]
Put both factors in (a) on the same base
Why: Four is two squared, so four to the t is two to the two t.
\[ 4^{t}\cdot 2^{t^{2}} = 2^{2t}\cdot 2^{t^{2}} = 2^{2t+t^{2}} \]
Differentiate with Theorem 6.20 and the chain rule
Why: The factor is the logarithm of the base, then the derivative of the exponent.
\[ \frac{d}{dt}2^{2t+t^{2}} = 2^{2t+t^{2}}\ln 2\,(2+2t) \]
Write (b) with the general logarithm rule
Why: One over the inside, times the logarithm of the base, times the derivative of the inside.
\[ \frac{d}{dx}\log_{8}u = \frac{1}{u\ln 8}\cdot\frac{du}{dx}, \qquad u = 7x^{2}+4 \]
Substitute the inside and its derivative
Why: The derivative of seven x squared plus four is fourteen x.
\[ \frac{d}{dx}\log_{8}\left(7x^{2}+4\right) = \frac{14x}{\left(7x^{2}+4\right)\ln 8} \]
Check both at the value one
Why: The numerical slope of the first at t equal to 1 is 22.1807, and eight times the logarithm of two times four is 22.1807. The numerical slope of the second at x equal to 1 is 0.612052, and fourteen over eleven times the logarithm of eight is 0.612052.
\[ 2^{3}\ln 2\,(4) = 22.1807, \qquad \frac{14}{11\ln 8} = 0.612052 \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 634 — Example 6.40
Part (a) begins with an algebraic move rather than a calculus one. The two factors have different bases, so combine them onto a common base first; four is two squared, so the whole product becomes a single power of two, and then one chain rule finishes it.
You could instead use the product rule and the general derivative formula on each factor separately, and you would reach the same answer, but with two logarithms floating about and much more scope for error. When bases can be matched, match them.
Part (b) is a direct application of the general logarithm rule. The only thing to watch is that the constant in the denominator is the logarithm of the base, eight, and not the logarithm of anything inside. Both checks at the end use a single input, which is enough to catch a wrong constant or a missing chain-rule factor.
Fill the middle
Fill in the blanks
\frac3___\log____\sqrt___+1} = \frac______+1\right)\ln ___}
Why: Pull the square root out first: the logarithm of a square root is half the logarithm, so the expression is one half of the logarithm to base three of x squared plus one. The general rule contributes one over the inside times the logarithm of the base, the chain rule contributes two x, and the one half cancels the two. The base of the logarithm is three, so that is the constant left in the denominator; at x equal to 1 the value is 0.45512.
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 634 — Checkpoint 6.40
Simplify before you differentiate. The argument of the logarithm is a square root, which is a power of one half, so the property from Part 3 lets you pull one half out in front and leave a much friendlier expression inside.
After that the general rule supplies one over the inside times the logarithm of the base, and the chain rule supplies the derivative of the inside, which is two x. The one half cancels the two, which is why the final numerator is just x.
The blank asks only for the constant left in the denominator, and that constant is always the logarithm of the base of the original logarithm, whatever happens inside. If you wrote e there, you were thinking of the natural logarithm case where the constant is one and invisible.
Worked example
Evaluate the integral.
\[ \int\frac{3}{2^{3x}}\,dx \]
Move the power up to the numerator
Why: A reciprocal power is the power of the negative exponent.
\[ \int\frac{3}{2^{3x}}\,dx = \int 3\cdot 2^{-3x}\,dx \]
Substitute
Why: Take the exponent as the new variable.
\[ u = -3x, \qquad du = -3\,dx \]
Rewrite the integral
Why: Three dx is minus du, so the constant three is absorbed and a minus sign appears.
\[ \int 3\cdot 2^{-3x}\,dx = -\int 2^{u}\,du \]
Integrate with Theorem 6.20
Why: Divide by the logarithm of the base.
\[ -\int 2^{u}\,du = -\frac{1}{\ln 2}2^{u} + C \]
Substitute back
Why: Restore the exponent.
\[ \int\frac{3}{2^{3x}}\,dx = -\frac{1}{\ln 2}2^{-3x} + C \]
Check by differentiating
Why: The derivative brings down the logarithm of two, which cancels the one in the denominator, and the chain rule supplies minus three; at x equal to 1 both sides are 0.375.
\[ \frac{d}{dx}\left(-\frac{2^{-3x}}{\ln 2}\right) = -\frac{2^{-3x}\ln 2\,(-3)}{\ln 2} = 3\cdot 2^{-3x} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, pp. 634-635 — Example 6.41
Two separate simplifications happen before any integration. First the power moves up from the denominator by flipping the sign of its exponent. Then the substitution takes the exponent as the new variable, and the constant three in the numerator is exactly what the substitution consumes.
