The hyperbolic functions as the even and odd parts of the exponential, the identity that names them, their derivatives and integrals, the inverse functions with their logarithmic formulas, and the catenary.
Subject: Calculus I · 67 slides · symbolic lesson
Open the interactive version of this deck
Title
Calculus I · Section 6.9
A family built from the exponential, and the shape of a hanging chain
Objectives
Section 6.7 built the exponential and the natural logarithm from scratch. Two particular combinations of the exponential turn out to behave so much like the sine and cosine that they borrow their names, and so little like them in one crucial place that the difference is worth a whole section.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, pp. 646-656 — learning objectives 6.9.1 to 6.9.3
You have spent Chapter 6 applying integration, and Section 6.7 gave you the exponential and the natural logarithm on a secure footing. This section takes two specific combinations of the exponential and gives them names, and then does calculus with them.
Do not let the borrowed names mislead you. These are not the sine and cosine in disguise, and nothing here is periodic. What is true is that the family satisfies identities and derivative rules that look uncannily like the trigonometric ones, with a single sign changed, and the whole lesson is organised around finding that sign and remembering where it lives.
The payoff is at the end. A cable hanging under its own weight takes a shape that no polynomial and no trigonometric function describes, and the hyperbolic cosine describes it exactly. Computing the length of such a cable turns out to be one of the very few arc-length problems in all of calculus that comes out in closed form, and the reason is a single identity you will meet in Part 2.
Warm-up
Discussion prompt
Every function on the whole line splits, in exactly one way, into an even function plus an odd one. Write down the recipe, then apply it to the exponential. What do the two pieces look like?
Write your answer down before revealing. The recipe is worth rediscovering rather than recalling, because the rediscovery explains why the two halves are useful.
Take any function and average it with its reflection. Anything odd in the original cancels, so what is left is even. Take half the difference instead and anything even cancels, so what is left is odd. Add the two pieces back together and the reflections cancel, returning the original.
Now do it to the exponential. The even piece is the average of the exponential and its reflection, and the odd piece is half their difference. Those two expressions are the whole content of the next slide, and everything else in this lesson follows from them by algebra you can already do.
Section
Part 1
Concept
The hyperbolic cosine is the even half of the exponential and the hyperbolic sine is the odd half. That is the entire definition; everything else in this lesson is a consequence of it.
\[ \cosh x = \frac{e^{x}+e^{-x}}{2}, \qquad \sinh x = \frac{e^{x}-e^{-x}}{2} \]
Figure (svg): The curve y equals the exponential, dashed, with the hyperbolic cosine and hyperbolic sine solid beneath it; at x equals 1 three dots mark the heights 2.718, 1.543 and 1.175.
Because the definitions are elementary, so is every proof in this section. There is no new theory here, only a new pair of names for two combinations you can already handle.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 646 — the definitions
Look at the figure before reading any formula. The dashed curve is the exponential. The two solid curves are its halves, and at every value of x their two heights add up to the dashed height. At x equal to 1 that reads 1.543 plus 1.175 equals 2.718, which is the number e itself.
On the right both solid curves press in toward half the dashed curve, because the reflected exponential has almost died away and the two definitions differ only by it. On the left they separate: the reflected term dominates, so the even half climbs while the odd half plunges.
The lesson to carry forward is that there is no new theory here. Any claim you want to make about these functions can be settled by writing them out in exponentials and doing a line or two of algebra. You will do exactly that four or five times in this lesson, and every time it will work.
Notation
Annotate
On: \( \cosh x = \frac{e^{x}+e^{-x}}{2}, \quad \sinh x = \frac{e^{x}-e^{-x}}{2} \)
Step through the annotations one at a time. The first two are the whole design of the pair: one sign makes an even function, the other an odd one, and nothing else differs between them.
The third annotation explains why there is a two underneath. Without it you would get the sum and the difference rather than the average and half the difference, and the values at zero would come out as 2 and 0 rather than 1 and 0. The 1 and the 0 are what make the analogy with cosine and sine work at the origin.
The last annotation is the one that saves you later. There is no angle in sight, and no radians. If you ever catch yourself expecting a period, or looking for where the hyperbolic sine next returns to zero, go back and read this line again.
Intuition
Two reasons, and both will be visible before the lesson ends. First, they satisfy identities and derivative rules that copy the trigonometric ones almost exactly, so everything you know about sines and cosines transfers with one sign changed.
Second, they answer questions the circular functions cannot. A hanging chain, the speed of a falling body with air resistance, the shape of a soap film between two rings: each is a hyperbolic cosine or tangent, and none of them is a sine.
The names are borrowed because the parallel is exact, not decorative. It runs all the way down to what the parameter means in each family.
It is fair to ask why these particular combinations deserve names when infinitely many others do not. The honest answer has two parts, and you should hold both.
The first is internal: the family is closed under differentiation in the same neat way as the trigonometric family, and it satisfies matching identities. That means every technique you learned for sines and cosines transfers, which is a lot of value for one definition.
The second is external. Problems in the physical world hand you these functions whether you wanted them or not. A hanging cable, a body falling with quadratic drag, the shape of a soap film stretched between two rings, the special relativity of velocities that add: none of those is a sine, and all of them are hyperbolic. You will meet two of them before the lesson ends.
Worked example
Add the two definitions, then subtract them, and see what comes back.
\[ \cosh x = \frac{e^{x}+e^{-x}}{2}, \qquad \sinh x = \frac{e^{x}-e^{-x}}{2} \]
Add the two fractions
Why: One denominator, so add the numerators.
\[ \cosh x + \sinh x = \frac{(e^{x}+e^{-x}) + (e^{x}-e^{-x})}{2} \]
Cancel the reflected terms
Why: The two copies of the reflected exponential have opposite signs.
\[ \cosh x + \sinh x = \frac{2e^{x}}{2} = e^{x} \]
Subtract instead
Why: Now the two forward copies cancel.
\[ \cosh x - \sinh x = \frac{2e^{-x}}{2} = e^{-x} \]
Replace x by its negative in each definition
Why: Swapping the two exponentials leaves the sum alone and reverses the difference.
\[ \cosh(-x) = \cosh x, \qquad \sinh(-x) = -\sinh x \]
Check by multiplying the two results
Why: The product of the exponential and its reflection is one, and the left side is a difference of squares, so this is the fundamental identity arriving three slides early.
\[ (\cosh x + \sinh x)(\cosh x - \sinh x) = e^{x}e^{-x} = 1 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 654 — Exercise 377
This is the shortest possible first exercise, and it is worth doing carefully because it establishes the two facts you will use constantly. Adding the definitions returns the exponential; subtracting them returns its reflection.
The parity computation in the fourth step is equally cheap. Replacing x by its negative just swaps the two exponentials. In the even function nothing changes, because addition does not care about order. In the odd one everything changes sign, because the two terms exchange places in a subtraction.
The check at the end is the real prize. The product of a sum and a difference is a difference of squares, and the product of the exponential with its reflection is 1. So you have proved the fundamental identity of Part 2 without meaning to. Watch for that when you reach it: you will already have done the work.
Concept
The remaining four are the same quotients as in trigonometry, formed from the two definitions above.
\[ \tanh x = \frac{\sinh x}{\cosh x}, \quad \coth x = \frac{\cosh x}{\sinh x}, \quad \operatorname{sech} x = \frac{1}{\cosh x}, \quad \operatorname{csch} x = \frac{1}{\sinh x} \]
\[ \tanh x = \frac{e^{x}-e^{-x}}{e^{x}+e^{-x}} \]
Figure (svg): Three panels: the hyperbolic sine rising through the origin, the hyperbolic cosine a U-shape with its lowest point at height one, and the hyperbolic tangent flattening between the lines y equals minus one and y equals one.
The hyperbolic cosine is never zero, so the tangent and secant are defined for every real input. The hyperbolic sine vanishes only at the origin, so the cotangent and cosecant break in exactly one place.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, pp. 646-647 — the definitions and Figure 6.81
The four quotients are defined exactly as in trigonometry, so there is nothing new to learn about the pattern. What is new is where they are defined.
Look at the three graphs. The first is strictly increasing and passes through the origin. The second is a U with its floor at height one, never lower. The third is squeezed between two horizontal lines, approaching one on the right and minus one on the left, which is the behaviour that makes the hyperbolic tangent so useful for modelling anything that saturates.
Now compare with the circular case. The circular tangent blows up at infinitely many places, because the circular cosine keeps returning to zero. The hyperbolic cosine never reaches zero at all, so the hyperbolic tangent and secant are defined for every real number. That single difference is worth noticing now, because it will come back when you invert these functions.
Picture it
Figure (svg): Three panels: the hyperbolic cotangent with a break at the origin and horizontal lines at plus and minus one, the hyperbolic secant a smooth bump of height one, and the hyperbolic cosecant with a break at the origin.
Compare this with the circular picture, where the tangent, secant, cotangent and cosecant each have infinitely many vertical asymptotes. Here there is at most one, at the origin, and the secant has none at all.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 647 — Figure 6.81, panels (d) to (f)
Three more graphs, and one thing to notice in each. The hyperbolic cotangent has a single break at the origin, with the two branches approaching the lines at height one and minus one from outside. The hyperbolic secant is a smooth bump with a maximum of exactly one, defined everywhere. The hyperbolic cosecant breaks at the origin and decays toward zero on both sides.
None of these has more than one problem point, and that is the whole difference from trigonometry, where the same four functions each have infinitely many asymptotes marching off in both directions.
Keep the hyperbolic secant in mind in particular. It is the derivative of the hyperbolic tangent, squared, and it will turn up in every integral involving the hyperbolic tangent for the rest of the lesson.
Fill the middle
Fill in the blanks
\cosh x = \frac-x + e^-x}}___, \qquad \sinh x = \frac___ - e^___}}___
Why: Both blanks are the reflected exponential. The only difference between the two definitions is the sign in front of it: a plus makes the combination even, a minus makes it odd.
Type both blanks before checking. If you have to think, reconstruct rather than recall: the even half must be unchanged when x is replaced by its negative, and only the reflected exponential can make that happen.