That last coincidence is not luck; the exercise was built so that it works out. If the numerator had been four instead of three, you would carry a factor of four thirds through to the end, and nothing else would change.
The check is the important habit here. Differentiating the answer brings down a logarithm of two that cancels the one in the denominator and a minus three from the chain rule that cancels the minus sign, leaving exactly the integrand. If any of those cancellations fails, you have found your error without having to re-read the whole solution.
Worked example
Evaluate the integral.
\[ \int x^{2}2^{x^{3}}\,dx \]
Substitute the exponent
Why: Its derivative is three x squared, and x squared is already present.
\[ u = x^{3}, \qquad du = 3x^{2}\,dx, \qquad x^{2}\,dx = \frac{du}{3} \]
Rewrite the integral
Why: One third rides out in front.
\[ \int x^{2}2^{x^{3}}\,dx = \frac13\int 2^{u}\,du \]
Integrate with Theorem 6.20
Why: Divide by the logarithm of the base.
\[ \frac13\int 2^{u}\,du = \frac{2^{u}}{3\ln 2} + C \]
Substitute back
Why: Restore the exponent.
\[ \int x^{2}2^{x^{3}}\,dx = \frac{2^{x^{3}}}{3\ln 2} + C \]
Check by differentiating
Why: The derivative gives two to the x cubed times the logarithm of two times three x squared, all over three times the logarithm of two, which collapses to the integrand; at x equal to 1 both sides are 2.
\[ \frac{d}{dx}\frac{2^{x^{3}}}{3\ln 2} = \frac{2^{x^{3}}\ln 2\,(3x^{2})}{3\ln 2} = x^{2}2^{x^{3}} \]
OpenStax Calculus Volume 1, §6.7 Integrals, Exponential Functions, and Logarithms §6.7, p. 635 — Checkpoint 6.41
This is the cleanest substitution in the section. The exponent is x cubed, its derivative is three x squared, and x squared is sitting right there outside the power, so only a factor of one third is missing and it comes out in front.
Once the substitution is made, the integral is the general exponential formula from Theorem 6.20, nothing more. Divide by the logarithm of the base and you are done.
If you wrote the answer with a three multiplying rather than dividing, differentiate it and watch the factors pile up instead of cancelling. Doing that once is usually enough to fix the habit permanently.
Trap
The variable is in the exponent, but the power rule gets applied anyway:
\[ \frac{d}{dx}2^{x} = x\,2^{x-1} \]
Wrong. The power rule needs a constant exponent.
\[ \text{at } x=0: \; 0\cdot 2^{-1} = 0 \]
Rewrite the base through the definition and the chain rule gives the right answer. The slope where the curve crosses the vertical axis is not zero; the picture in this part shows it is the logarithm of the base.
\[ 2^{x} = e^{x\ln 2} \]
\[ \frac{d}{dx}2^{x} = 2^{x}\ln 2 \]
\[ \text{at } x=0: \; \ln 2 = 0.693147 \]
This error is common because the two shapes look alike on the page: something raised to something. What decides which rule applies is where the variable sits. Variable in the base, constant in the exponent: power rule. Constant in the base, variable in the exponent: this section's rule.
The wrong answer is easy to refute at a single point. At the input zero it predicts a slope of zero, which would mean the curve of two to the x is momentarily flat there. Look at the family picture from earlier in this part and you can see that it is not; it crosses the vertical axis climbing at 0.693.
The cure is to rewrite the base through the definition before differentiating. Once the expression is an exponential of a constant times x, only one rule can possibly apply and the ambiguity disappears.
Comparison
Comparison matrix
| object | built from ln and exp | derivative |
|---|---|---|
| ln x | the integral of 1/t from 1 to x | 1/x |
| e to the x | the inverse of ln | e to the x |
| a to the x | e to the (x ln a) | a to the x, times ln a |
| log base a of x | ln x divided by ln a | 1/(x ln a) |
Fill in all three blanks before checking. The table is the whole lesson compressed into four rows, and being able to reconstruct it is a fair test of whether the construction has stuck.
Read the middle column downwards as a dependency chain. Each object is built from the ones above it, and the first row is built from an integral alone. Nothing in the column refers to anything that has not already appeared above it.
The derivative column has a pattern worth noticing. Differentiating an exponential multiplies by the logarithm of the base; differentiating a logarithm divides by it. The base e is the case where that constant is one, which is why the formulas you learned first were the simplest ones in the family.
Section
Part 7
Pattern
Figure (svg): A flow diagram of seven boxes. Top row: ln x by integral, then derivative one over x, then increasing and onto the reals, then the number e. The chain then drops and runs back along a second row: exp, the inverse, then a to the x, then log base a. Each arrow is labelled with the theorem that justifies it.
This is the slide to revisit if you are ever asked to explain what Section 6.7 achieves. Seven objects, six theorems, and a strict order in which they must appear.
The order is forced, not chosen. The derivative must come before monotonicity, monotonicity before invertibility, invertibility before the exponential, and the exponential before any general power. Swap any two adjacent steps and the earlier one starts using something that does not exist yet.
The last two arrows are the ones students most often forget are theorems at all. Defining a general power as an exponential of a multiple of a logarithm, and a general logarithm as the natural one divided by a constant, are not notational conveniences. They are what make the two most-used function families in applied mathematics well defined for every real input.
Ranking
Put in order
Each step depends on the ones before it. Put them in the order the section builds them.
Why: The derivative has to come before anything else, because the proof that the logarithm is increasing, and therefore invertible, rests on it. The algebra has to come before e, because the existence proof doubles the logarithm of two. And e has to exist before there is anything to raise to a power.
Drag the steps into order before checking. If you get it right first time, you have understood the architecture, and if not, the explanation will tell you which dependency you missed.
Two orderings are genuinely tempting and both are wrong. Putting the algebra before the derivative fails, because the proof that the logarithm is increasing, which is what makes the later inversion legal, comes from the derivative. Putting e before the algebra fails, because the existence proof for e doubles the logarithm of two and that doubling is one of the algebraic properties.
The general power has to be last for the simplest reason of all: it is written using the exponential, and the exponential is written using e, and e is defined using the logarithm.
Real world
Acidity is reported on the pH scale, which is minus the logarithm to base ten of the hydrogen-ion concentration in moles per litre. A calculator stores no separate base-ten logarithm; it stores one routine and rescales.
\[ c = 3.2 \times 10^{-4}, \qquad \ln c = -8.04719 \]
Discussion prompt
Using the change-of-base formula from this part, and only the natural logarithm of the concentration given above, work out the pH.
Work the number out before revealing. The arithmetic is one division and one sign change, and the point is what the division means rather than the answer.
Chemists, acousticians and seismologists all work in base ten, and musicians work in base two, but no calculator or programming language stores a separate routine for each. There is one routine for the natural logarithm, and every other base is that value divided by a fixed constant, which is exactly the change-of-base formula from this part.
So the formula is not an exam curiosity. It is the reason a single algorithm serves every logarithmic scale in science, and the reason the constants 2.302585 and 0.693147 appear buried in so much numerical code.
Check
Check your understanding
Which statement is the definition of the natural logarithm used in this section?
Answer: B
Why: The section deliberately defines the logarithm as an integral, so that nothing about exponentials is assumed. Everything else about the logarithm, including the fact that it is the inverse of an exponential, is then proved from that integral.
The question is about what this section defines, not about what is true. Two of the wrong options describe statements that are perfectly true by the end of the lesson; they are simply not the starting point.
The option describing the logarithm as an exponent is the definition from Chapter 1, and it is the one this section replaces, precisely because it presupposes an exponential for all real exponents. By the end of the section it becomes a theorem, and a theorem is a very different thing from a definition.
Check
Check your understanding
What is the derivative of the logarithm to base five of x?
Answer: A
Why: Write the logarithm to base five as the natural logarithm divided by the natural logarithm of five. The denominator is a constant, so the derivative is one over that constant times one over x, which is one over x times the logarithm of five.