Getting this to be automatic matters more than it looks. Almost every proof in this section begins by writing these two lines, so an unreliable memory of them will cost you on every problem rather than just this one.
Sorting
Sort into buckets
Replace x by its negative in each and decide.
Sort all six before revealing. The method is the same in every case: replace x by its negative and see whether the whole expression is unchanged, reversed in sign, or neither.
For the quotients you do not need to expand anything. A quotient of an even function by an even function is even; a quotient of an odd function by an even function is odd; and a quotient of an even by an odd is also odd, since exactly one factor flips.
The last item is the point of the exercise. The exponential is neither even nor odd, and that is precisely why splitting it was worth doing. Every one of the other five inherits its parity from the split.
Prediction
\[ \tanh x = \frac{e^{x}-e^{-x}}{e^{x}+e^{-x}} \]
Predict first
As x grows large and positive, what happens to the hyperbolic tangent?
Correct: It approaches 1
Why: Divide top and bottom by the forward exponential: the quotient becomes one minus a vanishing term, over one plus a vanishing term, so it tends to 1. Going the other way it tends to −1. The circular tangent has no limit at all, which is one of the places the two families come apart.
Commit to one option before working it out. Then do the computation: divide the numerator and the denominator by the forward exponential and see what survives.
What survives on top is one minus a term that dies away, and on the bottom one plus a term that dies away. So the quotient approaches one. Going the other direction, the reflected exponential dominates instead and the quotient approaches minus one.
This is the behaviour that makes the hyperbolic tangent the standard model for anything that grows and then saturates: a falling body reaching terminal speed, a neuron firing rate, a chemical reaction running out of reagent. The circular tangent could never do this job, because it has no limit at all.
Section
Part 2
Concept
Square both definitions and subtract. The whole proof is two expansions and a cancellation.
\[ \cosh^2 x = \frac{e^{2x} + 2 + e^{-2x}}{4}, \qquad \sinh^2 x = \frac{e^{2x} - 2 + e^{-2x}}{4} \]
\[ \cosh^2 x - \sinh^2 x = \frac{(e^{2x}+2+e^{-2x}) - (e^{2x}-2+e^{-2x})}{4} \]
\[ \cosh^2 x - \sinh^2 x = \frac{4}{4} = 1 \]
the fundamental hyperbolic identity — The square of the hyperbolic cosine minus the square of the hyperbolic sine is one, for every real input.
The outer terms are identical in the two expansions, so they cancel; only the cross terms survive, and they differ in sign, so subtracting doubles them.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 650 — the identity used in the implicit differentiation
Follow the algebra line by line and notice which terms do the work. Both squares expand to the same three pieces: a forward exponential of twice x, a constant, and a reflected exponential of twice x, all over four.
The two outer pieces are identical in the two expansions, so subtracting kills them. The constants are plus two and minus two, so subtracting doubles them to four, and four over four is one. That is the entire proof.
Now compare with the circular identity. There the two squares add to one; here they subtract to one. One sign, and it is the sign that gives the family its name. Every other difference you will meet in this lesson traces back to this line, so it is worth being able to reproduce in under a minute.
Picture it
Figure (svg): Left: the unit circle with the point at angle one radian marked and the sector from the positive x-axis shaded. Right: the right branch of the unit hyperbola with the point whose coordinates are the hyperbolic cosine and sine of one marked, and the region between the axis, the curve and the ray to that point shaded.
Because the identity has a minus sign, the point whose coordinates are the hyperbolic cosine and sine of a number satisfies the equation of a hyperbola, not a circle. That is where the family's name comes from, and the hyperbolic cosine is never less than one, so the point always lands on the right-hand branch.
Both panels show a point moving along a curve as a parameter changes, and both show the region swept out from the positive horizontal axis.
On the left, the coordinates are the cosine and sine of the parameter, the curve is the unit circle, and the shaded sector has area equal to half the parameter. That is a fact you may never have been told: the parameter of the circular functions is an angle, yes, but it is also twice a sector area.
On the right the coordinates are the hyperbolic cosine and sine, the curve is the unit hyperbola, and the shaded region again has area equal to half the parameter. Here there is no angle at all, so the area is the only honest interpretation. That is the deep version of the analogy, and it is why the names are apt rather than merely convenient. Notice too that the point always stays on the right-hand branch, because the hyperbolic cosine is never less than one.
Notation
Annotate
On: \( \cosh^2 t - \sinh^2 t = 1 \)
Work through the four annotations in order. The first is the headline: the sign is the whole story, and if you remember only one thing from Part 2 it should be that the hyperbolic identity subtracts where the Pythagorean one adds.
The third annotation answers a question people usually do not think to ask. If the point lies on a hyperbola, which of the two branches is it on? Always the right-hand one, because the first coordinate is a hyperbolic cosine and that is never below one. The left branch is simply not reachable by this parametrisation.
The fourth is the one to sit with. Once you know the parameter is twice a sector area in both families, the parallel stops feeling like a coincidence and starts feeling like two instances of one construction.
Concept
Move the squared hyperbolic sine to the other side and you get the version used every time a hyperbolic function appears under a square root.
\[ 1 + \sinh^2 x = \cosh^2 x \]
\[ \sqrt{1 + \sinh^2 x} = \cosh x \]
The square root needs no absolute value, because the hyperbolic cosine is positive everywhere. Compare the circular case, where one plus a squared tangent gives a squared secant and the sign does need care.
\[ 1 - \tanh^2 x = \operatorname{sech}^2 x, \qquad \coth^2 x - 1 = \operatorname{csch}^2 x \]
Those last two come from dividing the fundamental identity by the squared hyperbolic cosine, and then by the squared hyperbolic sine.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 654 — the identity used in Example 6.51
This slide looks like bookkeeping and is actually the most practically useful rearrangement in the section. Whenever a hyperbolic sine appears squared under a square root, this identity removes the root completely.
Pay attention to the absence of an absolute value. When you do the same manoeuvre with circular functions, the square root of a squared secant is the size of the secant, and you have to worry about which quadrant you are in. Here the hyperbolic cosine is positive everywhere, so there is nothing to worry about, and that is one reason the catenary's arc length comes out so cleanly.
The last two identities come from dividing the fundamental one by a squared hyperbolic cosine, then by a squared hyperbolic sine. Compare them with the circular versions and note that the first has its signs swapped while the second reads almost the same. If you can derive them in ten seconds you never need to memorise them.
Worked example
Add the two squares instead of subtracting them, and a second identity falls out.
\[ \text{Show } \cosh^2 x + \sinh^2 x = \cosh 2x. \]
Write the two squares
Why: Both expansions have the same outer terms and opposite cross terms.
\[ \cosh^2 x = \frac{e^{2x}+2+e^{-2x}}{4}, \quad \sinh^2 x = \frac{e^{2x}-2+e^{-2x}}{4} \]
Add them
Why: The constants cancel this time instead of the exponentials.
\[ \cosh^2 x + \sinh^2 x = \frac{2e^{2x} + 2e^{-2x}}{4} = \frac{e^{2x}+e^{-2x}}{2} \]
Recognise the right-hand side
Why: That is the definition of the hyperbolic cosine with twice x in place of x.
\[ \frac{e^{2x}+e^{-2x}}{2} = \cosh 2x \]
Differentiate both sides, as Exercise 382 asks
Why: Chain rule on the right brings down a factor of two.
\[ 2\cosh x\sinh x + 2\sinh x\cosh x = 2\sinh 2x \]
Divide by two
Why: The double-angle formula for the hyperbolic sine, identical to the circular one.
\[ \sinh 2x = 2\sinh x\cosh x \]
Check at x equal to 1
Why: Twice 1.543081 times 1.175201 is 3.626860, and the hyperbolic sine of 2 is 3.626860.
\[ 2\cosh(1)\sinh(1) = 3.626860 = \sinh 2 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 654 — Exercises 381 and 382
You have already expanded both squares once, so the first two steps cost nothing. This time the constants cancel instead of the exponentials, which is the whole difference between this identity and the fundamental one.
The recognition step is the one to slow down on. Once the sum is written as a forward exponential of twice x plus a reflected one, all over two, you should see the definition of the hyperbolic cosine with twice x inside it. Training yourself to read the definition backwards like this is what makes the rest of the section quick.
Exercise 382 then gets a second identity for free by differentiating. Notice that the result, the hyperbolic sine of twice x equals twice the product, is identical to the circular double-angle formula with no sign change. Not everything differs between the families, and knowing which things match is as valuable as knowing which things do not.
Worked example
Change everything to exponentials, expand, and watch the mixed terms cancel.
\[ \text{Prove } \sinh(x+y) = \sinh x\cosh y + \cosh x\sinh y. \]
Write the first product in exponentials
Why: Two fractions with denominator two multiply to one with denominator four.
\[ \sinh x\cosh y = \frac{(e^{x}-e^{-x})(e^{y}+e^{-y})}{4} \]
Expand it
Why: Four terms, with the signs coming from the first bracket.
\[ = \frac{e^{x+y} + e^{x-y} - e^{-x+y} - e^{-x-y}}{4} \]
Expand the second product
Why: The same four exponentials, with the two mixed ones carrying the opposite signs.
\[ \cosh x\sinh y = \frac{e^{x+y} - e^{x-y} + e^{-x+y} - e^{-x-y}}{4} \]
Add the two expansions
Why: The mixed terms cancel in pairs and the others double.
\[ \frac{2e^{x+y} - 2e^{-x-y}}{4} = \frac{e^{x+y} - e^{-(x+y)}}{2} \]
Read off the definition
Why: That is the hyperbolic sine evaluated at the sum.
\[ \sinh x\cosh y + \cosh x\sinh y = \sinh(x+y) \]
Check with numbers
Why: At x equal to 1 and y equal to one half the left side gives 2.129279, and the hyperbolic sine of 1.5 is 2.129279.
\[ \sinh(1)\cosh(0.5) + \cosh(1)\sinh(0.5) = 2.129279 = \sinh(1.5) \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 654 — Exercise 383
This is the longest pure algebra in the lesson, and it is worth doing once by hand so you trust the pattern. Everything becomes exponentials, everything expands, and the terms that should cancel do.
Watch the four mixed terms in particular. In the first expansion the terms with mixed signs in their exponents carry a plus and a minus; in the second they carry a minus and a plus. Adding the expansions therefore removes both, and the two surviving terms double.
Compare the result with the circular addition formula for the sine: identical, with no sign change anywhere. The addition formula for the hyperbolic cosine, by contrast, does change sign relative to its circular twin. That pattern of which formulas keep their signs and which flip is worth collecting as you go, rather than memorising as a list at the end.
Trap
Copying Pythagoras straight across:
\[ \cosh^2 x + \sinh^2 x = 1 \]
Wrong. That sum is not constant.
Adding the squares gives the hyperbolic cosine of twice the input, which grows without bound. The identity that actually holds carries a minus sign.
\[ \cosh^2 x - \sinh^2 x = 1 \]
\[ \cosh^2 x + \sinh^2 x = \cosh 2x \]
At the input 1 the wrong version claims 1; the true value is 3.762196.
This is the single most common error in the section, and it usually comes from good instincts rather than bad ones. The names are trigonometric, the pattern of the identity is familiar, and the hand writes a plus before the brain intervenes.
The defence takes two seconds. Put x equal to 1 and compute: the sum of the squares comes out to 3.762196, nowhere near 1. Any identity you are unsure about can be killed or survived by a single numerical test, and the numbers here are easy ones.
The deeper defence is to remember what the sum actually equals. It is the hyperbolic cosine of twice x, an unbounded function. If a claimed identity says an unbounded expression equals a constant, something has gone wrong.
Comparison
Comparison matrix
| quantity | circular | hyperbolic |
|---|---|---|
| derivative of the cosine | −sin t | sinh t |
| derivative of the sine | cos t | cosh t |
| sign in the identity | plus | minus |
| curve traced by the pair | circle | hyperbola |
| period | 2π | none |
| value of the cosine at 0 | 1 | 1 |
Fill in the right-hand column before checking. Every entry is something you have already seen on a slide in the last few minutes, so this is a memory check rather than a calculation.
Two rows deserve a second look. The derivative of the cosine loses its minus sign in the hyperbolic column, and that is the only place in either derivative table where a sign disappears. The period row has no entry at all on the hyperbolic side, which is not an oversight: nothing built from real exponentials repeats.
Keep this table. When you meet a hyperbolic identity you half remember, check it against the circular one and then ask which of these six differences might apply.
Two truths and a lie
Eliminate the wrong options
Two of these statements are true. Eliminate those and keep the false one.
Survives elimination: C
Why: The circular sine vanishes at every whole multiple of pi. The hyperbolic sine, built from real exponentials, is strictly increasing and crosses zero exactly once. No hyperbolic function is periodic, and that single fact kills off most of the false analogies people bring to this section.
Knock out the two you are sure of first, and read your reason for each one as you go; the reasons matter more than the verdict.
The false statement is the one about how often the hyperbolic sine vanishes. It is strictly increasing, so it can cross any horizontal level at most once, and the level zero is crossed at the origin. There is exactly one solution, not infinitely many.
This is the periodicity misconception in its most testable form. The circular sine vanishes at every whole multiple of pi, and if you carry that expectation across you will believe this false claim. Every hyperbolic function is monotone or one-bump, and none of them repeats.
Section
Part 3
Concept
There is nothing to memorise yet. Differentiate the definition term by term and the answer appears.
\[ \frac{d}{dx}\sinh x = \frac{d}{dx}\left(\frac{e^{x}-e^{-x}}{2}\right) \]
\[ = \frac12\left[\frac{d}{dx}e^{x} - \frac{d}{dx}e^{-x}\right] \]
\[ = \frac12\left[e^{x} + e^{-x}\right] = \cosh x \]
The minus in front of the reflected exponential meets the minus from its own chain rule, and the two cancel. Doing the same to the other definition gives the hyperbolic sine back, with no sign change at all.
\[ \frac{d}{dx}\cosh x = \sinh x \]
Exercise 379 asks for the consequence: differentiate either of them twice and you get back where you started, so both solve the growth equation rather than the oscillation equation.
\[ y'' = y \quad\text{for } y = \sinh x \text{ and } y = \cosh x \]
\[ \text{compare: } y'' = -y \quad\text{for } y = \sin x \text{ and } y = \cos x \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, pp. 647, 654 — the derivation before Table 6.2, and Exercise 379
There is nothing to memorise on this slide, and that is the point. You already know how to differentiate the exponential and its reflection, so you already know how to differentiate both hyperbolic functions.
Follow the signs in the first derivation closely, because they are what people get wrong. The definition of the odd half has a minus in front of the reflected exponential. Differentiating the reflected exponential brings down another minus from the chain rule. Two minuses make a plus, so both terms end up positive, and the result is the even half.
Doing the same to the even half is one line shorter, and the answer is the odd half with no sign change. Write out that second derivation yourself before moving on; it takes fifteen seconds and it will stop you from ever writing the wrong sign under exam pressure.
The last two lines are the deepest statement of the difference between the two families. Differentiate a hyperbolic function twice and you return to it unchanged; differentiate a circular one twice and you return to its negative. One equation describes runaway growth, the other describes oscillation, and that is why one family models a hanging cable and the other a vibrating string.
Concept
The other four follow from the quotient rule. Exercise 380 asks for the first of them, and the identity from Part 2 does the last step.
\[ \frac{d}{dx}\tanh x = \frac{\cosh x\cosh x - \sinh x\sinh x}{\cosh^2 x} = \frac{1}{\cosh^2 x} = \operatorname{sech}^2 x \]
| f(x) | its derivative |
|---|---|
| sinh x | cosh x |
| cosh x | sinh x |
| tanh x | sech² x |
| coth x | −csch² x |
| sech x | −sech x tanh x |
| csch x | −csch x coth x |
Figure (svg): Two panels. Left: the cosine curve with its derivative, the negative of the sine, drawn beneath it. Right: the hyperbolic cosine with its derivative, the hyperbolic sine, drawn with no sign change.
Compare row by row with trigonometry. The sine rows match exactly. The cosine rows differ by a sign, and so do the two secant rows and the two cotangent rows. Every minus sign that appears in the circular table appears here too, except the one on the cosine, which disappears.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, pp. 647-648 — Table 6.2 and Exercise 380
The first line does Exercise 380 in one go. The quotient rule gives a numerator of a squared hyperbolic cosine minus a squared hyperbolic sine, and by the fundamental identity that is just 1, leaving one over the squared hyperbolic cosine.
Read the table against the trigonometric one row by row rather than memorising it. The first two rows are the ones that matter: the sine row matches exactly, the cosine row loses its minus. Every other minus sign in the circular table survives here too.
The picture makes the difference visible. On the left, the circular cosine and its derivative are out of step: where the cosine is falling, its derivative is negative, and the dashed curve mirrors the solid one about the horizontal axis on each half period. On the right, the hyperbolic cosine and its derivative are in step: both grow on the right, and the derivative is simply the odd companion. Nothing is reflected.
Worked example
Use the table together with the chain rule.
\[ \text{(a) } \frac{d}{dx}\Big(\sinh(x^2)\Big) \qquad \text{(b) } \frac{d}{dx}(\cosh x)^2 \]
Differentiate the outer function in (a)
Why: The derivative of the hyperbolic sine is the hyperbolic cosine, left at the same inside.
\[ \frac{d}{dx}\sinh(x^2) = \cosh(x^2)\cdot\frac{d}{dx}(x^2) \]
Differentiate the inside
Why: The power rule.
\[ \frac{d}{dx}\sinh(x^2) = 2x\cosh(x^2) \]
Start (b) with the power rule
Why: Outer square first, inside untouched.
\[ \frac{d}{dx}(\cosh x)^2 = 2\cosh x\cdot\frac{d}{dx}\cosh x \]
Differentiate the hyperbolic cosine
Why: No minus sign appears, unlike the circular case.
\[ \frac{d}{dx}(\cosh x)^2 = 2\cosh x\sinh x \]
Check (a) numerically
Why: A central difference of the hyperbolic sine of x squared at x equal to 1 gives 3.086161, and the formula gives twice the hyperbolic cosine of 1, which is 3.086161.
\[ 2(1)\cosh(1) = 3.086161 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 648 — Example 6.47
Both parts are chain-rule problems with a hyperbolic function in them, and neither needs anything beyond the table. Part (a) has the hyperbolic function outside and a polynomial inside; part (b) has a power outside and the hyperbolic function inside.
In part (a) the order matters: differentiate the outer function first, leaving its inside untouched, and only then multiply by the inside's derivative. Writing that first line before simplifying is a habit that prevents most chain-rule errors.
In part (b) notice what does not happen. Differentiating the hyperbolic cosine gives the hyperbolic sine with a plus sign. If this had been the circular cosine the answer would have carried a minus, and the resulting expression would have been minus the sine of twice x rather than plus the hyperbolic sine of twice x. One sign, two very different functions.
Worked example
\[ \text{(a) } \frac{d}{dx}\Big(\tanh(x^2+3x)\Big) \qquad \text{(b) } \frac{d}{dx}\left(\frac{1}{(\sinh x)^2}\right) \]
Name the outer derivative in (a)
Why: The hyperbolic tangent differentiates to the square of the hyperbolic secant.
\[ \frac{d}{dx}\tanh u = \operatorname{sech}^2 u\cdot\frac{du}{dx} \]
Put the inside and its derivative in
Why: The inside differentiates to two x plus three.
\[ \frac{d}{dx}\tanh(x^2+3x) = (2x+3)\operatorname{sech}^2(x^2+3x) \]
Rewrite (b) as a power
Why: A reciprocal square is the power minus two.
\[ \frac{1}{(\sinh x)^2} = (\sinh x)^{-2} \]
Apply the power rule and the chain rule
Why: Bring the minus two down and differentiate the inside.
\[ \frac{d}{dx}(\sinh x)^{-2} = -2(\sinh x)^{-3}\cosh x \]
Write it as one fraction
Why: Collect the negative power into the denominator.
\[ = -\frac{2\cosh x}{\sinh^3 x} \]
Check the sign of (b)
Why: For positive x the hyperbolic sine grows, so one over its square falls, and the derivative must be negative; the formula is negative there because both the hyperbolic cosine and the cube of the hyperbolic sine are positive.
\[ x > 0: \; \cosh x > 0, \; \sinh^3 x > 0 \;\Longrightarrow\; -\frac{2\cosh x}{\sinh^3 x} < 0 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 648 — Checkpoint 6.47
Part (a) is a direct application of row three of the table. The only thing to be careful about is that the squared hyperbolic secant keeps the same inside as the hyperbolic tangent it came from, and the extra factor is the inside's derivative, not something involving the secant.