Convert to the natural logarithm before differentiating and the question answers itself. The logarithm to base five is the natural logarithm divided by a constant, so its derivative is the derivative of the natural logarithm divided by the same constant.
The tempting wrong answer multiplies by the constant instead of dividing. If you are unsure which way round it goes, check the size: a base-five logarithm grows more slowly than the natural one, so its slope must be smaller, and dividing by a number bigger than one is what makes it smaller.
Check
Check your understanding
Which is the antiderivative of five to the x?
Answer: B
Why: Differentiating five to the x produces an extra factor of the logarithm of five, so the antiderivative must divide by it. Check by differentiating option B: the logarithm of five comes down and cancels the one in the denominator.
Every one of the four options is the result of some rule; the question is which rule applies. The variable is in the exponent, so the two power-rule options are out immediately.
Between the remaining two, the choice is whether to multiply or divide by the logarithm of five. Differentiating produces that factor, so integrating must remove it. Differentiate the answer you chose and watch whether the constant cancels or squares.
Explain it to yourself
Discussion prompt
In two or three sentences, explain why the natural logarithm is built from the integral of one over t and not from the integral of any other power of t. Mention both the power rule for integrals and the substitution used in the product proof.
Write your answer out in full sentences before revealing. Being able to say this is a better test of understanding than any calculation in the lesson, because it connects two things that look unrelated.
The first half is about the power rule: every power has an antiderivative given by one formula except the exponent minus one, so that exponent is the one that needs a function of its own. The second half is about the substitution: the scale factor introduced by stretching cancels only when the denominator carries the variable to the first power.
The insight to reach for is that these are not two coincidences. Both are consequences of the same exponent, and that is why the function forced on us by the gap in the power rule turns out to be exactly the function that turns products into sums.
Exit ticket
\[ x^{\pi}, \qquad x>0 \]
Discussion prompt
Say precisely what this expression means, differentiate it, and name the definition or theorem that licenses each move. Note what your answer says about the power rule you learned in Chapter 3.
Do the whole thing on paper before revealing: state the meaning, differentiate, and name the justification for each move. Naming the justification is the part that matters, because it is the habit this section is training.
The meaning comes from the definition of a general power, with a positive base and a real exponent. The differentiation is a chain rule on an exponential, and the inner derivative is the derivative of the natural logarithm, which is where the factor of one over x comes from. Collecting the two factors turns the exponential back into the original power.
Now look at what you have proved. The power rule holds for an irrational exponent. In Chapter 3 that statement could not even be made, because the expression it is about had no meaning; here it is a two-line consequence. If you want one sentence to take away from Section 6.7, that is a good candidate.
Recap
| step | what defines it | what it buys |
|---|---|---|
| ln x | the integral of 1/t from 1 to x | a derivative of 1/x, and the algebra of logarithms |
| e | the unique input with ln e = 1 | a base whose exponential is its own derivative |
| e to the x | the inverse of ln, defined for all real x | the laws of exponents, proved not assumed |
| a to the x | e to the power x times ln a | a meaning for every real exponent, including irrational ones |
| log base a | ln x divided by ln a | a derivative of 1/(x ln a), and change of base |
\[ \int\frac{du}{u} = \ln|u| + C, \qquad \int e^{x}\,dx = e^{x} + C, \qquad \int a^{x}\,dx = \frac{a^{x}}{\ln a} + C \]
The gap opened in Chapter 1 is now closed, and the power rule finally holds for every real exponent. Chapter 6 ends here; the next volume uses these functions constantly, and it uses them knowing they exist.
Stewart, Calculus: Early Transcendentals 8e, Appendix G — The Logarithm Defined as an Integral Appendix G, pp. A50-A56 — the same construction in Stewart
Read the table downwards as a single argument. Each row is defined using only the rows above it, and each row buys something the next row needs. That structure, rather than any individual formula, is what the section is for.
Three integration formulas are worth committing to memory, and they are on the slide together for comparison. The reciprocal gives a logarithm with absolute value bars; the exponential gives itself; a general base gives itself divided by the logarithm of the base. Notice that the third reduces to the second when the base is e, which is a useful way to check you have remembered it correctly.
If you want one sentence to carry away, make it this: the natural logarithm is an area, and everything else in the exponential family is built from it. That is not how the functions are usually introduced, but it is the only order in which every step can be justified, and it is the order a later course will assume you have seen.
Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.