Part (b) is the one that catches people, because the natural instinct is to reach for the quotient rule. Rewriting the reciprocal square as a power of minus two is faster and less error-prone; then it is the power rule and the chain rule, and no quotient rule at all.
The check is a habit worth building. Before you accept a derivative, ask whether its sign makes sense. Here the function is falling for positive x, so the derivative must be negative, and the formula obliges because everything in it except the leading minus is positive.
Trap
Differentiating by analogy:
\[ \frac{d}{dx}\cosh x = -\sinh x \]
Wrong. Nothing here produces a minus.
Differentiate the definition and the chain rule's minus cancels the minus already in front of the reflected exponential. The hyperbolic cosine differentiates to the hyperbolic sine with a plus.
\[ \frac{d}{dx}\cosh x = \sinh x \]
The minus signs in this family live in the reciprocal rows instead:
\[ \frac{d}{dx}\operatorname{sech} x = -\operatorname{sech} x\tanh x \]
If you take one thing from Part 3, take this. The derivative of the hyperbolic cosine has no minus sign. The instinct that it should is strong, it comes from years of trigonometry, and it costs marks on almost every test that covers this section.
The permanent defence is the derivation, not the memory. Differentiate the definition and watch the chain rule's minus cancel the definition's minus. Two lines, and then you never have to trust your memory again.
It also helps to know where the minus signs really are. They are in the reciprocal rows, on the secant and the cosecant and the cotangent, exactly as in trigonometry. The family did not lose its minus signs; it lost exactly one of them.
Concept
Every derivative formula is an integral formula turned around. These are the six the section lists, written with a general inside so they are ready for substitution.
\[ \int\sinh u\,du = \cosh u + C, \qquad \int\cosh u\,du = \sinh u + C \]
\[ \int \operatorname{sech}^2 u\,du = \tanh u + C, \qquad \int \operatorname{csch}^2 u\,du = -\coth u + C \]
\[ \int \operatorname{sech} u\tanh u\,du = -\operatorname{sech} u + C, \qquad \int \operatorname{csch} u\coth u\,du = -\operatorname{csch} u + C \]
Note the top row: both come out with a plus, so there is no sign to remember there. Every minus sign sits in the reciprocal formulas, exactly as in the derivative table.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 648 — the integral formulas after Table 6.2
Every antiderivative you will ever need in this family is already on the previous slide, written the other way round. This slide just performs that reversal, with a general inside so the formulas are ready for substitution.
Notice the shape of the list. The first pair is clean: the hyperbolic sine integrates to the hyperbolic cosine and the hyperbolic cosine to the hyperbolic sine, both with a plus. Every minus sign in the list belongs to a reciprocal function, and you can read off which one from the derivative row it came from.
When you meet a hyperbolic integral, your first move should be to ask whether it is one of these six with a substitution, not to reach for any cleverer technique. In this section it almost always is.
Worked example
\[ \text{(a) } \int x\cosh(x^2)\,dx \qquad \text{(b) } \int \tanh x\,dx \]
Substitute in (a)
Why: Let u be the inside of the hyperbolic cosine.
\[ u = x^2, \quad du = 2x\,dx, \quad x\,dx = \tfrac12\,du \]
Integrate in u
Why: The antiderivative of the hyperbolic cosine is the hyperbolic sine.
\[ \int x\cosh(x^2)\,dx = \tfrac12\int\cosh u\,du = \tfrac12\sinh u + C \]
Undo the substitution
Why: Put x squared back in.
\[ \int x\cosh(x^2)\,dx = \tfrac12\sinh(x^2) + C \]
Write (b) as a quotient
Why: The hyperbolic tangent is the hyperbolic sine over the hyperbolic cosine.
\[ \int\tanh x\,dx = \int\frac{\sinh x}{\cosh x}\,dx \]
Substitute the denominator
Why: Its derivative is exactly the numerator, so the fraction becomes du over u.
\[ u = \cosh x, \quad du = \sinh x\,dx \]
Integrate and drop the absolute value
Why: The hyperbolic cosine is never below one, so it is always positive and the bars are unnecessary.
\[ \int\frac{du}{u} = \ln|u| + C = \ln(\cosh x) + C \]
Check by differentiating
Why: The derivative of the logarithm of the hyperbolic cosine is the hyperbolic sine over the hyperbolic cosine, which is the integrand; numerically the integral from 0 to 1 is 0.433781, and the logarithm of the hyperbolic cosine of 1 is 0.433781.
\[ \frac{d}{dx}\ln(\cosh x) = \frac{\sinh x}{\cosh x} = \tanh x \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 649 — Example 6.48
Part (a) is the standard substitution pattern: the derivative of the inside is sitting outside, up to a constant. Spotting that the x out front is half the derivative of x squared is the whole problem.
Part (b) is the more interesting one. The hyperbolic tangent is not on the list of antiderivatives, so you have to make it into something that is. Writing it as a quotient exposes the fact that the numerator is exactly the derivative of the denominator, which is the classic signal for a logarithm.
Then comes a small luxury. Normally the antiderivative of one over u carries an absolute value, and you have to keep it. Here the hyperbolic cosine is never below one, so it is always positive, and the bars can go. Say why you are dropping them rather than just dropping them; on an exam that sentence is worth a mark.
Worked example
\[ \text{(a) } \int \sinh^3 x\cosh x\,dx \qquad \text{(b) } \int \operatorname{sech}^2(3x)\,dx \]
Substitute in (a)
Why: The hyperbolic cosine is the derivative of the hyperbolic sine, so the sine is the right choice for u.
\[ u = \sinh x, \quad du = \cosh x\,dx \]
Integrate the power
Why: The power rule in u.
\[ \int u^3\,du = \frac{u^4}{4} + C = \frac{\sinh^4 x}{4} + C \]
Substitute in (b)
Why: A linear inside, so u is three x and dx is a third of du.
\[ u = 3x, \quad du = 3\,dx \]
Integrate the squared hyperbolic secant
Why: Its antiderivative is the hyperbolic tangent.
\[ \tfrac13\int \operatorname{sech}^2 u\,du = \tfrac13\tanh u + C = \tfrac13\tanh(3x) + C \]
Check (a) by differentiating
Why: Bring the fourth power down, divide by the four already there, and multiply by the derivative of the inside.
\[ \frac{d}{dx}\frac{\sinh^4 x}{4} = \sinh^3 x\cosh x \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 649 — Checkpoint 6.48
Part (a) tests whether you can spot the derivative sitting beside the function. The hyperbolic cosine is exactly the derivative of the hyperbolic sine, so the integral is a power of u times du and the power rule finishes it.
Part (b) is even shorter, and the only thing to get right is the constant. A linear inside means the differential picks up the coefficient, so the answer carries a third out front. Forgetting that third is the single most common slip in linear substitutions.
Check both by differentiating in your head. A fourth power over four differentiates to a cube times the inside's derivative, which is the integrand. A third of the hyperbolic tangent of three x differentiates to a third times the squared secant times three, which is the integrand again.
Step zero
\[ \int\frac{e^{x}}{\sqrt{e^{2x}-1}}\,dx \]
Discussion prompt
Exercise 417. Before doing anything, decide what u should be and what shape the integral will take once you substitute. Which of the section's formulas will finish it?
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 655 — Exercise 417
Write down your substitution before revealing anything. Thirty seconds of planning here saves several minutes of misdirected algebra.
The signal is the numerator. It is the exponential, and the radical contains the exponential squared. Whenever the numerator is the derivative of something appearing inside the rest of the integrand, that something is your u.
Once the substitution is made, the integrand is one over the root of u squared minus one, and only one formula in this section produces that. Learning to look at the radical and name the answer before doing any work is the skill the next few slides are building.
Discrimination
Sort into buckets
Sort each integral by the kind of antiderivative it produces.
Sort all six before checking. The question is not what each answer is, but which of the two tables you would reach for, and that is decided by what the integrand looks like rather than by how hard it is.
The rule is simple once you see it. If a hyperbolic function appears in the integrand, you are reading Table 6.2 backwards and the answer is another hyperbolic function. If the integrand is purely algebraic, with a radical or a difference of squares and no hyperbolic function in sight, you are reading Table 6.4 backwards and the answer is an inverse.
That second case is the surprising one, and it is the reason this section belongs in an integration chapter at all. An integrand with no exponentials and no hyperbolic functions can still have a hyperbolic answer.
Real world
Figure (svg): The speed of a falling body with drag plotted against time: it starts along the straight free-fall line and then bends over to a horizontal dashed line at 3.13, the terminal speed.
Discussion prompt
Exercise 419. A falling body whose drag is proportional to the square of its speed obeys dv/dt = g − v². Show that the hyperbolic tangent below solves it, and say what the picture's horizontal line means.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 655 — Exercise 419
Work the verification before revealing it. Differentiating the proposed solution is one chain rule, and then a single identity finishes the job.
The identity you need is one minus a squared hyperbolic tangent equals a squared hyperbolic secant, which you met in Part 2. Once the derivative is written as g times the squared secant, rewriting the secant in terms of the tangent turns the right-hand side into g minus the square of the solution itself, which is exactly the equation.
Now read the picture. At the very start the speed is small, the drag term is negligible, and the curve rides along the straight free-fall line. As the speed grows the drag term eats into the acceleration, the curve bends over, and it approaches the level where drag exactly balances gravity. That level is the terminal speed, and it is the square root of g in the units this exercise uses. The hyperbolic tangent is doing what it always does: saturating.
Section
Part 4
Concept
The hyperbolic sine and tangent are strictly increasing, so they invert without any fuss. The hyperbolic cosine and secant are even, so each needs half its graph thrown away first, exactly as the circular cosine does.
Figure (svg): Three panels: the inverse hyperbolic sine defined for every x, the inverse hyperbolic cosine starting at x equals one, and the inverse hyperbolic tangent shooting up and down at the vertical lines x equals plus and minus one.
| function | domain | range |
|---|---|---|
| sinh⁻¹ x | all real x | all real y |
| cosh⁻¹ x | x at least 1 | y at least 0 |
| tanh⁻¹ x | between −1 and 1 | all real y |
| coth⁻¹ x | |x| greater than 1 | all y except 0 |
| sech⁻¹ x | greater than 0, up to 1 | y at least 0 |
| csch⁻¹ x | all x except 0 | all y except 0 |
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, pp. 649-650 — Table 6.3 and Figure 6.82
Inverting a function needs it to pass the horizontal line test, and the graphs tell you immediately which of the six qualify without help. The hyperbolic sine, tangent, cotangent and cosecant are each monotone on their domains, so they invert directly.
The hyperbolic cosine and secant are even, so each takes every value twice, and each needs half its graph discarded before it can be inverted. The convention is the same one you met with the circular cosine: keep the half where the input is non-negative.
Read the table alongside the graphs rather than memorising it. Every domain in the table is just the range of the original function, and every range is the part of the original domain that was kept. If you can say that sentence you can rebuild the whole table from the six pictures on the previous slides.
Intuition
Figure (svg): The right half of the hyperbolic cosine curve and its mirror image in the dashed line y equals x, which is the inverse hyperbolic cosine starting at the point one, zero.
Inverting a function means solving an equation for the input. For the hyperbolic family that equation is built from exponentials, and an equation in one exponential is a polynomial in disguise: clear the fractions, multiply through, and it becomes a quadratic you can solve.
For the circular family the same attempt produces an equation in two different periodic pieces, with infinitely many solutions and no algebraic way to pick one. That is why there is no elementary formula for the inverse sine, while the inverse hyperbolic sine is a single logarithm.
This is a question worth asking of every inverse function you meet, and the answer here is unusually clean. Inverting means solving an equation for the input, and whether that is possible depends entirely on what kind of equation it is.
For a hyperbolic function the equation involves one exponential and its reciprocal. Multiply through by the exponential and it becomes a quadratic in that exponential, which you can solve by formula. Take a logarithm and you are done. Every inverse in this family is therefore a logarithm of something algebraic.
For a circular function the same attempt gives an equation in a quantity that returns to the same value infinitely often, and there is no algebraic way to single out one solution. That is why the inverse sine has no formula in terms of roots and logarithms, while the inverse hyperbolic sine has a formula you can write on one line. The picture on this slide shows the other half of the story: the restriction to non-negative inputs is what makes the hyperbolic cosine invertible at all, and reflecting in the dashed line is what the inverse does geometrically.
Worked example
Turn the defining equation into a quadratic, as the exercise suggests, and solve it.
\[ y = \sinh^{-1}x \iff \sinh y = x \]
Write the definition out
Why: The hyperbolic sine of y is a difference of exponentials over two.
\[ \frac{e^{y} - e^{-y}}{2} = x \]
Clear the fraction and multiply through by the exponential
Why: This is the move the exercise names; it removes the negative exponent.
\[ e^{2y} - 1 = 2x\,e^{y} \]
Read it as a quadratic
Why: Let w stand for the exponential of y.
\[ w^2 - 2xw - 1 = 0 \]
Solve it
Why: The discriminant is four x squared plus four, whose root is twice the root of x squared plus one.
\[ w = \frac{2x \pm \sqrt{4x^2+4}}{2} = x \pm \sqrt{x^2+1} \]
Discard the minus root
Why: The root of x squared plus one is larger than the size of x, so the minus root is negative, and an exponential never is.
\[ e^{y} = x + \sqrt{x^2+1} \]
Take logarithms
Why: The natural logarithm undoes the exponential and leaves y alone.
\[ \sinh^{-1}x = \ln\left(x + \sqrt{x^2+1}\right) \]
Check at x equal to 2
Why: The formula gives the logarithm of two plus the root of five, which is 1.443635, and the hyperbolic sine of that number is 2.000000.
\[ \ln\big(2+\sqrt5\big) = 1.443635, \qquad \sinh(1.443635) = 2.000000 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 656 — Exercise 433
Follow the plan before following the algebra. You want y in terms of x, you have an equation containing the exponential of y and its reciprocal, and your route is to turn that into a polynomial.
The pivotal move is multiplying through by the exponential of y. That clears the reciprocal and leaves a quadratic in the exponential, which the quadratic formula solves. Everything before this step is arranging, and everything after it is tidying.
The rejection of the second root is not a technicality; it is where the mathematics happens. The root of x squared plus one is always bigger than the size of x, so the minus root is negative for every x, and an exponential is never negative. Saying that out loud is what makes the derivation a proof rather than a manipulation. Then check the answer at x equal to 2 and watch it return exactly 2.
Error analysis
Annotate
On: \( e^{y} = x - \sqrt{x^2+1} \;\Longrightarrow\; \sinh^{-1}x = \ln\left(x - \sqrt{x^2+1}\right) \)
Read the two lines and try to name the error before revealing the annotations. Notice that nothing has been computed incorrectly: the quadratic really does have two roots and this is one of them.
The error is an omission rather than a mistake. The step from the quadratic to the answer needs the extra fact that an exponential is positive, and that fact is what eliminates one root. Leaving it out does not produce a wrong number; it produces an expression that is not a number at all.
The one-second test in the last annotation is the habit to copy. Put a convenient value into any formula you are unsure of. At x equal to 0 the correct formula gives the logarithm of 1, which is 0, and the hyperbolic sine of 0 is 0. This formula asks for the logarithm of minus 1, and stops there.
Ranking
Put in order
Put the steps of the logarithmic formula for the inverse hyperbolic sine in order.
Why: The multiplication has to come before the quadratic, because the negative exponent is what stops it from being a polynomial. Rejecting the root has to come before the logarithm, because the logarithm of the rejected root is not defined at all.
Drag the five steps into order before checking. Two of the orderings are genuinely forced, and it is worth knowing which.
Multiplying through has to come before treating it as a quadratic, because until the reciprocal exponential is cleared the equation is not a polynomial at all. And rejecting the negative root has to come before taking the logarithm, because the logarithm of that root is not defined, so there is nothing to take.
The other steps could not sensibly be reordered either, but they follow naturally. If you can reproduce this sequence you can derive any of the six logarithmic formulas, not just this one.
Worked example
Implicit differentiation, and then the identity from Part 2 to remove the y.
Undo the inverse
Why: Apply the hyperbolic sine to both sides.
\[ y = \sinh^{-1}x \;\Longrightarrow\; \sinh y = x \]
Differentiate both sides with respect to x
Why: The left-hand side needs the chain rule, since y depends on x.
\[ \cosh y\,\frac{dy}{dx} = 1 \]
Solve for the derivative
Why: The hyperbolic cosine is never zero, so dividing is safe.
\[ \frac{dy}{dx} = \frac{1}{\cosh y} \]
Remove y with the identity
Why: The hyperbolic cosine is positive, so the square root takes the plus sign.
\[ \cosh y = \sqrt{1+\sinh^2 y} = \sqrt{1+x^2} \]
Put it together
Why: Only x is left.
\[ \frac{d}{dx}\sinh^{-1}x = \frac{1}{\sqrt{1+x^2}} \]
Check against the logarithmic formula
Why: Differentiating the logarithm found two slides ago gives the same function, which confirms both results at once.
\[ \frac{d}{dx}\ln\!\left(x+\sqrt{x^2+1}\right) = \frac{1 + \frac{x}{\sqrt{x^2+1}}}{x+\sqrt{x^2+1}} = \frac{1}{\sqrt{x^2+1}} \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 650 — the implicit differentiation before Table 6.4
This is the standard implicit-differentiation argument for any inverse, and you have seen it before for the inverse trigonometric functions. What makes it pleasant here is the last step.
After differentiating you have the derivative as one over the hyperbolic cosine of y, which is useless while y is still there. The fundamental identity converts the hyperbolic cosine into a root of one plus a squared hyperbolic sine, and the squared hyperbolic sine is just x squared. So y disappears entirely and the answer is algebraic.
The check closes the loop nicely. Differentiate the logarithmic formula from a few slides ago and you get the same function, which confirms both the formula and the derivative at once. Do that differentiation yourself: the numerator simplifies because the fraction inside the bracket shares a factor with the denominator.
Concept
The same implicit-differentiation argument, run on the other five, gives the rest of the table. Every one of them is algebraic: no hyperbolic function survives in any answer.
| f(x) | its derivative | valid where |
|---|---|---|
| sinh⁻¹ x | 1 / √(1 + x²) | all x |
| cosh⁻¹ x | 1 / √(x² − 1) | x greater than 1 |
| tanh⁻¹ x | 1 / (1 − x²) | |x| less than 1 |
| coth⁻¹ x | 1 / (1 − x²) | |x| greater than 1 |
| sech⁻¹ x | −1 / (x √(1 − x²)) | between 0 and 1 |
| csch⁻¹ x | −1 / (|x| √(1 + x²)) | x not zero |
Rows three and four carry the same expression. The two functions are not the same, though: their domains do not overlap at all, so when you integrate that expression you must decide from the values of x which antiderivative you are entitled to name.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, pp. 650-651 — Table 6.4 and the remark after it
Every entry in this table is algebraic. That is remarkable and easy to walk past: you start with a transcendental function, invert it, differentiate it, and out comes something with nothing but roots and squares in it.
The coincidence in rows three and four is the part to take seriously. The two expressions are identical, so the table alone cannot tell you which antiderivative to write when you integrate that expression. The domains settle it, and they do not overlap: one function lives strictly inside the interval from minus one to one, the other strictly outside it.
In practice that means an integral of one over one minus x squared has two possible answers and you must look at the values of x to choose. The book flags this explicitly, and it is the sort of detail that separates a complete solution from an almost-complete one.
Worked example
\[ \text{(a) } \frac{d}{dx}\left(\sinh^{-1}\frac{x}{3}\right) \qquad \text{(b) } \frac{d}{dx}\left(\tanh^{-1}x\right)^2 \]
Apply the formula with the chain rule in (a)
Why: The inside differentiates to one third.
\[ \frac{d}{dx}\sinh^{-1}\frac{x}{3} = \frac{1}{\sqrt{1+\frac{x^2}{9}}}\cdot\frac13 \]
Pull the three inside the root
Why: Three times the root of one plus x squared over nine is the root of nine plus x squared.
\[ = \frac{1}{3\sqrt{1+\frac{x^2}{9}}} = \frac{1}{\sqrt{9+x^2}} \]
Start (b) with the power rule
Why: Outer square first, with the inverse left alone.
\[ \frac{d}{dx}(\tanh^{-1}x)^2 = 2\tanh^{-1}x\cdot\frac{d}{dx}\tanh^{-1}x \]
Insert the inverse's derivative
Why: One over one minus x squared, from row three of the table.
\[ = \frac{2\tanh^{-1}x}{1-x^2} \]
Check (a) at x equal to zero
Why: The formula gives one third; independently, the derivative of the inverse hyperbolic sine at 0 is 1, scaled by the inside derivative of one third, which is also one third.
\[ \frac{1}{\sqrt{9+0}} = \frac13 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 651 — Example 6.49
Part (a) looks worse than it is. Apply the formula, multiply by the inside's derivative of one third, and then do the algebra that clears the fraction inside the radical.
That algebra step is worth watching. A three outside a square root can be taken inside as a nine, and nine times x squared over nine is x squared, so the whole thing collapses to the root of nine plus x squared. The result is much tidier than the line before it, and this simplification is expected in a final answer.
Part (b) is a power rule with an inverse inside, and it is the same shape as the reciprocal-square problem from Checkpoint 6.47. Differentiate the outer power, leave the inverse alone, and then multiply by its derivative from row three.
Worked example
\[ \text{(a) } \frac{d}{dx}\Big(\cosh^{-1}(3x)\Big) \qquad \text{(b) } \frac{d}{dx}\left(\coth^{-1}x\right)^3 \]
Apply the inverse hyperbolic cosine formula
Why: One over the root of the inside squared minus one, times the inside's derivative of three.
\[ \frac{d}{dx}\cosh^{-1}(3x) = \frac{3}{\sqrt{(3x)^2-1}} = \frac{3}{\sqrt{9x^2-1}} \]
Start (b) with the power rule
Why: Outer cube first.
\[ \frac{d}{dx}(\coth^{-1}x)^3 = 3(\coth^{-1}x)^2\cdot\frac{d}{dx}\coth^{-1}x \]
Insert the derivative
Why: Row four of the table, which reads the same as row three.
\[ = \frac{3(\coth^{-1}x)^2}{1-x^2} \]
Check the domain of (a)
Why: The answer needs nine x squared above one, and the inverse hyperbolic cosine needs three x at least one; both conditions say the same thing, so the formula and its function agree about where they live.
\[ 3x > 1 \iff x > \tfrac13 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 651 — Checkpoint 6.49
Both parts are the same two patterns as the worked example, so treat this as the repetition that makes them automatic. Part (a) is the formula plus a chain rule factor of three; part (b) is a power rule with the inverse inside.
The thing to notice in part (a) is that the three appears twice, once from the chain rule in the numerator and once inside the radical as a nine. Those are different threes and both belong there.
The check on domains is the habit this checkpoint is really building. An answer whose domain differs from its function's domain is a sign that something has gone wrong. Here both say x is bigger than a third, and the agreement is reassuring rather than accidental.
Concept
Read the inverse derivative table backwards and you get a short list of algebraic integrands with hyperbolic answers. Three radical shapes, and that is all.
Figure (svg): A mapping diagram: five integrand shapes in boxes on the left, each joined by an arrow to the inverse hyperbolic antiderivative it produces, on the right.
\[ \int\frac{du}{\sqrt{1+u^2}} = \sinh^{-1}u + C, \qquad \int\frac{du}{\sqrt{u^2-1}} = \cosh^{-1}u + C \]
\[ \int\frac{du}{1-u^2} = \begin{cases} \tanh^{-1}u + C & |u| < 1 \\ \coth^{-1}u + C & |u| > 1 \end{cases} \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 651 — the integral formulas before Example 6.50
This is the reverse of the previous table, and the mapping diagram is the version to keep. Three radical shapes on the left, three answers on the right, and two extra rows for the cases with an isolated factor outside the radical.
The diagnostic is the sign under the radical. A plus, and you have the inverse hyperbolic sine. A minus with the square first, and you have the inverse hyperbolic cosine. No radical at all, just a difference of squares in the denominator, and you have the inverse hyperbolic tangent or cotangent, chosen by domain.
The last two rows, with a lone factor outside the radical, both carry a minus sign in the answer. Those are the secant and cosecant cases, and the minus is the same minus that appears in their derivative rows.
Matching
Match the pairs
Why: Set u equal to x over 5 in the first three. A plus under the radical calls the inverse hyperbolic sine, a minus calls the inverse hyperbolic cosine, and no radical at all, just a difference of squares, calls the inverse hyperbolic tangent. Only the third keeps a leftover constant, because pulling 25 out of the denominator leaves a fifth outside as well as a fifth inside.
Pair all four before checking. Each left-hand item is one of the three radical shapes in disguise, hidden behind a constant.
The substitution is always the same: divide the variable by whatever makes the constant become one. Here that is five in the first three, and nothing at all in the fourth.
The third pair is the one that carries an extra constant. When you pull the twenty-five out of a difference of squares in a denominator, a factor of twenty-five comes out, and the substitution only puts back a five, so a fifth is left in front. Compare that with the first two, where the factor from the radical and the factor from the differential cancel exactly. Knowing when a constant survives and when it cancels is what this exercise is for.
Worked example
\[ \text{(a) } \int\frac{dx}{\sqrt{4x^2-1}} \qquad \text{(b) } \int\frac{dx}{2x\sqrt{1-9x^2}} \]
Substitute in (a)
Why: Let u be twice x, so four x squared becomes u squared.
\[ u = 2x, \quad du = 2\,dx \]
Rewrite the integral
Why: Half a du on top, the root of u squared minus one below.
\[ \int\frac{dx}{\sqrt{4x^2-1}} = \int\frac{du}{2\sqrt{u^2-1}} \]
Recognise the form
Why: A minus under the radical calls the inverse hyperbolic cosine.
\[ = \tfrac12\cosh^{-1}u + C = \tfrac12\cosh^{-1}(2x) + C \]
Substitute in (b)
Why: Let u be three x; then two x is two thirds of u.
\[ u = 3x, \quad du = 3\,dx, \quad 2x = \tfrac23 u \]
Rewrite
Why: A third from the differential and three halves from the denominator leave one half in front.
\[ \int\frac{dx}{2x\sqrt{1-9x^2}} = \tfrac12\int\frac{du}{u\sqrt{1-u^2}} \]
Recognise the form
Why: A single u outside the radical calls the inverse hyperbolic secant, which carries a minus sign.
\[ = -\tfrac12\operatorname{sech}^{-1}|u| + C = -\tfrac12\operatorname{sech}^{-1}|3x| + C \]
Check (a) by differentiating
Why: The derivative of half the inverse hyperbolic cosine of two x returns the integrand; numerically, the integral from 1 to 2 is 0.373240, and the antiderivative changes by 0.373240 over the same interval.
\[ \frac{d}{dx}\left[\tfrac12\cosh^{-1}(2x)\right] = \tfrac12\cdot\frac{2}{\sqrt{4x^2-1}} = \frac{1}{\sqrt{4x^2-1}} \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 652 — Example 6.50
Both parts are linear substitutions, and in both the point is to make the integrand match a row of the table exactly before naming any answer.
In part (a) notice how the constants work out. The substitution puts a two in the differential, which becomes a half in front, and the radical needs no extra factor because four x squared is exactly u squared. So the answer carries one half.
Part (b) has the extra subtlety. There is a lone factor outside the radical, which means the secant row rather than the sine or cosine row, and that row carries a minus sign and an absolute value. Substituting turns the two x into two thirds of u, and combined with the third from the differential it leaves a half in front. Track those constants carefully; two of them nearly cancel and you have to be sure which way.
Worked example
\[ \text{(a) } \int\frac{dx}{\sqrt{x^2-4}},\; x>2 \qquad \text{(b) } \int\frac{dx}{\sqrt{1-e^{2x}}} \]
Factor the radicand in (a)
Why: Pull the four out so the shape matches u squared minus one.
\[ \sqrt{x^2-4} = 2\sqrt{\left(\tfrac{x}{2}\right)^2 - 1} \]
Substitute
Why: Let u be half of x, so dx is twice du.
\[ u = \tfrac{x}{2}, \quad dx = 2\,du \]
Integrate
Why: The twos cancel and the inverse hyperbolic cosine appears.
\[ \int\frac{2\,du}{2\sqrt{u^2-1}} = \cosh^{-1}u + C = \cosh^{-1}\!\left(\tfrac{x}{2}\right) + C \]
Substitute in (b)
Why: Let u be the exponential of x; then dx is du over u.
\[ u = e^{x}, \quad du = e^{x}dx, \quad dx = \frac{du}{u} \]
Rewrite
Why: The exponential of twice x is u squared, and the leftover u goes outside the radical.
\[ \int\frac{dx}{\sqrt{1-e^{2x}}} = \int\frac{du}{u\sqrt{1-u^2}} \]
Integrate
Why: The inverse hyperbolic secant again, with its minus sign.
\[ = -\operatorname{sech}^{-1}|u| + C = -\operatorname{sech}^{-1}(e^{x}) + C \]
Check (b) numerically
Why: Between x equal to minus 2 and minus 1 a numerical integral gives 1.031082, and the antiderivative changes by 1.031082 over the same interval; note that the exponential is below one there, which is exactly where the inverse hyperbolic secant is defined.
\[ \Big[-\operatorname{sech}^{-1}(e^{x})\Big]_{-2}^{-1} = 1.031082 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 652 — Checkpoint 6.50
Part (a) begins with a move you will use constantly in the next course: factor the constant out of the radicand so that what remains has the standard shape. A four inside a square root comes out as a two.
Part (b) is the same substitution you planned two slides ago on Exercise 417, with a minus instead of a plus under the root. The numerator is not visibly the derivative of the exponential, but writing the differential as du over u supplies exactly the lone factor that the secant row needs.
The check at the end does two jobs. It confirms the antiderivative numerically, and it points out that the exponential is below one whenever x is negative, which is precisely the domain on which the inverse hyperbolic secant is defined. An antiderivative that is only valid on part of the line is still a correct antiderivative, but you should know which part.
Section
Part 5
Concept
Figure (svg): Two upright posts on a ground line with a chain hanging between their tops; the chain dips to its lowest point midway, and a dashed vertical segment marks the sag from the post height down to that lowest point.
catenary — The curve taken by a flexible cable of uniform density hanging under its own weight alone, with no other load. Its equation is a hyperbolic cosine.
\[ y = a\cosh\!\left(\frac{x}{a}\right) \]
Power lines, chains between posts and the threads of a spider's web are all catenaries. The one number a is the height of the lowest point above the origin, and it is set by the ratio of the horizontal tension to the weight carried per unit of length.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 653 — the catenary and Figures 6.83 and 6.84
This is the application the section is built toward, and it is a genuinely surprising result. Nothing about a chain suggests exponentials, and yet the shape it takes is a hyperbolic cosine exactly, not approximately.
The reason, which Exercises 422 and 423 work out, is that the slope of the cable at a point is proportional to the weight of cable hanging below that point, and that weight is proportional to the arc length, which brings in a square root of one plus the slope squared. The differential equation that results is solved by the hyperbolic sine, and integrating once more gives the hyperbolic cosine.
Look at the picture and fix the meaning of the one parameter. It is the height of the lowest point of the chain above the origin, and physically it is the horizontal tension divided by the weight per unit length. Pull the chain tighter and that number grows; let it go slack and it shrinks.
Notation
Annotate
On: \( y = a\cosh\!\left(\frac{x}{a}\right) \)
The annotations are all about the single parameter, because that is the only thing you can change. Notice that it appears twice, once as a vertical scale and once as a horizontal one, and always the same value.
That double appearance has a consequence worth stating. Every catenary is the same curve, seen at a different magnification. There is no family of different shapes here, only one shape at different sizes, which is not true of parabolas or of most curves you have met.
The last annotation covers what real problems look like. Exercises 425 and 426 both subtract a constant so the chain hangs at a convenient height. That subtraction shifts the whole picture down and changes neither the slope at any point nor the length of the cable, so you can ignore it while doing calculus and put it back at the end.
Picture it
Figure (svg): Two curves on the same axes from minus 15 to 15: the catenary ten times the hyperbolic cosine of x over ten, and the dashed parabola ten plus x squared over twenty. They are indistinguishable near the middle and separate near the ends, where the marked heights are 23.52 and 21.25.
Galileo thought a hanging chain was a parabola, and the picture shows why the mistake lasted. Expanding the hyperbolic cosine in powers of x gives a plus x squared over twice a, plus a fourth-power term and smaller ones; the parabola is what you get by stopping early.
\[ a\cosh\!\left(\frac{x}{a}\right) = a + \frac{x^2}{2a} + \frac{x^4}{24a^3} + \cdots \]
Galileo asserted that a hanging chain is a parabola, and it took most of a century and the invention of calculus to show that it is not. The picture explains why nobody caught the error by eye.
The expansion on the slide is the reason. The hyperbolic cosine agrees with the parabola in its constant and its squared term, and the first disagreement is at the fourth power, which is tiny near the middle. Across the central portion of a span the two curves are within a few hundredths of each other.
By the ends of this thirty-foot span the gap has opened to more than two feet, and it widens quickly from there. So the parabola is a good approximation for a shallow sag and a bad one for a deep one, which is exactly what the fourth-power term predicts.
Tweak it
Parameter explorer
The curve is a times the hyperbolic cosine of x over a, for a span of thirty units. Slide a from tight to slack. What happens to the lowest point, and what happens to the sag between the middle and the ends?
\[ y = {a}\cosh\!\left(\frac{x}{{a}}\right) \]
Start at the middle and move the slider slowly toward the small end. Watch two things at once: the height of the lowest point, and how much the curve climbs between the middle and the edge of the window.
At large values the curve is nearly a flat line high above the axis; the cable is tight and barely sags. At small values the lowest point drops toward the axis and the ends shoot up steeply; the cable is slack and sags deeply. One number does both, because it scales the curve horizontally and vertically together.
Now connect this to the physics. Large parameter means large horizontal tension relative to weight. That is why a tightly strung power line looks almost straight and a loose chain between two fence posts hangs in a deep curve, and why you cannot make a real cable perfectly horizontal no matter how hard you pull.
Worked example
A cable hangs in the shape below, with x measured in feet. Find its length.
\[ f(x) = 10\cosh\!\left(\frac{x}{10}\right), \quad -15 \le x \le 15 \]
Write the arc-length formula
Why: From Section 6.4.
\[ L = \int_a^b\sqrt{1 + [f'(x)]^2}\,dx \]
Differentiate the catenary
Why: The ten in front and the one tenth inside cancel exactly.
\[ f'(x) = 10\sinh\!\left(\frac{x}{10}\right)\cdot\frac{1}{10} = \sinh\!\left(\frac{x}{10}\right) \]
Substitute into the formula
Why: Square the derivative and put in the limits.
\[ L = \int_{-15}^{15}\sqrt{1 + \sinh^2\!\left(\frac{x}{10}\right)}\,dx \]
Use the identity
Why: One plus a squared hyperbolic sine is a squared hyperbolic cosine, and the hyperbolic cosine is positive, so the radical disappears completely.
\[ L = \int_{-15}^{15}\cosh\!\left(\frac{x}{10}\right)dx \]
Integrate
Why: The antiderivative of the hyperbolic cosine of x over ten is ten times the hyperbolic sine of x over ten.
\[ L = 10\sinh\!\left(\frac{x}{10}\right)\bigg|_{-15}^{15} = 10\left[\sinh\tfrac32 - \sinh\!\left(-\tfrac32\right)\right] \]
Figure (svg): The catenary from minus 15 to 15 with eight straight chords drawn across it, and a right triangle on one chord whose horizontal leg is labelled dx and vertical leg dy; a double arrow below marks the 30-foot span.
Use oddness and evaluate
Why: The hyperbolic sine is odd, so the two terms add instead of cancelling.
\[ L = 20\sinh\tfrac32 \approx 42.586 \text{ ft} \]
Check against two crude paths
Why: Joining each end to the lowest point gives a two-segment path of length 40.393, and any sagging cable is longer than that; going straight across and then straight down gives 57.048, which it cannot exceed. The answer sits between them, closer to the taut bound, as a shallow sag should be.
\[ 40.393 < 42.586 < 57.048 \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 654 — Example 6.51
This is the model problem of the section, and the whole reason it works is the identity from Part 2. Follow the fourth step closely: one plus a squared hyperbolic sine is a squared hyperbolic cosine, so the square root disappears entirely and you are left integrating a hyperbolic cosine.
That almost never happens. Arc-length integrands normally contain a square root that no substitution removes, which is why most arc lengths in this course are computed numerically. The catenary is the famous exception, and the reason is that it was built out of exactly the function whose square plus one is a perfect square.
The figure shows what the integral means: the sum of the tiny hypotenuses of triangles with horizontal leg dx and vertical leg dy. The check at the end brackets the answer with two paths you can measure by hand, one shorter than any sagging cable and one longer. Getting 42.586 between 40.393 and 57.048 is not a proof, but it is a strong signal that no arithmetic went astray.
Estimation
\[ f(x) = 15\cosh\!\left(\frac{x}{15}\right), \quad -20 \le x \le 20 \]
Predict first
The span is now 40 feet instead of 30. Estimate the cable's length before computing it.
Correct: About 53 ft
Why: The same three steps give the derivative as the hyperbolic sine of x over 15, the radical collapses to the hyperbolic cosine, and the integral is 15 times the hyperbolic sine of x over 15, evaluated from minus 20 to 20. That is 30 times the hyperbolic sine of four thirds, which is 52.951 feet. The span grew by a third and the length grew by about a quarter, because this cable is relatively tighter: the ratio of half-span to the parameter fell from 1.5 to about 1.33.
\[ L = 15\sinh\!\left(\frac{x}{15}\right)\bigg|_{-20}^{20} = 30\sinh\tfrac43 \approx 52.951 \text{ ft} \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 654 — Checkpoint 6.51
Guess before computing. The span has grown by a third, from thirty feet to forty, so a first instinct might be that the length grows by a third too, from about 43 to about 57.
It does not, and the reason is worth understanding. The parameter also grew, from ten to fifteen, so relative to its own scale this cable is tighter, not slacker. The ratio of half-span to parameter fell from one and a half to about one and a third, which means proportionally less sag and proportionally less extra length.
The computation is the same four steps as the worked example. Differentiate, use the identity to kill the radical, integrate to get the parameter times the hyperbolic sine, and double it using oddness. The answer is a little under 53 feet.
Worked example
A high-voltage power line hangs as the catenary below. Compare the area under it with the length of the line itself.
\[ y = 10\cosh\!\left(\frac{x}{10}\right), \quad -15 \le x \le 15 \]
Integrate for the area
Why: The antiderivative of ten times the hyperbolic cosine of x over ten is one hundred times the hyperbolic sine of x over ten.
\[ A = \int_{-15}^{15}10\cosh\!\left(\frac{x}{10}\right)dx = 100\sinh\!\left(\frac{x}{10}\right)\bigg|_{-15}^{15} \]
Evaluate
Why: Oddness doubles the value at the top limit.
\[ A = 200\sinh\tfrac32 \approx 425.856 \]
Recall the length
Why: The previous example, with the same cable and the same limits.
\[ L = 20\sinh\tfrac32 \approx 42.586 \]
Figure (svg): The region under the catenary from minus 15 to 15 shaded, with its area of 425.86 square feet marked inside and the length of its curved top edge, 42.59 feet, marked above.
Divide
Why: The hyperbolic sines cancel exactly, leaving a pure number.
\[ \frac{A}{L} = \frac{200\sinh\frac32}{20\sinh\frac32} = 10 \]
Check that the limits did not matter
Why: Redo it over any symmetric interval from minus b to b: the area is 200 times the hyperbolic sine of b over ten and the length is 20 times the same thing, so the ratio is always ten, which is the parameter a in the formula.
\[ \frac{A}{L} = \frac{200\sinh(b/10)}{20\sinh(b/10)} = 10 = a \]
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 655 — Exercise 427
This exercise ends with the words what do you notice, which is an invitation to look past the number. Do the two integrals, divide, and watch what cancels.
Both the area and the length are constant multiples of the same hyperbolic sine, so when you divide, the hyperbolic sine goes away entirely and a pure number is left. That number is ten, which is the parameter in the formula.
The last step shows that the limits never mattered. Take any symmetric interval you like and the same cancellation happens, so the ratio is always the parameter. There is a nice geometric reading: the area under a catenary is exactly the parameter times the length of the curve, as if the region were a strip of constant width laid along the cable.
Socratic
\[ y = 2\cosh\!\left(\frac{x}{2}\right) - 1, \quad \text{posts 2 m apart} \]
Discussion prompt
The posts are two metres apart, so the chain runs from x equal to minus 1 to x equal to 1. Find the slope at the left-hand post, and say what the minus 1 in the formula does to your answer.
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 655 — Exercise 425
Answer before revealing. Two things need deciding: what the interval is, and whether the subtracted constant matters.
The posts are two metres apart and the formula is symmetric about the vertical axis, so the chain runs from minus one to one and the left post is at minus one. The subtracted constant lowers the whole chain and has no effect on any derivative, so you can ignore it entirely for this question.
Differentiating gives the hyperbolic sine of half of x, and at minus one that is the hyperbolic sine of minus a half, about minus 0.521. The sign is the sanity check: moving left to right the chain is still descending toward its lowest point at the middle, so a negative slope is exactly what the picture demands.
Counterexample
Discussion prompt
The main cable of a suspension bridge really is a parabola, not a catenary, and this is not an approximation. What is different about how it is loaded, and which assumption in the catenary derivation fails?
OpenStax Calculus Volume 1, §6.9 Calculus of the Hyperbolic Functions §6.9, p. 655 — Exercises 422 and 423
Think about this before revealing, because it is a rare case where two famous curves are both correct answers to nearly the same question.
The catenary derivation assumes the only load is the cable's own weight, spread evenly along the cable itself. That is what puts the arc-length element, and therefore the square root, into the differential equation, and the square root is what produces a hyperbolic function.
A suspension bridge's main cable carries a roadway hung beneath it, and the roadway's weight is spread evenly along the horizontal, not along the cable. The square root never appears, the slope grows linearly, and integrating gives a quadratic. So the parabola is not an approximation to the bridge cable; it is the exact answer to a different loading problem. If you ever see a cable with nothing hanging from it, expect a catenary.
Section
Part 6
Pattern
Figure (svg): A flow diagram: a hyperbolic expression branches four ways, to differentiating with Table 6.2, to integrating by reading that table backwards, to a radical that calls an inverse hyperbolic form, and to a hanging cable whose arc length uses the identity.
This is the order to work in when a hyperbolic function turns up, and the diagram sorts problems into the four kinds this section contains.
The most valuable line is the second. Before you do anything with a radical, look at the sign under it: a plus calls the inverse hyperbolic sine, a minus with the square first calls the inverse hyperbolic cosine, and no radical at all calls the inverse hyperbolic tangent or cotangent. That single glance decides most of these problems.
The last line is your escape hatch. If you cannot remember which identity you need or which sign goes where, write everything in exponentials. Every result in this section is two or three lines of algebra away from the definitions, which is not true of the trigonometric material it resembles.
Check
Check your understanding
Exactly one of these derivative formulas is correct. Which?
Answer: C
Why: Differentiating the odd half of the exponential term by term gives the even half, with no sign change at all: the derivative of sinh is cosh. This is the one row of the table that matches trigonometry exactly, and it is also the row people most often doubt.
Take each option and ask where it came from. Three of them are circular formulas with the letter h inserted, and each has picked up or lost a sign in the process.
The correct one is the sine row, which is the one row that transfers unchanged. If you hesitated over it, that hesitation is worth noting: people so often expect the hyperbolic table to differ from the circular one that they start doubting the rows that match.
The most dangerous distractor is the first, because it is the natural guess. Differentiating the even half of the exponential gives the odd half with a plus, and the two-line derivation is the only defence you should rely on.
Check
Check your understanding
Which of these is an identity, true for every real x?
Answer: B
Why: Adding the two squares leaves twice the forward exponential plus twice the reflected one, over four, which is the definition of cosh with twice x inside. Subtracting instead is what gives 1.
You can eliminate two options in a few seconds by testing at x equal to 1, where the hyperbolic cosine is about 1.543 and the hyperbolic sine about 1.175.
The sum of their squares is about 3.762, which rules out the first option immediately and matches the hyperbolic cosine of 2. The difference of the squares is 1, which rules out the third option, since the hyperbolic cosine of 2 is not 1.
The fourth option would force the two squares to be equal, which never happens because their difference is always 1. Numerical testing is the fastest way to handle any identity question, and it costs nothing.
Check
Check your understanding
Which is an antiderivative of 1 / √(x² + 16)?
Answer: A
Why: Substitute u equal to x over 4: the radical becomes 4 times the root of u squared plus 1, and dx is 4 du, so the fours cancel and the integral is the inverse hyperbolic sine of u. A plus under the radical always calls the inverse hyperbolic sine.
First read the sign under the radical: it is a plus, so before doing any algebra you know the answer involves the inverse hyperbolic sine, and two options disappear.
The choice between the two survivors is about the constant. Substituting a quarter of x gives a four from the radical and a four from the differential, and they cancel, so nothing is left in front. The extra quarter in one option is the trap.
If you picked it, redo the substitution writing every factor of four explicitly rather than in your head. That is where the cancellation becomes visible, and it is the same pattern as part (a) of Example 6.50.
Explain it to yourself
Discussion prompt
In three or four sentences, explain to someone who has just met these functions why they are named after the sine and the cosine, and name the single algebraic change that separates the two families.
Write your answer before revealing, and aim it at someone who has never seen these functions. If you cannot explain the parallel in plain words, you are still holding the two families as separate lists of formulas rather than as one idea.
A complete answer names the matching structure and the single difference. The structure is the pair of functions, the identity relating their squares, the derivative rules that turn each into the other, and the sector-area meaning of the parameter. The difference is one sign.
The best answers also say what follows from that sign: a hyperbola instead of a circle, no period at all, a missing minus in one derivative, and logarithmic formulas for the inverses where the circular family has none.
Exit ticket
\[ y = 5\cosh\!\left(\frac{x}{5}\right), \quad -10 \le x \le 10 \]
Discussion prompt
For this cable, find how far it sags from the posts down to its lowest point, and find its total length. Then say which identity did the work.
This combines the whole of Part 5 in one problem. Find the height at the ends, subtract the height at the middle for the sag, then run the four-step arc-length computation.
The sag is the height at the end minus the height at the lowest point. At the ends the input to the hyperbolic cosine is two, so the height is five times 3.762, about 18.811, and the lowest point sits at five. The sag is about 13.811 feet.
The length comes out as ten times the hyperbolic sine of two, about 36.269 feet, over a span of twenty. That is a deep sag, as the numbers say: the cable is more than eighty per cent longer than the straight distance between its supports. And the identity that made the integral possible is the same one every time: one plus a squared hyperbolic sine is a squared hyperbolic cosine.
Recap
| what | the formula | the thing to remember |
|---|---|---|
| the definitions | cosh x = (eˣ + e⁻ˣ)/2, sinh x = (eˣ − e⁻ˣ)/2 | the even and odd halves of the exponential |
| the identity | cosh² x − sinh² x = 1 | a minus, so a hyperbola, and no period anywhere |
| the derivatives | sinh′ = cosh, cosh′ = sinh | the only missing minus sign in the whole table |
| the inverses | sinh⁻¹ x = ln(x + √(x² + 1)) | solve a quadratic in eʸ and reject the negative root |
| the inverse integrals | 1/√(1 + u²), 1/√(u² − 1), 1/(1 − u²) | three radical shapes, three answers |
| the catenary | y = a cosh(x/a) | a hanging chain, and its arc length is 2a sinh(b/a) |
\[ 1 + \sinh^2 x = \cosh^2 x \]
That one line is the workhorse. It removes the radical from every arc-length integral in this section, and it is the reason the catenary is the one curve whose length you can write down without effort.
Stewart, Calculus: Early Transcendentals 8e, §3.11 Hyperbolic Functions §3.11, pp. 259-265 — the same material in Stewart
Six lines, and the first two carry the other four. The definitions are the even and odd halves of the exponential, and the identity subtracts where Pythagoras adds. If you can write both from memory you can rebuild everything else in the section in a few minutes of algebra.
For calculus, the one thing to overlearn is that the hyperbolic cosine differentiates to the hyperbolic sine with a plus sign. That is the only place the hyperbolic table differs from the circular one, and it is the error examiners expect to find.
The inverse functions matter for two different reasons. Their logarithmic formulas are elegant and worth seeing once, but their derivatives are what you will actually use, because they turn a short list of purely algebraic integrands into integrals you can do. Section 3.7 of Calculus II will lean on that list heavily when it takes up trigonometric and hyperbolic substitution in earnest.
Want this taught 1-on-1? Alexander tutors Calculus I — $55/session, free consultation.